Algorithms Discover Build Order From Hell

Humans: Still a pathetic creatures of meat and bone.

This is mad. Over on StarCraft 2 forum Teamliquid, a poster who goes by Lomilar has been talking about a program he’s coded called EvolutionChamber. It uses genetic algorithms to find powerful build orders, meaning his program takes a population of build orders, kills off the useless ones, and has the most successful ones reproduce asexually to create a new population, which tests itself again, and so on. I’m taking all this from this blog post by programmer Louis Brandy, wherein he breaks down what Lomilar’s done so that lay folk can understand it.

EvolutionChamber’s already come up with one ludicrous build order, which I’ve posted beneath the jump.

10 extractor-trick to 11
11 overlord
11 spawning pool
15 extractor
16 queen (stop drones here)
18 overlord
18 roach warren
17 overlord (yes, two)
spawn-larva on queen when she pops
roach x7

Which will get you a 7 roach rush in less time than anybody thought possible. “Extractor-trick” refers to telling a drone to build an extractor, building a drone, then cancelling the extractor’s construction to get the drone back, making your population of drones 11 out of a maximum of 10.

To summarise Louis’ blog post, this is interesting because:

(1) This is such a devastating build order that for a while the discussion on centred around whether Protoss could survive it even if they knew it was coming. Turns out they can, but only with very specific counter-strategies or if the person they’re playing fails to fulfil the potential of the build order.

(2) Despite its effectiveness, this build order wasn’t known to the SC2 community. It’s quickly become labelled as a “cheese build,” though, meaning… well, it works.

(3) The extractor trick used by the program has been tested by real players, and was thought to be “economically inferior” compared to just buying an overlord and increasing your population cap. However, using this trick the program created a build order that made 7 roaches in less time than the standard roach rush strategy produces 5.

(4) …and yet while the strategy neglects to buy an overlord to begin with, it later produces two at once, far overshooting the population cap the strategy requires but allowing it to squeeze those 2 extra roaches into the rush. It’s all counter-intuitive, but the result is inarguable. Louis says that “This is the type of non-obvious optimization that genetic algorithms excel at.”

Truly, we are living in the future. Here’s a tutorial on how to do the 7 roach rush.


  1. ShineDog says:

    A textbook example of why I do not like Starcraft.

    • wogzi says:

      Because other people really like it and it makes it hard for you to play and get any real recognition at just the beginner’s level?

      Or is it because you think Zerg is OP?

      In either case, you craaaazy!

    • TotalBiscuit says:

      Yeah, games with that degree of depth and metagame suck. Let’s play something simpler and less interesting.

    • Londinistan says:

      A little hurt are we TotalBiscuit?

    • TotalBiscuit says:

      That’s a totally logical leap to make, obviously.

      Ask me which other criticisms of consumer products I find hurtful. A particular sorespot would be microwave ovens.

    • ChampionHyena says:

      TotalBiscuit’s just mad because he sucks at StarCraft II.

      No, wait, I spelled that wrong. He Sucks At StarCraft II(TM).

      (ETA: Keep up the good work, o’course)

    • Alikchi says:

      I think what ShineDog is getting at is: how formulaic and predictable does an RTS have to be for newsposts about it to be literally formulas?

    • TotalBiscuit says:

      You can turn any game into that if you break it down into it’s basic elements.

      The fallacy is assuming that’s all there is to it and that there are no counters and no skill involved in the proper execution of said optimal strategy.

      This build for instance, which people seem to claim is the ultimate build, can be scouted and countered.

    • Dave says:

      @Alikchi: Chess has opening moves, and Go has joseki.

    • Tetragrammaton says:

      “You can turn any game into that if you break it down into it’s basic elements. ”

      Except for Men of War. A game for real men (of war)

    • Starky says:

      I think Chess is the correct comparison here – Math nerds have for YEARS been using computer algorithms to calculate the most optimal openings, counters and moves – because that early game is so important.
      Later in the game the possibilities become much more varied and open up many different strategies, opportunities and variation.

      So you might gat 10 games of chess in a row where the first 10-20 moves are exactly the same, then they branch out from there.

      Starcraft 2 is the same.

      hell ANY multilayer competitive RTS, or non-random skill based game (like chess) is the same, there are optimal beginnings, and people who want to be good at the game learn them, and their counters, and use them.

    • Hidden_7 says:

      The issue with non-random games, that is, games without any random elements, is that they can all, with enough computing power, be “solved.” Consider how if you know what you’re doing even a little bit in Tic-Tac-Toe you needn’t ever lose. Probably most of the time it will end in ties. Put another way, a computer will never ever lose in Tic Tac Toe, unless it’s programmed to, or it breaks.

      Obviously that’s a very simple game, but the principle is the same for every single game that is determined in that there are no random elements. The computing power required just goes up and up the more complex the game is. Chess will eventually be solved. Meaning that at every point for every board position there is the optimal move that will lead you toward the best possible outcome. Seeing as how with a lot of these sorts of games a huge factor is who moves first, it may be that depending on which colour you are, you can’t win. Essentially reducing the game down to a coin toss to see who wins (plays first).

      Obviously that’s a ways off, and the optimal decision tree will probably be so large and complex that for a human to follow it they’d need to have a big reference book with them on hand to make sure they do it right. But the more you make a game like chess in that it has no random elements, the more you are actually turning it into a puzzle that has a solution. Which is fine, it’s not no-skill, no thought, since obviously it requires a lot of work to find those solutions, but you are just moving the skill to another place. Once the algorithm is found, however, the actual playing the game becomes somewhat pointless.

    • Hidden_7 says:

      I should clarify here that SC2 is not like Chess in that it’s not turn based. The ability to play a particular strategy requires more than simply knowing the strategy, as in, say, Chess. Since SC2 is real time and thus requires all sorts of action-game like twitch skills to effectively play it will never really have a solution the way they are talked about in game theory. The same way that say, TF2 doesn’t and will never really have a solution. It sort of does in that you could tell someone to just be a sniper and shoot everyone in the head, but executing on that strategy requires so much skill that the strategy (more of a tactic, really) becomes somewhat secondary to optimizing your outcome.

    • Dean says:

      This is also why I hate SC2 multiplayer.

      It’s also why I hate Chess too though, to be fair. Having to work out openings and think ahead that far is really unappealing to a lot of us.

    • Gap Gen says:

      They argue with umpires
      They cheer when they’ve won
      And they practice beforehand
      Which ruins the fun

    • Nalano says:

      Depth? Sure, why not.

      Metagame? Definitely.

      A fun experience? Not to me.

      This reminds me as to why I could never – ever – get into fighting games. It’s not that it isn’t noob-friendly. FPSs aren’t noob-friendly and I love those. It’s not that it blends strategy and twitch-based mechanics. Any team-based FPS is basically that.

      It’s simply that, with the existence of an optimal build/strategy, the game devolves into finding and exploiting the flavor-of-the-month strategy, which may not – and usually is not – my preferred style of play.

      In an FPS, I can pull out my SMG and kill people with my SMG. I can pull out my sniper rifle and kill people with my sniper rifle. Obviously some are better than others for certain circumstances, but FPSs largely allow me to dictate my own circumstances: If I’m carrying a shotgun and I chance upon a sniper-friendly zone, I don’t go there. If I’m a sniper and a guy ducks into a tight, windy corridor, I don’t follow him.

      Simple as that.

      My priority system must be skewed: I play competitive games to have fun first, and win second. If my win ratio is positive, yay! But nobody gives a flying shit about my win ratio unless it somehow adversely affects theirs, and if that’s the first thing they care about, I know that, ten minutes into the game, I’m gonna have to listen to them start bitching and moaning.

      And if that’s the case, what the fuck’s the point of playing?

    • pneuma08 says:

      @Hidden_7: Counterexample: Rock-Paper-Scissors. That game has no random elements and it cannot be solved. Your opponent simply makes a choice and so do you.

      Chess falls into the same category, actually. If you know someone is going to play any specific “ultimate” strategy, you can anticipate it and counter it. Much of the early game is spent trying to find out what your opponent is doing and how to appropriately react.

    • Xirvus.rei says:

      Roach nerf incoming.

      But on a more serious note:
      You don’t know what you’re missing out on.

      Sure, there might be an emphasis about “build orders”, but that’s hardly the focus. A terrible player may try to learn build orders off by heart, and go with that, but they’ll still be terrible. A great player may have a basic build order in their mind, and adjust off that.

      It takes literally a minute to learn a build order. 10depot, 12rax, 13gas, 15oc, k done. That’s IT. I’m a diamond 1v1 player, and I play by ear; I don’t need a build-order to make me gosu. It’s how I play AFTER I’ve begun the basic match that is what really matters. It’s how I REACT to what’s happening in the game. If you think that SC2 is all about memorising the perfect build, then you’re completely wrong, to put it simply.

      The beauty of SC2 is its depth. And the depth of SC2 is NOT about just “memorising build-orders”. Not only is it not even THAT HARD to memorise a straightforward build order, it’s not what is going to make you great. It’s a matter of executing, reacting, thinking off the wall, reacting, and executing. A build-order is only going to get you so far — i.e. not very.

      A build order is simply there to make sure you don’t build stupid units/buildings in the wrong order, in the first three to five minutes of the match. It’s not a “memorise what you’re going to do for the first 15 minutes of each game”. Hell no. If you did that, you’d never be promoted out of lolbronze league.

      Seriously, you don’t don’t know what you’re missing out on.

      And if you think that memorising/being accustomed to a very straightforward build order is too much for you, then hey, maybe you should stay away from SC2 after all.

    • Nalano says:

      That reminds me:

      There’s a fundamental difference in my eyes between SC2 and, say, chess.

      In chess, I do not attribute a value to any one piece for anything more than its strategic worth. I don’t like the rook more than the bishop.

      In an RTS, I like basic infantry, and I like tanks. Any winning strategy that doesn’t utilize hordes of infantry and tanks is not going to be fun for me, and being forced to either use said strategy or fail is not a rewarding experience for me.

      I like CoH because, when all is said and done, thanks to heavy use of soft-counters and micro-friendly squads, pretty much every strategy uses hordes of infantry and tanks. So in that sense instead of rock-paper-scissors, it’s more like rock-paper-scissors-shotgun, and rock is required.

    • Matthew Whittingham says:

      In an annoyingly pedantic reply to Hidden7 – his analogy breaks down. Technically there is no absolutely random element in a videgame, because all games use unambiguous algorythmic steps in their coding. All you can mean by random – quite rightly – is an unknown quantity: I don’t know if this will happen or this will happen, or what is happening over there. Chess is not random in this sense because the game is transparent; every move made by the other player is instantly telegraphed, and you have plenty of time to react without the situation changing. This is not the case for an RTS, where you not only cannot pay attention to every point on the battlefield, but there is also a fog of war, and you cannot know what items are in production. Furthermore, given that it is not turn based, much comes down to reaction times in thinking and enacting those thoughts. I don’t think an RTS like starcraft will ever be ‘solved’.

    • The Hon. Reverend Fred Gherkin says:

      Total Annihilation still has more depth than StarCraft II, more than a decade after its release. Deal with it, nerds.

    • Hidden_7 says:

      @ Matthew Wittingham

      You’re right of course, I should have been clearer with what I meant when I said “random.” It is basically that, actions that have unknown outcomes. When I move a pawn forward a spot, I know what will happen. It will move forward a spot. When I take a piece with that same pawn, I know the piece will be taken. When I roll a die, flip a coin, draw a card, I don’t know what the outcome will be. You can’t have a solved strategy for a game like that, or rather, you could, but it would need to account for always rolling the worst possible number / picking the worst possible card. It’s probably pretty rare that a game features dice rolling / card drawing and yet the effect on the game is such that a player can always win / draw despite always drawing the worst possible result. Also I granted that SC2 won’t be solved because of its real time actiony nature. You can have an optimal strategy but without the talent to execute it then you won’t win. Similar to how you can’t solve sports.


