Here's an article i wrote about some competitive game design theory. I figured some of you might like to read it, so here it is! <3 TL.net!!
-------------------------------------
The Marginal Advantage By Sean Plott
I recently was involved in a Mancala competition, where the entrants had to code an artificial intelligence program that could play Mancala. It taught me an important lesson about competitive game design.
Mancala is a game in which the winner is the player who “captures” more stones than his opponent. Thus the winner of the competition was the entrant whose AI program bested all others in stone capture. One of the coders devised a computer program that would maximize the number of stones captured in a given turn. Throughout the tournament, he steamrolled his opponents, repeatedly winning by fifteen stone margins and instilling fear and despair in the hearts of other coders. However, his program ultimately lost in the finals by 0-2, and lost each of those games by exactly one stone. In fact, I was shocked to hear that the winning program had consistently won every game it played by exactly one stone. How could the first program, which seemed to terrorize opponents, lose to another program that could only barely squeak out a victory each time?
The results were explained by subtle differences in their approaches to game play. The first player wrote a greedy program, one that would gobble up as many stones as possible. The coder reasonably theorized that maximizing the number of stones would maximize his chances of winning. However, the winning coder displayed even greater insight into the game: his goal was to have more stones after both players had taken a turn. In a sense, after taking the lead, he chose to maintain that lead, rather than extend it. For example, instead of capturing ten stones in a turn, he would capture four, knowing all the while that his opponent could only capture three, and that his lead would thus be extended by one. In a sense, the first programmer’s AI was successful in maximizing the number of stones captured in each turn, it just happened that this was always one less than the winner.
I found this incident to be particularly intriguing, as it reflects the nature of successful competitive game design. I’ve been involved in the competitive gaming community since 2001. Although my primary game is StarCraft, I have considerable experience with WarCraft 3, CounterStrike, Marvel vs Capcom 2, and a variety of other games. Despite the fact that these games function in drastically different ways and demand completely different skill sets, the expert players, the players who consistently win, always share a single commonality: they play comfortably with a marginal advantage.
The marginal advantage embodies the notion that one cannot, and should not, try to “win big.” In a competitive setting, the strong player knows that his best opponents are unlikely to make many exploitable mistakes. As a result, the strong player knows that he must be content to play with just the slightest edge, an edge which is the equivalent to the marginal advantage. More importantly, a one-sided match ultimately carries as much weight as an epic struggle. After all, the match results only in a win or a loss; there are no “degrees” of winning. Therefore, at any given point in a game, the player must focus on making decisions that minimize his probability of losing the advantage, rather than on decisions that maximize his probability of gaining a greater advantage. In short, it is much more important to the expert player to not lose than it is to win big. Consequently, a regular winner plays to extend his lead in a very gradual, but very consistent manner.
Amateur players, on the other hand, try risky, greedy strategies. In CounterStrike, for example, it is not unusual for amateurs to dash out into crowds of enemies trying to pull off a miraculous string of headshots in order to eliminate the opposing team. The majority of the time, this kind of amateur is fragged in a nanosecond. Expert Counterstrike players, on the other hand, patiently and carefully pick off enemies, knowing that such caution and precision virtually guarantee a win.
Moreover, amateurs often have no idea what to do with a marginal advantage once they gain one. I have personally watched countless games of StarCraft in which a player gained a massive lead but later lost the game. An opponent moves out a large force, and the amateur annihilates it with ease. At this point, the amateur has a marginal advantage: he has not yet won, but his opponent has lost his military and cannot apply any pressure for some time. The amateur in this situation will immediately try to win by launching a counterattack and will then crumble to a strong defense. Alternatively, the amateur will expand excessively, overextending his bases to the point where his defenses are too thinly spread. Such decisions violate the law of the marginal advantage, as they allow the opponent to get back into the game. They erroneously attempt to extend the lead, as opposed to maintaining it.