      Correct about Rock Paper Scissors (almost wrote shotgun there), although that was because I didn’t make myself clear enough when I used the term random. Because the game has only one move, both players make it at the same time, and the strength of all moves are equal, it basically consists of entirely what I mean when I say random elements. It either can’t be solved, or is very trivially solved by anyone who plays. I.e. pick rock, always. Or paper, or scissors, or mix it up. It doesn’t matter because every move has the exact same odds of winning and there is only ever one possible state you are playing in.

      However, I will reiterate my point about Chess though, to make it clear. The idea that a game is solvable means that no, the “ultimate” strategy cannot be countered. Because it takes into account every single possible move you could do at any point throughout the game. Consider Tic-Tac-Toe again. Most people who are actually posting on this site (i.e. not young children) know enough about this game such that if they don’t make foolish mistakes they will never be beaten. Ever. You can’t go into a game of TTT (noughts and crosses?) knowing your opponent plans to play the “ultimate” strategy and just counter it. Because their strategy accounts for every move you could make. Consider a different way; it is impossible to beat a computer at this game. There’s no counter to it, there’s no tricky way of playing that accounts for the way they are playing, because the way they are playing is simply to play to win the best possible way, and because it is a deterministic game there is a clear determined best possible way.

      Chess follows the exact same principles, it’s just many orders of magnitude more complex. Since Chess is unsolved at this point (though solvable) it’s unclear whether a optimal game will result in a win or a draw, or whether one needs to go first or second to have an optimal game, but for the sake of argument, let’s say that White can play a perfect winning game. If this were the case that means that if you were playing Black against a computer programmed to play perfectly and not make intentional mistakes, you would lose. Always. You can’t go into the game knowing it plans to play a certain strategy and counter accordingly, because the strategy it plans to play is the one to win regardless of what you do, and it takes in to account every single possible board position throughout the entire game.

      Now to be fair, the complexity involved in the eventual solution to Chess is probably such that a human is unlikely to be able to execute it unaided, simply because the task of storing every single possible board position and requisite ideal move in your head is not the sort of thing the human mind is designed to do, however once solved it would mean that it would be fairly trivial to make a computer chess program that, can only possibly be beaten half the time, dependent entirely on who is playing which colour.

      Again, to reiterate, Starcraft is in no danger of ever ending up in this situation, for reasons stated earlier, just something I find interesting to think about when stories like this get reported.

      Here’s some further reading on the subject written by someone far cleverer than me if people are interested:
      link to

    • ShineDog says:

      I think Nalano explains me well, playstyles, for someone who isn’t absolutely grade A brilliant at the game, being defined by flavour of the month build orders, just feels awful to me.

      On top of that, SC feels so artificial. So mechanical. That everything is so measured and clinical and arbitrary that I find my goodwill towards the game sapped very early on. Real world tactics don’t apply and the ones that replace them just feel wrong to me. The meter by meter precision to coax the most out of your units, the factory precise build orders… I want to play war, damnit! I want the chaos and intensity and insanity of the battlefield, if that means sacrificing a little precious balance for a random element? so be it.

      I’m not saying that SC is a bad game, it’s just a million miles away from what I’m looking for in an RTS. Which ends up being CoH (although what the hell are relic doing to it now argh) Men of War and Total War. Games which simulate the battle in a more naturalisitic and unpredictable way, are to me ten times more exciting

    • Thants says:

      There’s one big thing missing in this discussion though (unless I missed it): Fog of War. Games like Chess are at least theoretically solvable because they’re perfect information games. A game like Starcraft, where you can’t see what the other player is doing much of the time, doesn’t have a perfect solution.

    • Hidden_7 says:

      Yeah, Fog of War would be included under the banner of things I call “random” which really was a terrible term to use. Unknown elements is far better. You can get “optimal” choices for things with unknown quantities but they aren’t really solved.

      Examples of such routes include the Prisoner’s Dilemma, which has a min-maxing optimization, and card games like Poker, which uses choices based on statistical chance. The former attempts to find the highest “guaranteed” score you can get, while the latter attempts to optimize your win ratio over a long series of games.

    • Riesenmaulhai says:

      You guys keep saying Starcraft is realtime… but I have reason (and you should too) to believe that it has 60 turns a second, which are played and/or skipped simultaneously. And it has a board as well.
      And by this, there can be a ‘best’ strategy (although it is possible that other strategies are equally good). But due to the fact that you don’t have perfect information all the time (as you have in chess), it is not solveable.

    • James T says:

      “This reminds me as to why I could never – ever – get into fighting games.”

      Because you don’t comprehend how they’re played, apparently.

      “It’s simply that, with the existence of an optimal build/strategy, the game devolves into finding and exploiting the flavor-of-the-month strategy”

      The challenge in fighting games is to keep your head together enough to execute the right strategy for the given situation, very, very, very quickly — and there’s usually more than one solution; you have to change it up depending on what you want to do next, and so your opponent doesn’t adapt to your one-dimensional response. Of course, this only applies to the responsive game; on offense, even the simplest character will have heaps of ways to attack, some of them better than others depending on your opponent’s style and choice of character. There’s no ‘one strategy’; you always have to think on your feet.

    • brog says:

      @Riesenmaulhai: 60 turns a second, one turn an hour, either one is real-time; i.e. the progress of the game is tied to a real clock, rather than players taking as long as they like to make decisions. You’re bringing up a different issue, that it’s not continuous-time, but it’s at a high enough resolution that this doesn’t matter – it’s beneath human reaction times, so it would be impossible for a human to play perfectly even if it was a non-random perfect-information transitive game.

    • Nalano says:

      Hey, James T.

      I get it. I comprehend it completely.

      I don’t like it.

      Perhaps you don’t comprehend that.

  2. Rinox says:

    lol @ pokemon discussion in the vid. Fascinating.

  3. Rob says:


  4. Comment System says:

    It should also be noted that Evolution Chamber is open source (under the Apache License). Hopefully that means there is a bright future for this program. I’m looking forward to the Protoss and Terran builds it comes up with.

  5. Om says:

    I’m impressed in spite of myself. In hindsight though it does make a lot of sense, in a, “Why hasn’t someone else tried this before?” way

    Once reason why I’ve always stayed away from SC or competitive RTS MP is the reliance on build orders and strictly defined strategies. I’m not saying that these are bad but they’re too mechanical for me. Which is why they’re perfect for combinatorial optimisation. I’ll be very interested to see how this impacts the SC community; could we reach a point where the players are merely robotic instruments of simulated annealing or GA programmes?

    • Comment System says:

      Short answer, no.

      Basically the problem space is much more computable in the early game. So while the program may be able to find an optimized build, it does not handle anything in the late or middle game. Player skill and intuition is still a big part of Starcraft 2. If the game was an easy problem for a computer to solve Blizzard’s AI wouldn’t need to “cheat” by harvesting more resources then a player.

    • Rinox says:

      Why anyone hasn’t tried it before? Because the vast majority of SC2 players I have met online aren’t making their own strategies, they’re watching vids of other people and reading build orders and copying them into the game without a second thought, basically.

      Once their strategy takes a hitch somewhere, they often start raging or just quit. Or, in the best case scenario, they are clueless what to do to recover. It’s really quite disheartening to see how most SC2 competitive players don’t really know how to play the game. They can perfectly execture a couple of strategies, but they don’t understand the game. Alas, this makes a lot of games a complete crapshoot as you get unlucky and they put all their money on a single strategy that happens to be the perfect counter to yours. Shit happens. :-(

      On the whole, though, I found it interesting that after a relatively long period of getting my ass handed to me, I started ‘getting it’ and started winning huge streaks (20-3 eg) online without ever looking at build order strategies or memorizing my own play style. I steamed from lower bronze to high plat in a weekend like that. I think it was the moment where I’d developed a playstyle that was a) not in a textbook somewhere and b) was experienced enough to execute it on the field. The opponents had no idea wtf I was doing, and it threw them off completely.

    • Urthman says:

      It’s really a lot like chess. Pretty much every possible opening in chess has been discovered and named, so when you play chess at a high level, you’re choosing a particular “build order” and probably using a standardized defense in response to your opponent’s build order.

      But at some point after the opening moves, the game opens up and depends more on the individual players’ choices.

    • Dawngreeter says:

      Also, computer is demonstrably better than a human at chess.

    • Solcry says:

      Kotaku (well, Gizmodo but same diff) actually just ran an article on this. link to

    • Danarchist says:

      I find it ironic how many similarities sc2 has with WoW. In wow to be the best little raider you can be you have to know the order in which to hit your buttons to produce the maximum threat/dps/hps. For example the paladins “969” rotation. In sc2 to be the best little zergling you can be you just have to memorize the build order and know when to send you roaches a roaching. With both games true mastery comes from muscle memory and the ability to react to expected events.

    • pkt-zer0 says:

      “With both games true mastery comes from muscle memory and the ability to react to expected events.”

      Check game 2 of the Blizzcon BoxeR vs Fruit Dealer match. Losing your expo and then taking the gold base on Metalopolis is not a standard transition.

    • Raum says:

      “I find it ironic how many similarities sc2 has with WoW.”

      Oh, give me a break. Butthurt because you can’t handle SC2 much?

    • Thants says:

      Danarchist: You must have literally never seen a single high-level match if you think that’s the case.

  6. Rich says:

    A strange game. The only winning move is not to play.

    • Skurmedel says:

      Do you want to play a game of chess Falken?

    • DJ Phantoon says:

      Later. Let’s play Global Thermonuclear War.

    • Jez says:

      The question isn’t “what are we going to do,” the question is “what aren’t we going to do?” …

      …. D’oh!

  7. Langman says:

    Strong in this game, is Aspergers.

    • DrazharLn says:

      I imagine this Lomilar fellow just made it for kicks, I know I would. I wrote a genetic algorithm based timetabling system last year (fundamentally the same as this, maybe a little less difficult, depends how this thing works) and it was a pretty fun project.

      I’ve always intended to use Genetic Algorithms in a game that I make. I think they’re wonderfully suited to the computationally difficult problems that games often produce.

      I actually first learned about these from a book by Richard Dawkins, an expert on real world evolution (amongst other things :P).

  8. Nick says:

    The uber-competitive nature of the online games world in general, and the RTS genre in particular, really puts me off.

    • Brumisator says:

      Blizzard put a lot of effort into SC2 in order to make the entrance into MP as smooth as possible. coop Vs. AI, solo vs. AI, 50 training matches before you enter the league (skippable if you don’t need them). I terrified at 1st, but I finally completed my baby steps, and played (and won) my 1st placement match yesterday.
      the bliss I felt was overwhelming.

      In a nutshell, you don’t HAVE to play SC2 online like a maniac, there ar a lot of normal people playing too, that’s what the league ladder is for.

    • Nick says:

      Good to know, thanks.

    • Damien Stark says:

      I’ll second this.
      I was also a “single player only” RTS guy, but was drawn in by the huge list of things they added to SC2 to make the multiplayer more accessible and ease you in.
      It worked. I really enjoy it.

      As Urthman says above, chess is a good comparison. Just because you see a lot of named techniques “three rax reaper rush” or “MMMG” and such, doesn’t mean it’s a dull and rigid game with only a few static winning strategies.

      You’ll need to learn one or two simple techniques to hold of rushes and survive the early stages of a game, but then it’s pretty much up to you how to play it. If you’re smarter or faster than the other guy, if you notice your opponents strategy and cleverly counter it, hell even if you just hit a patch of good luck – you’re not going to lose because you failed to follow “22 – 3 gate mass void rays” or somesuch.

      Even if all that doesn’t reassure you, there’s custom games against the AI, and Co-op against the AI, unranked and fun.

    • pkt-zer0 says:

      The uber-competitive nature of the online games world in general, and the RTS genre in particular, really puts me off.

      I never got this sort of criticism. Uninteresting I can understand, but off-putting?

    • Brumisator says:

      Pkt-zre0, I don’t understand what YOU are saying.
      something highly competitive usually comes with a (un)healthy dose of complexity and nuance, so you can’t flat out say it’s uninteresting.

      On the other hand, it can be overwhelming, and thus, people not willing to devote time to understanding it properly are put off by it.

    • cliffski says:

      I tend to agree. I play games for fun. Winning is fun, but losing in an exciting or funny or interesting way is just as cool.
      Playing against random online players is often a nightmare because regardless of age, they have the manners of 6 year old kids, and the competitive spirit of a caged tiger.
      If there is a perfect build order in an RTS, I don’t want to know it, I want to discover it as I play the game and elarn from my mistakes and have fun.
      If that part of the game isn’t fun, then something is badly wrong with the design.

    • Neil says:

      “Build order” to me translates to “a bunch of stuff you have to do by rote as quickly as possible without thinking or interacting with the other player.”

      That part of the game is not interesting to me.

      The interesting part, gameplay-wise, comes later in the game, but as with most competitive online games, if you play casually, you’ll have maybe 2-3 months after launch to find other casual players before you start getting crushed on a regular basis by people who play a lot more than you do and who are reading up on strategies and tricks when they’re not playing.

      That’s why I don’t bother with multiplayer RTS games nowadays, unless it’s just with real-life friends.

    • pkt-zer0 says:

      @ Brumisator: I don’t see how that’s up to the game being competitive, instead of just having a steep learning curve or similar (which isn’t true of SC2 if you ask me)

      @ Neil: “you’ll have maybe 2-3 months after launch to find other casual players before you start getting crushed on a regular basis by people who play a lot more than you do”

      The matchmaking is pretty good in SC2, actually, your win rate will be damn close to 50%.

    • televizor says:

      @ Brumisator

      Don’t forget the challenges, they teach you about unit counters, (wrong) build orders, hotkeys and everything you need to survive in MP.
      For me, the order would be: Campaing -> Challanges -> vs AI -> Practice games -> Ladder

    • Dolphan says:

      @Cliffski – But SC2 is all about learning from your mistakes, losing in interesting ways, and having fun. You can deduce basic build orders from first principles and practice if you really want, but the main point is that the obsession with build orders in external coverage of the game is misplaced. They apply to what happens in the first 3-5 minutes of a match. And yes, this is a game where you can lose in the first 3-5 minutes, but it’s when you don’t (and with a bit of practice, you usually won’t) that the interesting stuff happens.

      Also, manners don’t really come into it. Other than ‘gl hf’ and ‘gg’ at the start and end, there’s generally not a lot of communication in ladder matches.

    • Starky says:

      DO NOT be scared of playing SC2 online – It’s really not bad at all.

      I had never played SC1 (not even single player), I’d only every really played RTS games in single player (not even much lan) and I got into Sc2 easily.

      The matchmaking system is fantastic (for the most part), and quickly puts you against reasonable opponents.
      So with only experience from the single player game, a few AI matches and 10 practice games I got into silver. I’m now top of my gold league, and trying to push into platinum 1v1 – I am platinum in 2v2.

      With the practice matches though don’t do any more than 10, just to get a feel of build orders, as doing more will do more harm than good. Better to throw yourself in at the deep end and lose a few games, then work your way up.

    • Jim Rossignol says:

      Starky is right. The real success of SC2 is that it both has brilliant match-making and a large enough community to be sure you can find someone as rubbish as you are.

    • strafe says:

      One issue with the matchmaking is that not only do people tend to throw their first 5 games so they can be lumped with newbies in bronze but late at night there are no bronze people on. So if you aren’t very good, you can expect matches vs platinum and diamond almost exclusively when you play after 10pm.

    • pkt-zer0 says:

      @strafe: Funnily enough, I seemed to get bronze leaguers only, the last time I played during 2-4AM. Or is that so late in the evening it’s early?

    • Thants says:

      @strafe – Well, the fact that you can only have one account per copy of the game helps a lot with that, anyone tanking their placements to beat up on bronze player would just very quickly be promoted up.

      And are you sure that bronze players aren’t on late at night? Are there statistics on that somewhere? Because if you’re getting matched up against high-league player it could be that you’re on your way up or they’re on their way down.

    • Drowlord says:

      This isn’t much of a single-player game. The campaigns are interesting for possibly three days. The story wasn’t very good. If you’re really tolerant of a mediocre narrative, you can play it a second time at harder difficulty and eek a week out of it. A couple dozen vs-AI games exhausts the single-player possibilities.

      Honestly, I find the first several minutes of any SC2 game to be frustratingly repetitive and league play requires you to be as structured and unwavering as possible in those early minutes, which exaggerates my boredom. With the ultimate goal being a swift end to the game during the most boring part of it, I can’t say that I really get the point.

      I played SC1 frequently for nearly 8 years without really tiring of it. I played SC2 for nearly 2 weeks before I was bored to tears.

    • Thants says:

      “This isn’t much of a single-player game”

      I’d be interested to hear about an RTS with more of a single-player game than SC2.

      I’m curious what it was about the first Starcraft that held your interest more than the sequel. It just seems to me that SC2 doesn’t pretty much everything that SC did, but better and more. The campaign has a lot more replayability (More than 1 difficulty, optional challenges), the skirmish AI is much much better, there’s a few challenge maps, the editor for custom maps is much better, and it has a multi-player matchmaker. The story is pretty weak though, granted.

  9. Cramineft The Docile says:

    I say just remove players from Starcraft 2. Add a bit of random childish insults and TA-DA you have a prefect SC2 high-level match.

  10. tomwaitsfornoman says:

    This is the coolest thing I’ll hear of all week.

  11. Brumisator says:

    I’m scared.. mommy! The zerg! they are coming in the night to take me!

  12. Tei says:

    I have always trough music is just a simple matematical problem, and I have never understand why no one has “solved” it, creating the “ultimate” music.

    Is somewhat weird that this has ben made with a genetic algorithm, … I don’t know much about that stuff, but sounds to me like a genetic algorithm would be happy with a optimum local, ignoring the optimum global. So is not a good algorithm to get THE better strategy, just to optimize to dead existing ones.

    • Brumisator says:

      Orly, Tei?

    • Om says:

      Genetic algorithms are specifically designed to avoid the local optima that simple hill climbing gets stuck in. Aside from the usual ‘reproduction’ methods, most evolutionary algorithms will also include a ‘mutation’ factor to increase variability in the system. Its just not as straightforward as it is in simulated annealing or other local search techniques

    • AndrewC says:

      Music is just maths and logic, but it is humans that listen to it, and it is them that are the problem – stupid, irrational, emotional creatures that we are.

      If we could just make music with computers, and have it listened to only by robots, everything would be fine.

    • yabonn says:

      At each generation, the algorithm random changes the tested solution. You can calibrate it to try anything, trivial or crazy, and possibly get you out of the local optimum.

      Grats Lomilar.

    • bob_d says:

      @Brumisator: The thing is, you can do that exact same thing with other chord combinations. Not to mention:

    • leafdot says:

      @AndrewC – Unless, of course, irrationality (i.e, genetic algorithms’ mutations) is somehow integral to a self-sustaining, “true” intelligence…

    • SuperNashwan says:

      Just for Tei: link to
      Needless to say, this caused quite a fuss among some musicians.

  13. blah says:

    Great, until the next patch…

    • jalf says:

      Eh, then you just run the algorithm again to find the *new* best build order.

      That’s the clever (or scary, or off-putting, depending on how you look at it) part. The big thing here is not “someone came up with the perfect build order”, but that “someone came up with a way to *find* the perfect build order”.

      So where are all the people praising SC for its perfect balancing now where everything has a counterstrategy and nothing is ever overpowered? ;)

    • brog says:

      not the “perfect build order”
      just the optimal way to execute a build order towards a particular strategy.
      it’s inevitable that such an optimum exists since there is no randomness or player interaction at that point of the game.
      no solution.
      not a problem anyway.
      cool, but nothing to get upset about.

    • Brumisator says:

      I don’t get your point. If everything has a counter, surely the game is balanced?

    • Mman says:

      “So where are all the people praising SC for its perfect balancing now where everything has a counterstrategy and nothing is ever overpowered? ;)”

      A method to pretty much hard-counter this if you know it’s coming (which good players can suspect pretty easily) has already been found, so I’m not sure what your point is.

    • mrmud says:

      Genetic algorithms arent about finding “the best” for anything. They are about finding something that is good.

      As has been mentioned, genetic algorithms are good at navigating out of local maxima but that doesnt mean they are immune to them.

      Doing an exhausitve search algorithm (if it was feasible to do, but its not) could very well reveal a more optimal solution.

    • subedii says:

      In addition to what everyone else said about this being a build, but not the ultimate build (if there is such a thing), people don’t say that Starcraft 2 is perfectly balanced.

      Starcraft 1 was really well balanced, but that was after years and years of iteration, balancing and tweaking. Starcraft 2 is naturally going to take a long time to balance as well, and it will be impossible to balance until all three packs are out anyway, since each is going to add in a new unit or two. As it is they’re still currently rolling out patches and bugfixes that make significant changes to the gameplay.

      Man I don’t even like playing Starcraft 2, but I felt that was a pretty disingenuous statement all round.

  14. pkt-zer0 says:

    Hypercharging the evolution of the metagame through algorithmic optimization. Colour me impressed.

  15. frags says:

    First step towards a Skynet.

  16. drewski says:

    Just zergling rush him. Don’t play zerg? How unfortunate.

    • undead dolphin hacker says:

      So you got 6-pooled once, cried, and quit. Thank you for removing yourself from the pool of bads that clog up random 2v2.

  17. loGi says:

    Interesting strat. But as it turns out, protoss can win if they know what they’re doing. And you don’t even have to know the roach rush is coming, you can adapt after you scout him.

    Quite strong, nonetheless.

  18. WiPa says:


  19. Gremmi says:

    I swear this article isn’t in English. Can someone translate?

    • Dawngreeter says:

      Some people who are competitive at something have come up with something related to that something which resulted in other people talking about a number of loosely related somethings, reaffirming the competitive nature of their relationship to the original something.

    • mod the world says:

      In short: Some people can now advance from the low-life-league to the no-life-league.

    • K says:

      So playing badly is “low-life” and playing competitively is “no-life”?

      You’re awfully unfriendly to all gamers here, might want to not hang around a gaming site.

  20. Jez says:

    Another example of the pathetic bullshit that people will no doubt adopt in order to gain the upper hand in a game that they either suck at, or a game that means more to them that it should. WoW has the same problem, elitist cuntholes re-rolling FOTM classes (RDruid, RPal, UhDK) and taking joy in the fact that they’re about to faceroll an opponent with a strategy they weren’t smart enough to come up with themselves… worse yet, no human had come up with. Dear me.

    • K says:

      Soooo, you dislike optimized openings then? I would highly suggest you never try to play Chess, Go or any other game that does not rely mostly on luck then.

    • jsdn says:

      Any human could come up with these optimal builds with some math and time. Who would choose to play suboptimally? Your anger has no relevance here.