So, what’s so special about the marginal advantage? It might seem that all I’ve done is imply that newbie players take unnecessary risks and experts do not. However, just as playing for a marginal advantage is the hallmark of the expert player, the presence of a potential marginal advantage in a game is the hallmark of excellent competitive game design. The Mancala tournament brought this sharply into focus for me. Building in and allowing for a marginal advantage leads to exciting and dynamic play.
Not that providing for a marginal advantage is the only critical element in competitive game design. I will concede that all reasonably designed competitive games share three basic traits: ambiguity of optimal play, diversity of play, and allowance for skill.
First, there is no such thing as a good competitive game that has an obvious optimal strategy. For example, there is no competitive tic-tac-toe gaming, as two logical players would tie every time. I remember playing an old real-time strategy game called KKND, where one of the units was clearly the strongest unit in the game. Playing against my brother degenerated into a fray where we hurled masses of the same unit at each other until one of us got bored and stopped playing. On the other hand, a superior game like StarCraft presents a completely intractable problem. There is no best race, no best strategy, and certainly no best way to win.
Second, a quality competitive game should have a variety of techniques that can be employed in order to win. That is, if a player performs strategy A, the opponent should have more than one reasonable response as an option. Games frequently have a system of built-in counters. Theoretically, game creators insert such counters to avoid the danger of an optimal strategy. However, these kinds of systems often cause games to be nothing more than a fancy multimedia version of rock-paper-scissors: Unit A counters Unit B counters Unit C counters Unit A. On the other hand, excellent competitive games, such as Marvel vs Capcom 2, allow for huge diversity in response. In Marvel vs Capcom 2, players can elect three characters to form a fighting team. With fifty-six characters from which to choose, Marvel vs Capcom 2 offers over 25,000 combinations of possible teams, presenting the player with virtually unlimited options. At the highest level of play, strong competitive players can be seen using drastically different teams and styles.
Third, a good competitive game should test a player’s skills and minimize the element of chance or luck. Ideally, the probability of a weak player defeating a good player should be as close to zero as possible. For example, in a well-designed game like WarCraft 3, it is highly unlikely that an amateur will be able to control his units or respond to his opponent’s tech patterns as well as an experienced player. In fact, the best way to test a player’s skill in a game is to present the player with more decisions. In WarCraft 3, a player not only has to make major decisions, such as which buildings to make or what hero to choose, but also has to make innumerable small decisions, such as how to precisely control each unit or time an attack. By presenting a player with more decisions, the game offers amateurs more opportunities to make mistakes and experts more opportunities to shine.
However, against this framework of competitive game design, we can understand why the marginal advantage gives a game flavor and excitement for both the player and observer. The marginal advantage not only provides the player with the joy of overcoming obstacles, of finding new and more effective methods of winning, but also allows a player to express himself, to have his own unique style. By exploiting the marginal advantage, the expert player is both a problem solver and an artist. In WarCraft 3, StarCraft, Marvel vs Capcom2, and CounterStrike alike, we see the individuality of the players shine through: some play aggressively, some play defensively; some are renowned for their solid, steady play, others for their unorthodox tactics. We respect the brilliance of their expert skills; we admire their ability to win. Yet, at the same time, we appreciate the aesthetic of each player’s technique, that each player finds a solution that is so different from the next player. If too many decisions are clear cut, the player has no need to discover his own marginal advantage over the field, and the competitive game collapses into redundant, unexciting play, unappealing to master and unappealing to watch. The brilliance of a competitive game is that the designer must limit his role to be the creator and balancer, to allow for the potential of innovation. In this way, each player can uncover his own marginal advantage and become the true pioneer of the game.
Great read, very interesting. I didn't intend to read it all right now, but it sucked me in. One thing I'd like to mention is that players whose strategy is to gain and maintain the marginal advantage are almost always boring as hell to watch.
On December 31 2007 12:42 SonuvBob wrote: Let's make babies together!