    • Jez says:

      Oh, don’t get me wrong; I’d be quite amazed if someone actually did come up with it.. and employed it to great effect. The part that bugs me – no anger here, excuse the loose French, it’s habitual – is the fact it’ll doubtless be copied by anyone looking for a cheap win. :)

    • jsdn says:

      If it would amaze you that a human would think of this, liquipedia would absolutely blow your mind.

    • Ribi says:

      Prepared choreography is pretty much inevitable in competitive gaming. I’m feeling like an old fart here, but my old competitive Unreal Tournament (’99 and 2004) group had something very close to “build orders” for team games; at the beginnings of matches, we generally executed one of several coordinated sets of movements, taking over critical parts of the map as a team and nabbing important items (the big health and armor upgrades, as well as clean-kill weapons like the sniper and shock rifle). After that, we often had coordinated, pre-planned ways for controlling segments of the map. Now, the reflex metric wasn’t commands-per-minute; it was movement fluidity and firing accuracy — it amounted to the same thing, really. Point is, the team very much knew who was going to be doing what early on in a match. We knew who was our best shootist, our fastest runner, our most dogged defender, and we put them into places where they could shine, along with enough backup that momentary lapses of concentration and bad luck didn’t ruin the match. We could definitely take on people whose tactical reflexes and in-fighting combat skills were better than ours so long as their strategy was inferior. Lose many battles, win the war, something like that.

      I don’t know what a good algorithm could have done for our performance; perhaps, it could have pointed out better ways to rotate force or exploit respawn timing. The enemy team temporarily grows weaker every time someone is killed, as the dead person has to run around getting weapons again; it’s quite possible that there were non-obvious ways to exploit this. We knew, generally, that sometimes it was better to leave one or two foes heavily wounded and out of ammo instead of slaughtering everyone — they’d be forced to either suicide (mox-nix on killing them) or run around gobbling up health and ammunition/weapon spawns to get back in fighting shape, which sometimes took longer than just re-arming from scratch and denied those refills to their lightly-depleted allies. Other times, we adopted a scorched-earth approach, trying to wipe them out in big waves so that they all had to re-arm together. Could an algorithm analyze team combat potential at any time, discovering hidden troughs (or peaks)? I’m tempted to say, yes.

      I also know (more from conversation than experience) that fighting games hinge significantly on sets of moves that work well together and well-established counterstrategies that most people in the competitive community are quite aware of. Who do you pick to best handle a known Blanka/Samus/Exdeath player? Same for most games that don’t amount to simply tallying up an individual performance metric and comparing end tallies (most rhythm games, some puzzlers and racing games). It’s been mentioned that MMORPGs work this way. Competitive team sports commonly have things like this — in US football, after considering athleticism and conditioning, most of the performance difference between teams lies in their execution of formation movement and the appropriate selection of counterstrategies. Martial arts are about leveraging the strength of the body to maximal effect; superior skill-in-movement-choice beats raw strength or speed in most cases, and strategic concerns crop up all over the place. The “jolt” falling-step jab is a punching strategy with benefits (more power and striking speed) and drawbacks (worse balance and slower recovery). I really do think that the fact is, most competitive games wind up having describable strategies that work better than instinctive reaction and quote-unquote “skill.”

    • DH says:

      Such a long post deserves some sort of response…

      What it seems like you’re touching on is the difference between technical skill and mental skill, and it’s an important distinction to make. In your case, technical skill is, as you said, fluidity and accuracy, while in a game like SC2, it’s in quick execution and APM. From the way you described it, you beat your opponents using mental skill, plain and simple.

      When you mention coming into the game with a pre-defined strategy, I’d say that’s *certainly* a subset of mental skill… the ability of your team to devise a working strategy before entering combat. More to the point, I’d call it a subset of mental skill because you claim to have devised it yourselves, despite repeating it over multiple battles. One of the problems that I (and apparently others) have with the SC2 early game (not the whole game, mind you, just the part where build orders are still relevant) is that most players aren’t inventing their own strategies, they’re either copying something an opponent did successfully or they’re copying something they saw watching clips of high-level players. When you’re just regurgitating a mechanical action that you didn’t think of yourself, the early-game stops being about your own mental skill (planning-wise) and starts being an exercise in technical skill/execution (which to some gamers, and Blizzard based on some of the changes they’ve made since SC1, holds much lower value than creative thought… though other gamers feel quite the opposite)…

      That said, without this algorithm, the early-game is at least an exercise in mental skill for the high-level players who invent certain build orders, and it’s always a potential exercise in mental skill in the event that you try something new and invent your own successful build order. But with the algorithm, assuming it attempts to run virtually every possible build order combination, creative planning in the early stages of the game becomes far less meaningful. (Though that’s hardly saying that there’s no creativity to SC2. There is, but personally, I’d rather avoid removing any of it from any stage of the game.)

    • The Geek says:

      Ribi, I was going to make the sports analogy but you beat me to it and with more words. Good job.

  21. K says:

    I also want to plug something that all those poor people who don’t understand competitive play should read. And those that do should read it too, to understand why “cheese” and “cheap plays” are part of the game:

    Playing to Win – Sirlin
    link to
    link to
    link to

    Everyone who plays any game against anyone else should have read these articles. They are some of the most insightful on the whole internet.

    • cliffski says:

      Not everyone does play to win. I often don’t play to win when I play Call of Duty, just to be a team player. And I play Company of heroes purely for fun and giggles.

      I take my business seriously, and my tax return seriously. I really don’t need to take computer games about aliens shooting laser guns seriously.

      Quite happy for others to do so, but don’t assume everyone plays the same way.

    • Malibu Stacey says:

      So you join a team & actively work against them so they lose cliffski? There’s a word (and an internet forum) for those type of people & I highly doubt they’d welcome you.

    • Skurmedel says:

      Lol, that’s not what he said at all.

    • Lendemain says:

      @Malibu Stacey

      “Not playing to win” doesn’t mean “playing to lose”. It means “I don’t care whether I win or not” or “Winning is important to me, but not very “.

      It ordinarily implies “My enjoyment of the game comes from the playing of it, and not from victory.” Griefers, as you’ve said, play to lose, and yes, they’re a pain in the ass.

    • Dinger says:

      He’s wrong, of course, and doesn’t appreciate how wrong he is. “Made-up rules” exist, and he recognizes that some of them exist, and even respects some of them. So he can’t really draw the line. Moreover his “scrub” psychology is puerile and makes him sound like a psycopath.

      On the other hand, the David and Goliath/Full-court press article linked in the Sunday Papers last year makes the point relatively clearly. There are rules of the game, and there are unwritten conventions. The latter are more vague (giving rise to accusations of “cheapness”) and exist for a variety of reasons. One of those reasons is to prevent the perceived less-strong player from winning.

      Victory, of course, means nothing without glory.

    • DH says:

      For those whose optimized robot brains can’t understand the idea of not playing to win, here’s an example:

      There are some shooters in which I might try an approach that is entirely absurd and impractical… but often, a strategy’s absurdity and impracticality is what makes it so awesome. I remember once in pre-class-update TF2, I ran into a base as a pyro and ran circles around a level 2 sentry just so I could kill the engineers that built it with my axe (I then got killed by the sentry)…

      And frankly, I wouldn’t have cared if those were my only two kills in that entire match, because those two kills were *hilarious*, *awesome*, and otherwise *totally worth it*.

    • Hyudra says:

      ‘Not everyone plays to win’

      This is true. I really like the Magic: The Gathering design philosophy, which originated when the designers came up with player archetypes so they could create selections of cards that had appeal to a wide range of players.

      ‘Spike’ is the player who thrives in SC2. Spike is the player who plays to win, who Sirlin is speaking to in his articles. Spike isn’t necessarily a good player, but he’s a player who gets the most raw enjoyment out of his victories. In its shortest and rawest form, it’s about ‘winning the game’. To a lesser extent, it can include players who get a kick out of winning confrontations in the short term or enjoy meeting goals, explicit or not (Xbox achievements, getting that piece of loot in WoW, etc.).

      ‘Timmy’ is the player who plays for the experience. In MTG, the purest incarnation of Timmy is the player who reaches for the ‘awesome’ cards, the biggest monsters, regardless of their efficiency. Outside of MTG, Timmy is more the player who plays for the social interaction, the immersion/interaction with the game and those moments you can tell a friend about days later. SC2 does have some appeal to Timmy (Ultralisks, those moments where you completely dominate), but it’s very dependent on Timmy having a circle of friends who are also into the game, because it’s not a game that lends itself to giddy war stories told to someone who’s not intimately familiar with it. Again, don’t assume ‘Timmy’ is a bad player (just like Spike isn’t necessarily good) – his approach to the game is different, but he can still be in that top 5%.

      ‘Johnny’, then, is the player who plays for the art. Not graphics, per se, but typically the creative combinations and the moments where the opponent says ‘I totally did not expect that’. Johnny plays for the unorthodox, with the typically ‘Johnny’ being the player who tries something off the wall 10 times, and is content to lose nine times for that one time when it works. To a lesser extent, Johnny includes the people in RPGs who like having a unique looking character, or strange builds, and the minecraft players who make stuff like 40 story mushrooms from Super Mario Brothers.)

      Speaking as a Johnny, I feel that SC2 really, really doesn’t appeal. It does have nifty combinations and unit interactions, true, but this article really highlights how so much of it is mechanical and set in stone, (and there’s also a rather daunting learning curve for entrants to any serious multiplayer) that a player with a lot of ‘Johnny’ in them can readily be turned off (with a Spike/Johnny hybrid likely finding the environment more hospitable).

    • DH says:

      That post was so good that, if you were a woman, I would want to sex you immediately… or at least after an intelligent discussion of competitive gaming theory and whatnot. The internet needs more people like you.

    • pkt-zer0 says:

      @Hyudra: “Speaking as a Johnny, I feel that SC2 really, really doesn’t appeal.”

      I’d also classify myself as a Johnny, and I disagree. Why do you think TheLittleOne is such a fan favourite? SlayerS BoxeR is probably also worth a mention.

      “and there’s also a rather daunting learning curve for entrants to any serious multiplayer”

      Disagreed on that as well, and SC2 was the first RTS I tried to play in “serious” multiplayer. Not my first “serious” multiplayer game, though, (as that would be TF2) but I don’t think that makes much of a difference here.

    • Hyudra says:

      I’ll elborate on the bit I had at the end in parentheses. No player is 100% one type or the other, and I suspect the reason you find SC2 palatable is because (despite being predominantly identified as Johnny) there’s other strong motivations at work when you’re in the game.

      I think any Johnny player who really gets into SC/SC2 (or, say, competitive level SF4, TF2, DOTA, HoN, etc.) is going to have to have a really strong streak of Spike in them. I’m of the opinion that Starcraft, because of the duration of the matches, the time/overall play experience it takes Johnny to learn the intricacies of the upper level strategies and the time (with each try taking a 5-30 minute match) investment in working & reworking a new/original strategy (or waiting for the situation to come up where you can try an offbeat unit combination, etc.)… well, it doesn’t lend itself to any but the most tenacious or brand-loyal Johnnies. So the player needs something else to drive them, and Starcraft doesn’t wholly lend itself to the social/immersive aspects that would motivate a Johnny/Timmy***. It’s a very competitive game, so it’s really Spike’s playing field.

      This isn’t a criticism of SC2 – very few mainstream games these days will appeal to a Johnny. There’ll be character customization options, but oftentimes the game that follows does not really provide the room for creativity in the game after that point. SC2 may allow creativity, but it’s a very different kind of creativity than many Johnnies want – it’s a creativity many compare to that in chess or Go – true, it may be something you can call creative and unorthodox, but that only exists in a way that is only fathomable by those who understand the game on a deep level. To someone watching from an outside perspective (and those of us wanting a more flexible ‘genuine’ creativity, without the presumption of the ‘spike’ mindset), it comes across as a rather boring game.

      ***There’s, of course, the Johnny/Timmy players who play almost entirely with their circle of friends, but this is the exception rather than the rule.

    • DH says:

      Aye, I’m very much a Johnny, but with some amount of respect for the Spike side of things… really, I suppose what I want most is a game in which the goal of winning and the goal of a Johnny (creativity and originality) are the same thing… That is, players compete and are scored or rated somehow based on what sorts original and creative things they can improvise during gameplay… Though at the same time, something as freeform as, say, a competitive version of Scribblenauts would be completely pointless to me as there’s not really enough of a defined objective or parameters.

      Really, given that games have a limited set of scoring methods, the creativity=points idea is always going to give way to some degree of Spike-ism… that is, attempting to devise an ‘optimal’ strategy for winning and then applying that when possible…

      But hell, there’s always hope. Assassin’s Creed: Brotherhood is looking to be quite the pro-Johnny competitive title, especially seeing as the more improvisational nature of stealth gameplay meshes well with the Johnny-ish strategy of creative and/or awesome improvisation. Also, Bulletstorm, if it has a multiplayer… because absurd dismemberment gives you points.

    • pkt-zer0 says:

      I think any Johnny player who really gets into SC/SC2 (or, say, competitive level SF4, TF2, DOTA, HoN, etc.) is going to have to have a really strong streak of Spike in them.

      Not sure what you consider competitive level in SF4, but 2000+ matches online I would consider semi-serious, at least. And yet I main Gouken (he’s mid-tier), because getting a counter off of a proper read, or better yet, a Shin Shoryuken, is just plain awesome. In BlazBlue, my character of choice was Hakumen, and after that, Bang. The former quite strong at lower skill levels, but ultimately low-tier, and latter just plain crap overall.

      I’m of the opinion that Starcraft, because of the duration of the matches, the time/overall play experience it takes Johnny to learn the intricacies of the upper level strategies and the time (with each try taking a 5-30 minute match) investment in working & reworking a new/original strategy (or waiting for the situation to come up where you can try an offbeat unit combination, etc.)… well, it doesn’t lend itself to any but the most tenacious or brand-loyal Johnnies.

      It seems odd to bring up game length as an argument in favour of Magic when compared to Starcraft. (I don’t play MtG, though, so maybe games are over in ten minutes). I’d also say that tenacity is one of the defining aspects of a Johnny – losing nine times out of ten doesn’t matter as long as that one win you pull off is sufficiently awesome.

      SC2 may allow creativity, but it’s a very different kind of creativity than many Johnnies want – it’s a creativity many compare to that in chess or Go – true, it may be something you can call creative and unorthodox, but that only exists in a way that is only fathomable by those who understand the game on a deep level.

      I do not see how Magic is different at all in this respect.

    • Dances to Podcasts says:

      This is one of the reasons I love TF2 so much. There’s so little emphasis on winning or losing, no one cares. I play it because I just like shooting stuff, not because I need some ego boost by being a winner.

      I tried playing FEAR (1) a while ago, because I felt like some doing some karateshoots. The players on the server suggested impolitely that if I didn’t score more, I should leave. So I did, and haven’t returned since. They can have their grim, silent ‘fun’.

    • Hyudra says:

      Not sure what you consider competitive level in SF4, but 2000+ matches online I would consider semi-serious, at least. And yet I main Gouken (he’s mid-tier), because getting a counter off of a proper read, or better yet, a Shin Shoryuken, is just plain awesome. In BlazBlue, my character of choice was Hakumen, and after that, Bang. The former quite strong at lower skill levels, but ultimately low-tier, and latter just plain crap overall.

      It seems odd to bring up game length as an argument in favour of Magic when compared to Starcraft. (I don’t play MtG, though, so maybe games are over in ten minutes). I’d also say that tenacity is one of the defining aspects of a Johnny – losing nine times out of ten doesn’t matter as long as that one win you pull off is sufficiently awesome.

      Not sure what you’re arguing here. It sounds more like you’re talking about a Timmy playing for that one cool experience than a Johnny reaching for the creative aspect. It’s not about that one win being awesome, but about making something happen. Thinking outside the box, to use two game elements together in a way that creates a whole new dynamic. That 1 win out of 9 is awesome because you did it by doing something nobody ever thought of & your opponent didn’t/couldn’t anticipate it.

      Hyudra: SC2 may allow creativity, but it’s a very different kind of creativity than many Johnnies want – it’s a creativity many compare to that in chess or Go – true, it may be something you can call creative and unorthodox, but that only exists in a way that is only fathomable by those who understand the game on a deep level.

      Response: I do not see how Magic is different at all in this respect.

      Magic caters to Johnnies on a number of levels.

      Johnny likes magic because he can set up combos involving interactions between 2 of the tens of thousands of cards out there, and because the game is set up in a way that allows you to create those dynamics. To be able to set up an infinite loop with two cards and deal 1 billion points of damage to the enemy, or to make the opponent draw every card in their deck. Compare to SC2, where each unit has fairly narrow, focused design, and there’s really no room for anything that would raise eyebrows from an observer who knows relatively little about the game.

      Johnny likes magic because the game allows offbeat play. You can do something unconventional (like a deck focused entirely around lands, or having a craptastic deck & effectively switching decks with your opponent so they have to play with the crap cards) and you can theoretically succeed. It’s delightful when your opponent is constantly wondering what you’re up to, and you pull off the grand reveal. Compare to SC2, where the game pretty much dictates things like build orders, resource maintenance, scouting and routine aspects, with deviation from the norm tending to result in utter failure.

      Johnny likes magic because you can be artistic. You can make a deck centered around a theme, like vampires, or cultists summoning devils, or with cards themed around the A team.

      Johnny likes magic because when someone says “It can’t be done”, you have the tools to prove them wrong. I can take cards that read “You gain 3 life, you lose the game in 3 turns” and make them work. In contrast, Starcraft 2 is very rigid.

      Fact is, SC2 is not a creativity-friendly game.

    • Lilliput King says:

      @Dances to Podcasts

      Indeed, especially over the halloween update. At one point during a firefight in the mansion, the gift popped up and everyone in the corridor instantly scattered in random directions, completely ignoring one another. I can see how that would irritate some people. Personally I love it.

      re: gamer archetype discussion

      I like the system, although it would appear that everyone wants to be a Johnny. This is completely understandable, though. I’m not sure what my several hundred hour in character play through of BG 1, 2 + ToB as a lycanthrope makes me. Possibly a freak.

    • JWill says:


      Firstly, let me say your post was excellent, well written, and the reason RPS has such a great community. I think as someone coming from the SC2 community, I can elaborate on your player types as they apply to starcraft 2 ladder players:

      ‘Spike’ is the player who uses ‘cheese’ strategies (like the roach rush designed by Evolution Chamber). These strategies focus on creating an opportunity for a win very early the game, and in most cases are ‘all-in.’ If the all-in cheese strategy doesn’t win in the first few minutes, Spike has already lost. These sorts of strategies take almost no understanding of the game, and you don’t become a better player as a result of using them. Spike may use these strategies to get a lot of easy wins quickly, and if he executes these mindless but mechanically demanding cheeses well, he will make his way to the highest league. Spike will be happy with this, because he has won. But he has learned nothing about the game, and when he reaches a level in the ladder where all players know how to deal with these cheese all-in strategies, he will be stuck.

      ‘Bobby’ is the player who plays for the game. He plays to improve, and with every loss he stops to think about what he did wrong. He doesn’t rage about how his opponent’s race is imbalanced, or how he just lost because his opponent did something cheap. Bobby always plays for the late game, trying to survive the Spikes who rush in the early game for an easy win. He knows that cheese or all-ins are viable strategies, but he doesn’t use them because he believes they won’t teach him about the game. Some day Bobby wants to get into the highest leagues, but until then he’s content to slowly and steadily improve.

      Of course there are variants of these types, and quite a lot of Johnnies around who try crazy fun strategies (mostly as terran, since they have the most tactically diverse units). TheLittleOne is absolutely amazing in this regard — watch some of his replays casted on youtube. If you know much about starcraft 2, your jaw will drop at his creativity and execution.

      And to anyone mislead by this article… the ‘perfect strategy’ created by Evolution Chamber wasn’t. Lomilar asked it how he could get 7 roaches as fast as possible, ignoring econ (making it an all-in rush).
      The early game has set build orders because there are so few variables. You have to scout what your opponent is doing, and dynamically adapt your opening based on his. Starcraft 2 is still too young for there to be a few known and formulaic build orders. And even if you do follow a build order by rote, it only lasts into at most the first 5 in-game minutes. Most games last about 15 to 30 minutes.

    • DH says:

      That’s kind of his point, though… the two types you’ve outlined there might have different methods (one is vastly smarter than the other, for example), but both are still playing to win the game. The goal of both is still to rank up, compete at a higher level, and enjoy their victories over their opponents.

      While you’ve got other subtypes and variants, that was really his whole point… that to play SC2 seriously (given that RTS games tend to have higher learning curves or require more time commitment), someone needs to have some of the goal-oriented and achievement-driven ‘Spike’ in their gaming personality. Again, that’s not to say that Johnnys can’t or shouldn’t play the game, but the Johnnys that do pay SC2 are going to have some amount of Spike in them as well, in order to be able to/to want to dedicate themselves to learning and improving.

    • JWill says:


      I see your point. Starcraft 2 is an absurdly competitive game, and in the end, it will only appeal to people who like to win.

      But as a ‘Bobby’ type player, I can state that I don’t get much enjoyment directly from winning. If I know I made a lot of mistakes and didn’t play well, I really don’t feel good about my victory. Starcraft 2 is a game of such complexity and sheer difficulty (even SC2 progamers don’t play perfectly) that playing purely to learn is a viable approach. You could play SC2 for years without learning everything about it.

      Conversely, Spike plays solely to win, without any introspection or thought. He’s the sort of player who crows about how his opponent’s a noob, and was owned.

      I fully agree with your point that anyone playing SC2 seriously has to have a fair amount of motivation for learning the game. It’s a singularly complicated title, and casuals will just bounce off the surface. But honestly, everyone wants to win in one way or another. The different gamer types described above simply define victories in different ways. Be it literal victories, the knowledge that you’ve improved, the fact that you made something creative and new, or simply using a very cool combo, everyone wants to win.

    • Tei says:


      Wait,… there are more people like me? .. I am a Johnny, in games and out of games. I play for the art, and to surprise myself and others. Wow, so there are more people like me? thats awesome

    • pkt-zer0 says:

      Not sure what you’re arguing here.

      Basically, this: “Johnny likes a challenge. Johnny enjoys winning with cards that no one else wants to use.” “For example, let’s say Johnny builds a new deck that has a neat but difficult way to win. He plays ten games and manages to get his deck to do its thing… once. Johnny walks away happy.”

      Compare to SC2, where each unit has fairly narrow, focused design, and there’s really no room for anything that would raise eyebrows from an observer who knows relatively little about the game.

      Depends on your definition of relatively little, I guess. I don’t think it takes much to appreciate the likes of TheLittleOne, really.

      Fact is, SC2 is not a creativity-friendly game.

      Not breakable in as insane ways as Magic, sure. I’d say there’s still plenty of room to develop your own style.