Great read, very interesting. I didn't intend to read it all right now, but it sucked me in. One thing I'd like to mention is that players whose strategy is to gain and maintain the marginal advantage are almost always boring as hell to watch.
i think you are taking it a little too much to the extreme
virtually every progamer plays in the margin; the games are extremely fragile. but, again, this is why starcraft is so fun to watch. its so intense to see players battle knowing that one or two errors can cost a game, or that one or two good decisions can swing a win
I'm in disagreement with Bob. The players who take huge gambles are the most boring to watch. 3 hatch hydra all-in RARELY makes a good game vs a FE toss. Whereas a long game where you know you have the advantage and calmly out expo your opponent while reaching 200 is fine and will give you lots of big battles that will look closer to the spectators than they really are, but you know you are winning unless some screw up happens.
Well, the skill level in professional Starcraft is ridiculously high, and it has the depth to keep every match exciting. Also, in SC players tend to resign when the advantage gets to be too great.
In games where the 'greedy' approach is viable, the 'slow and steady' players tend to be less exciting to watch, though as your article says, the latter will be more likely to win. And as skill levels rise, eventually that will be the only viable option. Sorry, I can't think of many examples. If you followed the CS scene from the beginning, think about watching games from the early days vs more recent ones. When a team was basically 5 individuals with no real cohesive strategy, there was a lot more action (and lots of aces), whereas nowadays it's all about a perfectly executed strategy, with lots of waiting (the shortened timelimit kinda fixed that though). It's slower, with less aces and all that, but the game has the depth to keep it interesting for the spectators.
The games that are the most successful when it comes to spectating are those that have enough depth that the marginal advantage strategy is still fun to watch.
Testie: I'm mainly referring to the average spectator. Obviously the more knowledge you have of a game, the more interesting the smaller details will be, whereas a lot of folks would probably rather watch a hydra all-in than a close 30 minute TvT.
Hi day, long time lurker first time poster. I'd just like to say this was a great read, and now I know why I never manage to finish off my opponent when I attack him after I just won a battle.
That was a really great, solid, and concise article. I enjoyed it and agree fully. I also feel this is not limited to competitive gaming alone but to other things like most sports (though i think it applies alot to poker and not to poker at the same time, depending on situations).
Punting on 4th and inches rather than going for it.
Safety shot in 9-ball (maybe this example applies the least)
Walking Barry Bonds.
Making your star baller sit out in the first half cause he has 3 fouls.
All of these things are examples of things that are not 'fundamentally correct' if you envision the game in it's purest form...but since we are human and there are always different situations and scenarios there are a lot of cute things that must be done to maintain the marginal advantage. Though, I must say that this definitely applies in games more than sports, in sports most of the time it's all about collecting as many rocks as possible.
First, there IS in fact competitive rock, paper, scissors. I'm not sure how they do it, but I guess it's all about reading people.
Second, I don't agree that one should always try and maintain a marginal lead. Certainly you should avoid becoming sloppy and overly aggressive, but maintaining that assertive edge is what gave you the lead in the first place, so you should continue to press it.
CS really is a good example of this, because momentum plays a huge role. Everyone has amazing aim at the expert levels, but when your confidence is high your aim becomes just a bit better and your reactions a bit faster. The opposite goes for when your confidence is struggling. A team might take a 10-2 lead in the first half and win the opening pistol round, basically guaranteeing them 12 rounds (13 to win.) From then on, there's three things that could happen. They could get overly confident, push impatiently and begin to lose rounds. They could maintain what they were doing before, and they could become too passive, and begin playing not to lose.
There's no fine line between these and ultimately it comes down to whether you're successful or not. We've all seen NFL teams get a big lead and then lose their aggressiveness, ultimately allowing the other team to come back. We've also seen the Patriots go full throttle for 60 minutes each game this season, demoralizing their opponents even further. Or "dagger" three pointers in the NBA. When they go in, of course it was the right decision, but when they miss it's always a mistake. It's really impossible to see the line between too aggressive and too safe, until after the game is over which is why sports (and esports) are so fantastic.