    • Hyudra says:

      Tei – I would say there’s a lot of Johnnies out there. Thing is, a lot of games really fail to appeal to Johnny, so those kinds of communities don’t tend to come about.

      What games did/do?

      Final Fantasy Tactics is a personal favorite, because of the sheer variety of team compositions/setups and how the game was balanced in such a way that you could really decide your own difficulty level; I could (and have) run through FFT with a team of 5 squires using only squire skills, and it was hard as hell.

      Minecraft? Up to date example. Build what you want, how you want. Use flammable goods and set the world on fire from horizon to horizon, or build a scale replica of the Eiffel tower.

      Scribblenauts? Deus Ex? Dwarf Fortress (very Timmy/Johnny)? Guild Wars?

      It’s a pretty untapped market, I think, and some attempts to tap it have failed in key ways. Spore springs to mind, given how the creativity in creature design was hampered by having a given creature’s mechanics fixated on said design.

      SC2 tends to attract a lot of rancor from players who likely come from a Johnny mindset, I think, because it really isn’t a Johnny-friendly game. So many criticisms (here and elsewhere) that tend to get directed at SC2 are based around the inflexibility of the game. Rote, routine, setup, application. Very rarely will a player win because they were more creative, when their opponent was flat out better.

      Dunno if I’m gonna return to this particular page/comments thread, but if anyone wants to chat more on the archetype subject, you can bug me on AIM, where I go by Domina Bellatrix.

    • archonsod says:

      “Basically, this: “Johnny likes a challenge. Johnny enjoys winning with cards that no one else wants to use.” “For example, let’s say Johnny builds a new deck that has a neat but difficult way to win. He plays ten games and manages to get his deck to do its thing… once. Johnny walks away happy.””

      The Johnny archetype doesn’t care about winning. He would in fact be perfectly happy losing 10/10 times as long as he managed to get that insane six card combo working in one of those matches, even if he still lost that match. For them the game isn’t about winning in the sense of beating the opponent, it’s about beating the rules – bending, breaking or otherwise stretching them to see just how far the system can be pushed. It’s why SC 2 and similar would have little appeal; the mechanics of the game are too rigidly enforced to allow for much in the way of bending those rules.

    • JWill says:


      Just a quick note… my all time favourite games are Minecraft, Dwarf Fortress, Deus Ex. So, I guess you could characterize me as a Johnny.

      However, Starcraft 2 has surpassed even those, and is a game I see myself playing five years from now. So I would hesitate to call SC2 repellent to creative types. Once you get to know the game, there’s an absurd amount of creativity you can put into it.

    • pkt-zer0 says:

      “For them the game isn’t about winning in the sense of beating the opponent, it’s about beating the rules”

      Considering that the Golden Rule of Magic is that any rule can broken if a card says so, that’s not a very meaningful definition.

    • archonsod says:

      That would be why they tend to play Magic, yes.

    • pkt-zer0 says:

      @ archonsod: That doesn’t have much to do with my point, but whatever.

      Anyway, breaking the rules is far from the only way to express style, quite the opposite. Self-imposed rules define the latter three of the four Johnny subgroups listed in the follow-up article.

    • Mitthrawn says:

      These classifications are ridiculous. Everyone’s going, I want to be a Johnny, I want to be special and unique and a beautiful creative butterfly. If everyone’s picking one classification, a classification the original poster put himself in, clearly its a flawed system. If no one says, hey I want to be a timmy or a spike, then clearly its just self congratulation and going, well, I’m a bit rubbish at these games but I like to think of myself as creative, therefore I’ll call myself creative and fun. You’ve made one classification (your johnny) highly desirable, and therefore no one wants to pick any of the others. Of course you’re going to all say you’re johnnys, its like when they do those astrological readings and the cards come back “your a caring kind person with a streak of good in your heart, except when you get angry.” It’s bullshit. Make a classification system that appeals to different types of people or that system is worthless.

      We all have bits of johnny, spike and timmy in us. but right now you’ve given anyone who is rubbish at games an excuse to be rubbish at games “oh, its ok that I’m not getting any better, I’m a creative sort who’d rather dribble over his painting samples than put the effort in to get any better.”

    • pkt-zer0 says:

      @Mitthrawn: “If no one says, hey I want to be a timmy or a spike, then clearly its just self congratulation”

      The folks who brought up Sirlin’s playing to win would likely consider themselves to be Spikes, so that’s not entirely true. But you do bring up a good point.

  22. Nathan says:

    I’ve been away from the game for a bit, but while this is a clever build it by no means imbalances the game. Building nothing but drones until 16 supply is really late in the game to start building troops. Any sort of early rush/pressure should defeat this. Not to mention that you are really susceptible to anti-air. If a terran gets even one banshee out he can lift buildings to minimize the damage. No doubt this is a powerful build, and great against an opponent who is too noob to actually scout out what his opponent is doing. The only reason people cry about builds like this is that too many people think they can just sit back and do a strong early macro build without bothering to see if the other guy is going to build troops early.

    • jsdn says:

      “Building nothing but drones until 16 supply is really late in the game to start building troops.”
      Have you played zerg… ever? How can you speak on balance when you make such a statement? You wouldn’t happen to work for Blizzard, would you?

    • loGi says:

      It’s a strat made to be used against protoss, who usually don’t rush for air and can’t lift buildings. Misleading title picture.

    • Dolphan says:

      And a normal Protoss build won’t get a gateway out until 16ish (build at 12/13) …

  23. Noterist says:

    This particular build is a bit scary and will definitely need to be toned down (good luck to Blizzard figuring out how without messing with the entire Zerg production mechanic).

    At least it’s easily scoutable (produces no units until the rush starts) and if it becomes common enough then the metagame will shift to accommodate it. The trick to playing against Zerg has always been to put them on the back foot and force them to ditch build order strategies anyway, rally your first unit to their base and pester them so that they make Zerglings instead of an economy.

    I personally love EvoChamber and think it’s good for the game, it helped me learn the race by refining my fast expansion into defensive roaches build that gets me through the “first rush” mark (6:30-8:30 mins). Things never go 100% to plan, but it’s a good framework for me to base my reactionary play around.

    Like Rinox said above, there comes a point in SC2 when you “get” the game and don’t win/lose by build order mismatches. Medium length macro games that go to 2-3 bases each are the game’s real strength and once you learn how to scout for and defend against one dimensional one-strat kiddies you’ll be having a larff.

    • Jutranjo says:

      What’s the problem here exactly? For months zerg has had to scout terran constantly for the 12 different openings that auto lose the game for zerg if not scouted and countered in advance.

      People who know one cheese and try to get higher in the ladder with it will just fall apart if they ever reach a gamestate where the cheese didn’t win them the game then and there. If it’s only cheese you can do and you fall apart later it’s a pretty bad gameplan from you, if you 6 pool and then transition into expanding and actually playing the game, that’s a great gameplan.

      I love playing SC2 because of builds/options like these.

    • Noterist says:

      Not sure you meant to reply to me there Jutranjo as we seem to agree on everything.

      Unless you mean what’s the problem with the 7 Roach Rush existing and why did I think it had to be toned down? Namely for the same reasons that Blizzard already toned down the build-order-wins that you mentioned for Terran, and because it’s barely mathematically possible for the other races to magic up the sizeable army needed to counter a 7RR even if they know it’s coming.

  24. Mark says:

    Look at all the people praising the game as deep and strategically interesting now that it has what amounts to a dominant strategy. Nothing about this rewards being good at the game – at outthinking the enemy, adapting to changing circumstances, misdirection – just being good at following a script. Makes you wonder how they have the gall to call it a STRATEGY game at all, when playing hotkeys like a piano is the only skill that matters.

    • Stijn says:

      If you scout this build (and any decent player should scout) it is possible to hold it off. There’s a few other early game tactics that are, when not scouted in time, very hard to hold off – called “cheese” by the community and often looked down upon. But, as said, it’s always possible to defend against it. Once it gets into later stages of the game there’s no such thing as a failsafe or almost failsafe strategy, and it really comes down to player skill and ingenuity.

    • K says:

      If this proves to be unbeatable and Blizzard does not change it, you’ve got a point. Otherwise, you’re just a whiner.

    • Mman says:

      Maybe read the given information before ranting? This strategy can be very powerful against people who don’t know about it, but it’s almost an an auto-lose against anyone good now it’s a known part of the metagame.

    • TotalBiscuit says:

      If you don’t know anything about Starcraft, try not to discuss it, it’ll make you look silly.

    • Huggster says:

      Total Biccie – how does SC2 compare to Men of War?
      If I like MoW will I like it?
      Its a shame but I went off base building RTS in the 90s only to revisit for CoH which did not overly impress after MoW.
      I did cane Dune 2 back in the day though.

    • Luke says:

      @ Mark:

      Pretty sure it’s been said above, but given that you don’t seem to have read much of that; I can only assume you’re not a fan of chess or any game/sport in which mechanics don’t overwhelmingly favour luck then? What is it that you’re a fan of which is so impeccably superior in it’s strategic scope to SC2?

  25. Ovvno says:

    I always wanted to make something like this for ship builds in eve, now that would be usefull….

  26. MacBeth says:

    There does seem to be quite the clash between those who can’t understand why someone would want to play “sub-optimally” and those who can’t understand why someone would want to (strive to) play “optimally” at all times. Clearly some of us have quite different perceptions of what is fun…

    Can’t we all just get along?

    • ix says:

      I think the clash is between people who don’t understand SC2 and those who actually play it. Seems to be a lot of people assuming this is somehow an “optimal” build order. It’s a local optimum sure, good for a certain purpose, but it’s no win by default.

      People talking out of their ass is always tiresome.

  27. Vinraith says:

    In fairness, I’m sure this kind of thing could be done with most any strategy game on the market. That said, it’s certainly one of many reasons I don’t play strategy games competitively online. I have no interest in turning what should be a fun exploration of the rules and decisions of the system into a race to click the “right” buttons before the other guy. Hell, to me something like this qualifies as a “mechanical spoiler,” in that knowing the best builds really sucks all the fun out of the game for me (which is why I avoided actually reading said build order).

    • Dawngreeter says:

      Mechanical spoiler is relevant only for enjoying a game as an interactive narrative experience. It is irrelevant to sports, which is what this is about. It is not a mechanical spoiler when, after a long while of engaging a sport in one way, a new and more efficient technique is discovered and adopted by everyone.

      You can play a sport casually, too. This determines the effort and time invested, not familiarity with the mechanics of the game. In bowling, even if I play extremely casually and am terrible at it, outperforming your opponents is the goal. Even if I don’t play very often and don’t try particularly hard when I do play, even if I don’t care if I’m the winner or not, I like scoring higher over scoring lower. I still can’t do a hook to save my life but It wasn’t a mechanical spoiler when I was told about it. Winning isn’t necessary to facilitate fun at bowling but the entire involvement in it revolves around attempting to perform better.

    • Vinraith says:


      If Starcraft 2 were no more complex than a sport, you’d have a point. Of course, if that were the case, I’d have no interest in it whatsoever. Strategy games are narrative interactive experiences to me, they’re about figuring out the best way to do things, not following a pre-defined script. Learning and manipulating the system, making difficult choices, that’s what makes a strategy game for me. If I already know the “right” way to do things, there’s usually little point in continuing to play.

    • pkt-zer0 says:

      @Vinraith: Knowing how to do an optimal roach rush does not automatically translate into knowing how to play a 30+ minute match optimally. Or not having to make decisions, for that matter: if you go with the same opening every time, your opponent can just go for a direct counter to win every time.

    • Dawngreeter says:

      I am not sure if I’m more baffled by sports being used as an example of something very simple or multiplayer Starcraft being considered a narrative experience.

      It’s been said here too many times, though – there is no “right” way. What’s presented here is not the “right” way to do anything. It’s just a very counter-intuitively strong opening. After which you play the game as you would in any other situation. Every strategy game ever built is consisted of three segments – opening, mid-game and endgame. The first and the last part are book-learning segments. Always. Even in curling.

    • liq3 says:


      “If I already know the “right” way to do things, there’s usually little point in continuing to play.”

      If you knew the right way to play SC2 competitively, you would be in Korea playing in the GSL and winning the prize money. You aren’t though, are you?

    • jsdn says:

      I know a lot of people like you, Vinraith. Not a single one of them has ever attempted to play an RTS against a human player, and thus they don’t understand what is involved. It’s like saying you think a food will taste terrible, so you just never taste it and tell us why you never plan to, as if to say that we should change how you think it tastes before you’ll try it. Well, we can’t.

    • Vinraith says:


      That point might have some relevance if I hadn’t played a fair bit of competitive Starcraft back in the day. To extend Dawngreeter’s analogy, there’s a reason I go bowling with friends, rather than angry 14 year-olds I don’t know.

    • pkt-zer0 says:

      @ Vinraith: We’re talking about Starcraft 2, here, though. SC1 was horribly broken initially AFAIK, with the 5pool being the strategy. The metagame has also matured considerably since then, i.e. people have a general idea about playing RTS games, “expanding” is no longer a new and wild idea.

      And regardless of all that, your experience with the game might not be necessarily representative.

    • Stijn says:

      It’s representative for him though.

      I mean, if you don’t like that kind of games, you don’t like them. Nothing wrong with that.

    • Vinraith says:


      I want to be clear, here, that I’m not criticizing SC2. I haven’t played SC2, I have no basis to criticize it. I have a fundamental problem with enjoying competitive RTS play, and the build order article is representative of it, that’s what I’m trying to convey.

      I like to tinker, I like to try new things, by the discussion up the thread a bit I’m decidedly a “Johnny” type, I don’t care about winning all that much (hell, I go out of my way to find strategy games that will kick my ass the first dozen times I play them). I can appreciate your earlier point that, in the case of SC2, the game “opens up” strategically after this early build-order driven period. My problem is that if I’m not willing to memorize and execute build orders of this sort, my sense is that I cannot be competitive during that later, far more interesting period of the game. Indeed, I may not get to see it at all. If this kind of “cheese” opening is representative of the competitive side of the game, and I suspect it is, then said competitive space is not one in which I can actually have any fun by my standards.

      This problem is resolved, for me, either by playing solo or co-op against AI, or competitively against like-minded friends. It’s a shame, though, that there’s no “casual league” for players more interested in tinkering with the system than actually being fiercely competitive. I’d play the hell out of that, at least until the “winning is all that matters” crowd inevitably crashed it.

    • pkt-zer0 says:

      @Vinraith: To be frank, I think your fundamental problem is that you don’t quite get competitive RTS play yet. You’ve yet to give it a proper shot, basically.

      I don’t think memorizing build orders is that big of an issue. There’s not much to memorize, for starters. And it all kind of “falls out” of just the basic principles of building supplies, workers, and units. The reason for their existence is, as with chess, simply due to the constrained game space at this point. There’s not a lot of options, nor a lot of ways in which the opponent can screw with you. I wouldn’t say optimizing the early game is absolutely crucial – I basically just wing it, and I’m still at the top parts of the silver league (was gold in the beta, before I switched to random). You seem to think this is a bigger problem than it really is, imo.

      I don’t see how tinkering is supposed to be the opposite of being competitive – the idea is to come up with a non-standard strategy that wins, after all. Or are you simply looking for completely terrible opponents that would lose to any strategy regardless?

      The lower half of the player pool I’d easily classify as “casual”, with some folks never bothering to attack or expand for the first 30 minutes. That should be ample time to prepare for whatever off-the-wall strategy you thought up – even though a standard approach would’ve already won by then (as turtling on one base for this long is generally a horrible idea).

    • Thants says:

      @Vinraith – There is a casual league, bronze league! Really, the matchmaker solves a lot of the problems people have with this game. If you want to play the game casually and worry about planing a perfect build order you’ll be matched with people who are the same.

      And memorizing a build order can be as simple as “When you have 9 drones make an overlord, at 13 make a spawning pool, then a few more drones and zerglings.”

    • jsdn says:

      It’s painfully obvious that you don’t know what is involved, Vinraith, much the same that the rest of the people I’m basing you off of that have never played an RTS competitively. It’s not easy to explain, and the rest of the replys here are doing a poor job of it. Read the comments about Chess for a decent explaination.

      There’s a considerable amount of room to tinker. There’s the bronze league to play essentially casual players. This build is not in the least bit representative of competitive play. Your reasons for why you dislike competitive RTS are simply inapplicable to SC2. I’m constantly refining ridiculous builds to solve their various weaknesses. It’s entirely possible to reach the highest league doing unconventional strategies, and it’s entirely possible that someone doing that could completely shift the metagame if they find something that works really well (something that happened several times in Brood War, and has happened several times in SC2 already).

      That’s all I have to say. If you’re going to continue to be closed minded because of an opinion likely briefly formed 10 years ago (when RTS was a fairly new genre and there was no such thing as esports) then this discussion is pointless for all of us. You can not like competitive RTS, but the reasons you’ve given for not liking it are dead wrong.

  28. crooon says:

    Dammit, I’m busy with work and school all week, so I won’t have time to even try this out until next week, at which point it won’t work anymore :(

  29. Diziet says:

    From a programming perspective I’m surprised an enterprising SC fan didn’t do this with the first game. I <3 genetic algorithms but outside of curiosity have yet to achieve anything useful with the technique; this post makes me think. Does solving puzzles count? Even if they are ones that are solved faster by a brute force approach….

    • Heynes says:

      The linked thread had a post linking to link to, which apparently also uses genetic algorithms to find optimal build orders. It also mentioned that it was pretty much ignored by the community, though that may be partially due to lack of promotion. Ultimately, it may simply be that the “edge” in using programs such as these do not really accumulate to much in the long run.

  30. HexagonalBolts says:

    If anyone fancies a team game:

    link to

  31. DJ Phantoon says:

    Seems a bit… formulaic.

  32. HexagonalBolts says:

    Also, Quintin you bastard, every game I play against zerg for the next month is going to be awash with sodding roaches.

  33. A-Scale says:

    Thank you so much for taking a look at this. This is why I love RPS, SC2, and gaming. Awesome.

  34. DH says:

    Ugh… stuff like this just strikes me as… I don’t know the right word for it… “game breaking”?

    I’m *not* trying to say that it’s imbalanced. As many others have said already, counters for the build order are entirely possible. I’m also *not* trying to say that an entire game of SC2 is just by-the-books APM-spamming mechanical action, because thanks to the mid-game, it’s totally not.

    What I am trying to say is that I don’t see why someone would be making this program (besides just to see whether or not they can, which is a good reason to do anything). It seems similar to the people in WoW who pay others to level up or get gold or run raids for them… you’re paying someone for the privilege of not actually playing the game. This algorithm is a lot of programming for the privilege of not actually playing the game yourself.

    Think about it… now no human person will ever create this build. There will never be anyone to come up with this by going against the grain and doing something counter-intuitive that ends up paying off and inventing a new build order to call their own. This algorithm is likely just going to keep on discovering powerful build orders, and as it does, it prevents real human players from discovering these things on their own! It’s eating creativity! It’s killing potential! If every possible build is discovered by some joyless machine, what use is it for humans to play around with early-game strategies for the sake of invention and innovation (or community recognition, or whatever makes you happy)?

    Again, that’s not to say that there’s no room for creativity in a single multiplayer match… there certainly is, after everyone’s finished their opener and the mid-game begins. What I’m saying is that the robot is stealing creativity from the meta-game and the community. My point is that there is an aspect of *fun* that exists in the game that will be eliminated as this algorithm runs its course. It is the anti-fun, and I fear its presence in gaming just as I worry about the mental health of those that support it.

    • Latro says:

      Eh… its a very interesting application of a ton of things you learn in Computer Science degrees and also can shed lights on aspect of SC2 strategy that are not obvious?

      I mean, thats more than good enough reasons for me. I find the idea & execution inspiring, as a failed computer programmer that stuidied genetic algorithms a decade ago :-)

    • DH says:

      Oh, I’m absolutely in favor of building it for the sake of building it. Especially if you’re the sort of computer science guy who gets his rocks off on just programming for the sake of programming, to see what cool stuff you can do. Fun is a good thing, and if you find programming fun, then program the hell out of some code.

      I suppose what I dislike about it is implementing the results… actually using it to discover new in-game strategies and then posting those strategies to the community at large. That’s why I’m saying it’s sapping fun away from others — there could have been a real person to come up with that build order out of the magic of their real human brain, but now all the glory has been taken by a machine that will feel neither pride nor enjoyment.

  35. Caleb367 says:

    Well, this sounds exactly like Charlie Chaplin’s Modern Times. Endless mechanical repetitions in the same sequence, over and over and over. You click the key sequence fast enough? You win. Not fast enough? You lose. That’s all there is. A Guitar Hero for obsessive-compulsive psychopaths. I know it sounds harsh and I’m not trying to offend anyone here, but these kinda things actually scare me to hell.

    • A-Scale says:

      Only someone who has not experienced high level SC2 play could argue this. It’s infinitely more about flexibility and creativity than it is speed and sticking to a single rule set.

    • Caleb367 says:

      Too bad the OCD’s obsessing on build orders do their best to prove otherwise, imho. Maybe (maybe) it does have all this “deep” gameplay you talk about, just it won’t ever see the loght ’cause when you try that awesome unconventional strategy you’re already been trashed by the uberquick by-the-book nerd. And that was my point.

    • pkt-zer0 says:

      “Maybe (maybe) it does have all this “deep” gameplay you talk about, just it won’t ever see the light”

      Watch the Day[9] daily. You might be surprised that it’s actually not 200 hours of “well, he lost because he just didn’t click fast enough!”.

    • Nallen says:

      Yeah that’s exactly what this would seem like to someone that doesn’t play SC

    • Caleb367 says:

      I’ll try that, thank you. I’m not expecting much to be honest, but I take pride in not being single-minded on anything. Rampant fanboyism sure does more harm than good to SC2’s reputation, as repeating endlessly “u dont liek it cuz u suk” or “you never played it thats why said this” actually prove the opposite in my book. (and, metagame? Puh-leeze. We’re talking about game mechanics, not some freaking cutscene, achievement, unlock or whatever crap has been thrown in. It’s like guy A says “wow, this car sure drinks a lot of gas per mile” and guy B “YOU DON’T UNDERSTAND!!! IT HAS MP3 PLAYER!”)

    • Stijn says:

      I think what’s meant by “metagame” is the way the game is generally played. Like, if you go on now and play a few matches against a Zerg player, you’ll probably see a lot of people using roaches in their strategy. That’s the metagame.

      It’s kind of an awkward term, but it’s part of the Starcraft 2 jargon, I guess…

    • Dolphan says:

      Yeah, the metagame means what it means in something like MtG – the effect of the popularity of particular strategies on how people play (Terrans rarely reaper harrass anymore, so Protoss might not feel like they should get an early stalker like they used to – that’s the metagame).

      And it isn’t a choice between flexibility and strategy vs. knowing a good build order. Build order is for the very start of the game. There’s nothing pre-specified or mechanical about what happens after the first 5 minutes.

    • Thants says:

      @Caleb367 – Rampant anti-fanboyism also does more harm than good to gaming sites like this one. I hate to be the one to say “you never played it that’s why you said this”, but what else can I say to someone criticizing a game based on mistaken assumptions?