Oh fuck I meant to post in this thread and then forgot to. Everything I have to say on this subject is already written more eloquently in this article; It was a great read.
this should go in the articles section i think, FE is specifically for starcraft progaming and involves progamers as far as i know, but then again i'm not sure...it is a worthy article though
very awesome article, well written and I agree completely. I really dont have anything to add on this subject, you covered everything already. This is so central to the concept of competitive gaming, and I think it's a major reason of why they are so much fun to play and watch. I remember clearly back in my WC3 ladder days I would feel the excitement of the marginal advantage that must be held or how my opponent was marginally ahead and I would try to see if I could take it back in any way that I can. God I love games.
Oh I also think this deserves to be in the FE, but that is not my call.
Wow, awesome article and I agree with you to the fullest. When I watch my friends play Warcraft 3 (amateurs, such as me) they do exactly what you said. After having won a battle, they often go creep or do unneccessary things instead of building up some pressure or anything.
I think you should have mentioned the inverse, the marginal disadvantage, and the kind of play it creates.
For instance, hockey teams switching their goalie for an attacker in the last minutes, Jaedong's muta-scourge rush vs Stork after losing 4 drones to zealot harass.
"First, there is no such thing as a good competitive game that has an obvious optimal strategy. For example, there is no competitive tic-tac-toe gaming, as two logical players would tie every time. I remember playing an old real-time strategy game called KKND, where one of the units was clearly the strongest unit in the game. Playing against my brother degenerated into a fray where we hurled masses of the same unit at each other until one of us got bored and stopped playing. On the other hand, a superior game like StarCraft presents a completely intractable problem. There is no best race, no best strategy, and certainly no best way to win."
Because there are optimal strategies in games like Chess and Go, and I'd venture to say that theyre some of the best competitive games in the world.
On January 03 2008 05:18 blagoonga123 wrote: The only part I didn't like about this was
"First, there is no such thing as a good competitive game that has an obvious optimal strategy. For example, there is no competitive tic-tac-toe gaming, as two logical players would tie every time. I remember playing an old real-time strategy game called KKND, where one of the units was clearly the strongest unit in the game. Playing against my brother degenerated into a fray where we hurled masses of the same unit at each other until one of us got bored and stopped playing. On the other hand, a superior game like StarCraft presents a completely intractable problem. There is no best race, no best strategy, and certainly no best way to win."
Because there are optimal strategies in games like Chess and Go, and I'd venture to say that theyre some of the best competitive games in the world.
err nono !! i think you misunderstood me.
chess and go don't have optimal strategies. that is, the game has not been "solved." the fact is, the best move is sometimes ambiguous. sometimes, there are CLEAR correct moves to make, but this does not imply an overarching correct way to play EVERY single move in the game.
if there was an absolute 100% correct move in every circumstance, THAT would be an "optimal strategy." (I'm using optimal in a mathematical sense, in that an optimal play will always at least draw all other types of play).
Yes, blagoonga, Day is using the world "optimal" in a context that means "one and only at all circumstances". Chess and GO doesn't work that way (I only played Chess on a moderately high level of the two though, bit clueless about GO).
On the article, I also can just add my support, I agree with the points you made, loud and clear. Blizzard employees should read this (I'm sure they do since we're told they keep a close-eye on TL.net anyway).
This reminded me of my games where I banged my head to the wall after doing the mistakes you describe, loosing the game because of it every now and then. I also figured this thing out on my own - isn't so hard imo - , and I always had to cool myself during games not to make hasty moves. Often you could see me having a huge advantage but instead of going for the kill - I just took the map , defended vs harass opportunities and waited for my opponent to act while expoing, waiting patiently. It isn't sparking, but it works
Great great article, but I REALLY couldn't help but nitpick one thing you said.
On December 31 2007 10:44 Day[9] wrote: On the other hand, excellent competitive games, such as Marvel vs Capcom 2, allow for huge diversity in response. In Marvel vs Capcom 2, players can elect three characters to form a fighting team. With fifty-six characters from which to choose, Marvel vs Capcom 2 offers over 25,000 combinations of possible teams, presenting the player with virtually unlimited options. At the highest level of play, strong competitive players can be seen using drastically different teams and styles.
now I know you've played MvC2 (i've seen your wcg replays where your name was striderdoom, got a kick out of that) and you cannot honestly tell me this statement is true. at tournaments theres really about 3 teams you ever see played, and about 9-10 characters that are actually used.
also a really small nitpick at the same statement.... i didn't do the math but i'd assume how you got 25,000 combinations from 56 characters means "Sentinel-Cable-Magneto" and "Cable-Sentinel-Magneto" count as different teams, wich would mean its not really 25,000.
agian i'd like to strongly emphasize that i loved the article and it was great, but having watched how actual mvc2 tournaments breakdown, i could not resist this nit-pick.
On December 31 2007 10:44 Day[9] wrote: On the other hand, excellent competitive games, such as Marvel vs Capcom 2, allow for huge diversity in response. In Marvel vs Capcom 2, players can elect three characters to form a fighting team. With fifty-six characters from which to choose, Marvel vs Capcom 2 offers over 25,000 combinations of possible teams, presenting the player with virtually unlimited options. At the highest level of play, strong competitive players can be seen using drastically different teams and styles.
now I know you've played MvC2 (i've seen your wcg replays where your name was striderdoom, got a kick out of that) and you cannot honestly tell me this statement is true. at tournaments theres really about 3 teams you ever see played, and about 9-10 characters that are actually used.
also a really small nitpick at the same statement.... i didn't do the math but i'd assume how you got 25,000 combinations from 56 characters means "Sentinel-Cable-Magneto" and "Cable-Sentinel-Magneto" count as different teams, wich would mean its not really 25,000.
agian i'd like to strongly emphasize that i loved the article and it was great, but having watched how actual mvc2 tournaments breakdown, i could not resist this nit-pick.
On December 31 2007 10:44 Day[9] wrote: On the other hand, excellent competitive games, such as Marvel vs Capcom 2, allow for huge diversity in response. In Marvel vs Capcom 2, players can elect three characters to form a fighting team. With fifty-six characters from which to choose, Marvel vs Capcom 2 offers over 25,000 combinations of possible teams, presenting the player with virtually unlimited options. At the highest level of play, strong competitive players can be seen using drastically different teams and styles.
now I know you've played MvC2 (i've seen your wcg replays where your name was striderdoom, got a kick out of that) and you cannot honestly tell me this statement is true. at tournaments theres really about 3 teams you ever see played, and about 9-10 characters that are actually used.
also a really small nitpick at the same statement.... i didn't do the math but i'd assume how you got 25,000 combinations from 56 characters means "Sentinel-Cable-Magneto" and "Cable-Sentinel-Magneto" count as different teams, wich would mean its not really 25,000.
agian i'd like to strongly emphasize that i loved the article and it was great, but having watched how actual mvc2 tournaments breakdown, i could not resist this nit-pick.
i get 25,000 by evaluating the binomial coefficient (56 choose 3) which evaluates to 56*55*54/3*2*1 = 27720, which, combinatorially, is the number of ways of having distinct subsets of 3 from a set of 56.
that is, cable-sentinel-magneto is not double counted with sentinel-cable-magneto would be.
Also, although you definitely do not see close to the 25,000 actual teams, you DO see a wide variety of different teams being used (3 is kind of an exaggeration). I mean, you could probably name a large amount yourself . Still, i liked using the number 25,000 in the article because its kind of a waste of effort to type "there are 25,000 possible teams, and although only a small subset is used, there's still a big variety." Any MvC2 player would agree that there is a great deal of diversity in teams chosen.
On January 03 2008 10:15 Woyn wrote: aha! KKND! I also played this, I still have my original copy that I picked up in a sale
omg lets play!! ahahah!!! yess!!! :D :D :D
do you remember the beetle for evolved? it did by far the most damage of any unit in the game, but it had a ridiculous cooldown rate. however, if you told it to attack a unit, then attack a different one, it would reset the cooldown. essentially, if you had a few beetles, you could attack units back and forth and it would be a constant stream of infinity-damage ooze ahahah.