      I think you’d find things like this a lot less scary if you took the time to learn what they’re really like, rather than sticking with knee-jerk reactions.

  36. Jaunty says:

    “Basically the problem space is much more computable in the early game. So while the program may be able to find an optimized build, it does not handle anything in the late or middle game. Player skill and intuition is still a big part of Starcraft 2. If the game was an easy problem for a computer to solve Blizzard’s AI wouldn’t need to “cheat” by harvesting more resources then a player.”

    This is true, but the idea behind finding the optimal build is to get a victory in the early game.

  37. Scandalon says:

    Full disclosure: I’ve never played SC2, only played a bit of Single-Player SC1.

    Those complaining that this somehow “breaks” the game…did you stop playing basketball when someone realized they could jump and throw the ball at the basket at the same time? When people got tall enough and could jump high enough to dunk?

  38. Son of Kong says:

    Is there a mod I can download that plays the first 3-5 minutes for me, then I can start to actually have some fun!

    • Kieron Gillen says:

      I suspect this is why I’ve never got back into Starcraft after playing with Quinns. Those first few minutes at the start are simultaneous dead time, complete panic time and incredibly punishing. I’d love to see a game mode where players get to choose from a variety of pre-built bases/with troops and get right into the more flexible, unpredictable stuff.


    • Lendemain says:


      That reminds me of Chinese chess. Because the setup is much more open than that of Western chess, the game begins with the tactical middle phase rather than the strategic opening one. This makes it more exciting for new players–which isn’t to say that Western chess is boring, but only slow in getting to the conventionally interesting part.

    • Kieron Gillen says:

      I’m reminded of that series of posts about boardgames and mechanics they should consider stealing.


  39. nmute says:

    sc2 zerg vs comments thread.

  40. AgentD says:

    To make a required car analogy, “I don’t like Starcraft because all that matters is who memorizes the cheap build order” is about as accurate as “I don’t like racing because all that matters is who pushes the gas pedal better.”

    Unless you are in the top 10%, your build order is probably the least important part of your game. If you are there, then you have enough experience to take it into context.

    • Boldoran says:

      I disagree. A basic build order is the beginners ticket to success. In the top 10% micro and execution will make the difference but below things like scouting an correctly building a counter to what your oponent does wil win you games. You don’t even have to copy it somewhere you can just apply some basic logic to the problem and go with what works. This won’t allow you to be a progamer but you will gain ranks fast.

      I was terrified of playing starcraft online but the matchmaking is pretty good and in the beginning you play against people that suck just as much as you do. But once you learn the most basic things like building some units instead of just teching up and hoping to get away with it or constructing a chokepoint at your baseentrance to hold off rushes you will be gaining ranks in no time (I was in platinum and belive me I don’t know much beyond how to hold off the most common rushes )

    • AgentD says:

      Secretly you agree with me – what you are describing is not what the unwashed would call a build order. What you are describing sounds to me like “good fundamentals,” the absolute basics – make workers constantly, produce out of buildings constantly, expand appropriately, don’t get supply blocked. Until you have those, whether you go 11pool or 16pool makes essentially no difference.

  41. protorp says:

    I have no strong feelings one way or another about SC2, but this still hit my COOL button…

    Mainly due to the fact that it’s about programming genetic algorithims for the fighting strategy of what are, let’s face it, tyranids. This Lomilar is actually some warp-teleported, millenia in advance hive-tyrant first contact agent of hive-fleet Behemoth, no?

  42. Iucounu says:

  43. Joof says:

    “It uses genetic algorithms to find powerful build orders, meaning his program takes a population of build orders, kills off the useless ones, and has the most successful ones reproduce asexually to create a new population, which tests itself again, and so on.”

    I have an issue with this. Now, I’ve only encountered genetic algorithms twice, once as an aside in a lab, and once in an AI class, but every time I’ve done them, the top percent that aren’t killed off are bred together to make a new generation and each of those children would have genes from both it’s parents. So wouldn’t that be the most successful sexually reproducing, since they mate with each other instead of spawning new answers from itself?

    • mike says:

      Haha, yes. What makes reproduction sexual is the swapping of DNA, not the rubbing together of naughty bits.

    • GHudston says:

      I’m pretty sure that it’s quite common for them to have the best specimens reproduce asexually. As long as there is some degree of random mutation in the offspring and some way to cull the weaker individuals then they will still evolve over several generations. Having them reproduce sexually would probably just be complicated and unneccesary in this case.

      Also, in before “that’s what she said”. ;)

  44. BobsLawnService says:

    This is why I’ll always suck at SC 2. My play style is just to wing it and have fun

  45. shinygerbil says:

    Wow. That’s awesome, and I’m amazed nobody has done this earlier. I’d be very interested to see this retroactively applied to SC1.

    On a slightly different note:

    Wherever anyone has said anything less than wholly positive towards SC2, there has always been a reply to the tune of:

    “OMG u so BUTTHURT becos u so FLURKING SCHNITT at SC2”

    This makes me a sad gerbil.

    • Wilson says:

      @shinygerbil – I expect some of the people reacting like that are just being preemptively aggressive since they assume that the people criticizing the game are fanatical anti-starcraft types (and to be fair, some of the comments here don’t suggest they fully understand the best parts of SCII/multiplayer RTS games), so they don’t bother with reasonable counter points but just jump straight to the angry defences. Starcraft seems to be one of those things which people love or hate, so people assume anyone who disagrees with them about it is likely to be totally biased, and many of them will be.

  46. Ted says:

    I just wish I didn’t need to mircromanage the building of my forces, after ‘learning’ RTS on TA and SupCom, I’m used to rates and setting up production, because I’m playing on a computer, somethign good and repeating menial tasks like unit production.

  47. smashed up atoms says:

    This has been pretty interesting to read. Thanks. Now my eyes hurt. But here’s some thinks:

    I wish I went into programming. Just the phrase ‘genetic algorithm’ gives me a semi.

    I’m glad so many people have made a comparison to chess. When it comes to multiplayer games, I always see modern gaming as a translation of the ‘conflict of the mind’ that chess perpetually is. If someone could make a game like chess with a sprinkling of twitch gameplay I’d be all over it like a hot towel. I would say (in my experience) WoW got close. Moves, counters, but also reactions.

    Finally, I think a lot of people have tried to point out how worrying this might be, the ‘pro-forma’ nature of this code. I think it’s called min-maxing. I was a min-maxer when I played wow and it’s a little unsettling to new players to see such degrees of commitment to squeeze every single improvement out of your character or team or whatever. In all highly competetive games, especially on computers where players seem to be more fanatical and invested, the degree of fine tuning can be scary. Maybe off putting to new players. Sometimes it seems to negate the entertainment value of playing in the first place. (Play co-op games instead! they are more fun).

    Lets just say there was a … hour long group quest you could do in wow to give yourselves a 5% stamina buff that stayed in the arena. How long do you think it would be before every single combatant did it, regardless of the quest in question being horribly dull and inconvenient, because they may live or die on that in a ranked match?

    I know I’m mentioning wow a lot it’s one of my best frames of reference.

    Also I don’t have opinions of my own. But would that algorithm work for DoW2?

    • DH says:

      That comparison of min-max gameplay is actually quite applicable to DnD. I don’t play it myself, mostly because I don’t know anyone else who does, but I know some of the mechanics and I’ve read a bit about game theory from the DnD perspective.

      In DnD, there exists a player archetype that’s generally considered undesirable to have in your game: The Munchkin.
      They are much like the ‘Spike’ gamer discussed in one of the threads up above, only rather than being just another style of gameplay, the Munchkin is a playstyle that is generally discouraged by the community. These are people who play to win and will make all sorts of compromises in the name of gameplay optimization (but with DnD, the point is not to optimize but to play a role and act as a character in the DnD world, not try to become that world’s god-king). The problem with them, from the perspective of many DnD players, is that there is no ‘win’ or ‘lose’ in DnD. It’s not a competitive game — hell, a player shouldn’t even be competing with their DM. It’s a co-operative/collaborative experience for the sake of getting the most excitement possible out of your game.

      That’s not to say that Munchkins shouldn’t be allowed to have fun, but the problem is that their win-at-all-costs style of fun tends to make things less fun for everyone else by annoying the other players (for example, by breaking generally-accepted rules, such as a character having to act using their own knowledge rather than the knowledge of the player). And if every player were a Munchkin, they still probably wouldn’t be having fun, because they’d all just be getting annoyed trying to top one another and ‘win’ their table.

      …I forgot where I was going with this. Hell, I’m barely sure where you were going with your post. Oh well…

      …Ah! Right!

      A lot of people think that games ought to be played for fun rather than victory, even in competitive settings. While it’s true that winning is fun for a lot of people, the argument against playing to win is akin to that against camping in a game like Modern Warfare: Yes, camping in a corner will keep you safe and give you the best kill/death if you’re any good, but it also makes the game suck, because sitting in virtual corners for hours is piss-poor entertainment.

      Though as many have mentioned, Starcraft isn’t quite that simple. While it does have a lot of that robotic ‘min-maxing’ in the first few minutes of the game, it does later become a battle of minds and such. Sorry to be long-winded to get to something so obvious.

      TL;DR version: Read the whole post, ya lazy ass.

    • Thants says:

      @smashed up atoms – “If someone could make a game like chess with a sprinkling of twitch gameplay I’d be all over it like a hot towel.”

      Speed chess? :)

      “Lets just say there was a … hour long group quest you could do in wow to give yourselves a 5% stamina buff that stayed in the arena. How long do you think it would be before every single combatant did it, regardless of the quest in question being horribly dull and inconvenient, because they may live or die on that in a ranked match?”

      On the plus side, there was something sort of like this (some intense micromanagement could make your workers 7% more effective early on) and it quickly got patched out.

    • Archonsod says:

      You can divide most RPG’ers into two categories – the ones who want to roleplay, and the ones who want to game. Munchkins tend to fall into the latter category, they’re interested purely in their character sheet, inventory and the game mechanics and not at all interested in portraying Sir Giles the Paladin. Roleplayers are the opposite, they’re there to portray the character first and foremost, and the character sheet is an annoying distraction at best, if they even remember it exists.
      The problem tends to be with them pulling in different directions. Munchkins want to game, they’d rather be rolling dice than conversing with NPC’s and they tend to describe everything in terms of mechanics. Roleplayers tend to look down on such behaviour as being opposed to the spirit of roleplaying. At the same time, roleplayers will happily spend a three hour session discussing the finer points of some obscure point of the lore with an unimportant generic NPC rather than actually doing anything game wise, and their insistence of staying in character as much as possible tends to be thought of as pretentious waffle by the munchkin types.

      The advantage the RPG has is the GM is there to smooth out or mitigate the differences between the players, and a good GM will design their campaign so that both player types are getting what they want from the game.
      The main difference is, with a pen and paper RPG it’s possible to discuss the expectations of the players beforehand. It’s impossible to do it with random matchmaking on the PC. What perhaps is needed is an opt in for competitive play to distinguish between those who are interested in playing against a human opponent, and those who simply want to drive tanks through buildings and the like.

    • Thants says:

      @Archonsod – You can always play co-op against some very easy computers if that’s what you’re into.

  48. Commissar says:

    Sure is Bronze league in here, actually I would even go that far.

    Plastic league: Where you finish the single player on casual and you talk about competitive multiplayer without actually knowing what you’re talking about.

  49. LionsPhil says:

    A game that can be won by frantic clicking in the first few minutes is kind of lacking in strategy.

    Oh, wait, Starcraft.