Wow this is a great article. This "marginal advantage" is something I believe everyone should keep in mind while playing any game. When I first started playing SC, after fending off the initial attack, I'd always take everything I had and attack the enemy base. My army was always either in my base or my enemy's base. It took me almost a year to realize that it's ok (or should I say ideal) to control the center and take expos. Thanks for your insight Day
Excellent remarks. I'd just like to point out that one of the biggest things that separates A-class player from others is their ability to maintain and extend their advantages. This does not necessarily mean they are cautious and don't attack. For example, Iris and July like to use their unit advantage to wreck the other player. Other players prefer being able to secure more expos instead of pushing their attack.
StarCraft has a strong inherent defender's advantage, usually due to travel distance, cliffs, positioning, etc. so there are often games when a player takes a moderate advantage early game or midgame and builds on that advantage for another 10 or 15 minutes before winning. The case is most extreme in PvP or ZvZ where a build order advantage is often enough for a skilled progamer to win.
S-class players are those that are able to overcome disadvantages. Jaedong's ZvZ is a good example. Regardless of what build he uses, he's able to maximize his position and ultimately win most of the time.
What a great write up. Throughly enjoyed reading that. The ending especially was pretty profound. It was something that all game developers should keep in mind as they create their games.
Furthermore, I believe that this philosophy should be keep in the back of the minds of everybody that plays games competitively.
While reading this, I realised that you can apply this to more than just a handful of well designed computer games. Take cricket for example: If you take a wicket, your next objective is to try to use the advantage you've gained by taking the next wicket while the batsman is new to the crease. I'm sure the philosophy could be applied to even more diverse fields than that.
On January 03 2008 10:15 Woyn wrote: aha! KKND! I also played this, I still have my original copy that I picked up in a sale
omg lets play!! ahahah!!! yess!!! :D :D :D
do you remember the beetle for evolved? it did by far the most damage of any unit in the game, but it had a ridiculous cooldown rate. however, if you told it to attack a unit, then attack a different one, it would reset the cooldown. essentially, if you had a few beetles, you could attack units back and forth and it would be a constant stream of infinity-damage ooze ahahah.
I will reinstall it when I am home from work and hopefully it will run fine. Hamachi!! yes I remember the beetle, unfortunately I was never smart enough to work out that you could reset the cooldown like that
I love the article. Being a competitive MvC2 player as well among other games you mentioned (since I play MvC2 best) I love how you clearly put your thoughts about marginal advantage, Playing to win rather than Playing to win while looking Good <3
Edit: About that thing in the second page with disarray, "3 teams... etc", + Show Spoiler +
while its exaggerated although it holds truth in it (Teams usually consists of two Top tiers + one good assist) There are still players who can consistently play well against the best players because of the dynamics of their team as well as the assists they choose (example: You play MSS-proj quite differently from MSS-ground )
Any low tier team with a good dynamic still has a good chance to beat a better team even when both players are of equal skill, This is also why most low-tier players pick Tronbonne as their primary assist. By doing this they give themself an advantage they wouldn't normally have.
But enough MvC2, this is a SC site. I can't really type well at 1:30 am here but I hope I made even a small point across =)
Wow, I really need to think about this kind of thing more often. Of course, I probably won't improve all that much just by reading it, because I have problems gaining any kind of advantage, including a marginal one
Ah, it was the beetle that was heinously overpowered in KKND one (I was hoping it would have turned out to be the autocannon tank. That tank was fucking badarse). I merely remember beating the whole thing in a desperate attempt to find some of the special units they hid in the game. That said, the article makes it's point strongly and I agree with it totally. The safe way may not always be the most exciting, but it's those who grasp the marginal advantage and play with it carefully who win more often than not.
I know northing about Mancala, but it sounds to me as if a more powerful strategy (more powerful than the 1 stone advantage) would be one that strives to always have the largest advantage after the opponents turn, but at least 1 stone.
So it wouldn't aim to make the greediest move, but a move that builds a steady advantage if possible and otherwise maintains the advantage.
On January 06 2008 12:17 Simplistik wrote: I know northing about Mancala, but it sounds to me as if a more powerful strategy (more powerful than the 1 stone advantage) would be one that strives to always have the largest advantage after the opponents turn, but at least 1 stone.
So it wouldn't aim to make the greediest move, but a move that builds a steady advantage if possible and otherwise maintains the advantage.
yes you are totally correct. this is essentially what the winning program did. rather than say this explicitly, i just wrote it in a very brief way since it was unnecessary to go into too much detail. (plus, there are so many billions of variations of mancala that it would depend on what version you were playing haha)
what's important is that video games >> board games ;] hahahah
It seems to me as if the "winning" program was simply examing more future moves than the "greedy" program. It looked at the possible enemy responses after all. In chess programs the program gets stronger the further and more completely it looks into the "future" of the game. The same should be true for Mancala I guess.
Great article, You got me thinking about game analysis and I would like to share some thoughts for your consideration. Initially I will be a bit abstract, but please bear with me. My points are eventually worth it (i think).
From the minimax theorem, a players best strategy is the strategy that minimizes potential loses against all opposing strategies. Mencala is a game of perfect knowledge. Here, every aspect of the game is visible to each player. This makes the best strategy easier to choose. However, unlike Mencala, Starcraft is a game of imperfect knowledge. Certain aspects of your opponents position and strategy are hidden from you for at least part of the game. Thus, you cannot always select the best counter to what your opponent is doing. You cannot be in your opponent's head visualizing the strategy he will choose. Thus, your strategy will not always be the best because it will not always minimize potential losses against your opponents strategy.
The players that do the best in Starcraft are the ones who scout incredibly well. Good scouting increases knowledge of your opponents strategy/position allowing you to select a suitable counter. However, there are still times you do not know what your opponent is doing especially in the early game. This leads me to ask what characteristics should we look for in early game strategies especially when scouting is limited due to blocks/map distance etc...
Initially I believed conservative strategies such as 2 gate >> obs are great starting builds simply because these strategies minimize initial potential losses. However, the eventual resulting potential losses from conservative strategies can be tremendous. Considering 2gate >> obs when facing a terran who quick expos. The p player's strategy puts him significantly behind in economy which often leads to a stomping generated by the terran player's superior economy.
Accepting the premise that obtaining and exploiting a marginal advantage should lead to victory, the best strategies in games of imperfect knowledge are the strategies that are most likely to give you a marginal advantage. This is why FE builds often lead to victory especially on larger maps. Thus when choosing strategies, we should look for strategies that are likely create marginal advantages.
The best games, in my opinion, are the ones with very subtle marginal advantages. Boxer's recent games on Blue Storm are great examples. Boxer took superior position (in one game with marines/medics, in other with vults/mines), and his following game play allowed him to exploit that position. He may have been slightly behind on economy initially, but his position was his marginal advantage which he exploited fully.
Keep in mind that randomness is part of Starcraft. You'll never know exactly what you should have done until you see the replay, but I hope this little bit helps when it comes to choosing strategies.
One has to remember that marginal advantage isn't completely in the hands of the game. The player himself has a lot to do with marginal advantage.
For example, players like July can turn what would be a marginal disadvantage for most players to an even ground or a marginal advantage. While Starcraft can be split up into essentially three areas where you can have advantage - economy, tactics and army size, players like July can sacrifice one of the three more readily. 8 Zerglings vs. 6 Marine/2 Medic could be Zergling blood for many players, but it could be easy pickings for someone like July.
Marginal advantage in one area can often be "traded off" or led to into an advantage for some other area. Like BoxeR's TvTs on Blue Storm - while he may not expand and could be a base behind the game, his Vults/Mines or his M&M hold the center so that he can expand if he needs to. In this case, his tactical advantage leads directly into an economy advantage. This economy advantage could then supplement his tactical advantage or both tactics and economy could supplement an army advantage, which is common in his play now.
A true good player knows how advantageous of a position he is over his player, and by allocating that advantage between the three different types, he can push the lead.
While the defender may have an advantage in Starcraft by virtue of travelling distance and "home field advantage" and such, the player is not to be discounted in the equation. By putting pressure on a player early game, the attacker gets the marginal advantage in that area - he is far more composed and can do more with his other advantages than the opponent. While the defender has to concentrate on fending off the attack and rebuilding any losses, the attacker (or pressurer, to be more precise) can continue to do what he would do in his base.
This leads to the difference between pressure and attack. Pressure is meant to trade advantage. It is, in example, the way to trade off an army size advantage for an economical or a tactical advantage. If you sneak 12 Zerglings into a base and kill off 6 SCVs, you've swapped your army for their econ. Perhaps, also, you are forcing their otherwise-attacking force to return to defend, also gaining you a tactical advantage. Attacking is the end of trades. You trade all three of your advantages to secure a victory. For example, take the 4 Pool Zergling rush. Your economy is automatically at a disadvantage as you are not building workers. Your army size is at a huge advantage by virtue of the 4-Drone Spawning Pool. If you fail to use this army advantage to destroy their economic advantage, then the player will have a very hard time to respond.
Aside from the starcraft side of that article, the interesting thing that I'd like to point out, is that Mancala is a game in which the first player should always win if both players play perfectly. In fact, I just solved it on my computer just now using one of my professor's research projects called GAMESMAN, which is a brute-force game solver that determines all possible positions (the current player's turn being part of that position) in a game and links each position to another position, until it reaches a winning or losing (or draw-ing) final position.
If you're interested.... here's a link to the site.
Summed up many of the thoughts I have gone over when I thought about competitive gaming. Basically I always thought when there is an option and one will have a better outcome than the other you have the basis for competitive gaming. The more options you get the more greatly talented people can distinct themselves from others.
there is a similar concept in chess, which is called maintaining the equilibrium.
the idea is that at low levels, two players are both playing poorly so to do well, each player needs to think about making the best move.
but at high levels, both players should assume that the other player is playing perfectly. in this case, when either player tries to force things to happen, they are actually overextending themselves and allowing weaknesses to flow into their player. at advanced levels where play is very close to perfect, a player can only win by taking advantage of another player's mistakes. in a chess game of 30 moves, even best players make very small mistakes maybe a few times a game.
if a player takes advantage of his opponents mistakes those few times, he builds a marginal advantage which can later accelerate and accumulate into a large enough advantage to win in the end game. otherwise, if a player tries to force a win when his opponent hasn't made a mistake, then the player only overexerts himself and gives his opponent the opportunity to take advantage of his mistake.
On January 06 2008 14:17 Simplistik wrote: Hehe, and to talk about this a little further...
It seems to me as if the "winning" program was simply examing more future moves than the "greedy" program. It looked at the possible enemy responses after all. In chess programs the program gets stronger the further and more completely it looks into the "future" of the game. The same should be true for Mancala I guess.
that's not completely true. the program that examines more moves isn't the program that will win.
in chess programs for example, for the most part until a few years ago, the program that can analyze the most positions, and analyze them most correctly, will have the highest winning percentage.
this changed in the past few years when a chess AI programmer paired up with a chess international grandmaster and instead of brute forcing future variations, they applied chess principles into the chess program. so the new chess program thinks like a person would think. with this shortcut, the program would analyze less variations than other brute force programs, but the variations it did analyze were the ones that were the best ones.
if you talk to chess players too, they will say that grandmasters for the most part, unless things get really complicated, tend not to analyze that many moves forward. rather, they have a very good knowledge of how the game will flow, and instead of analyzing move, they analyze the concept of the current game.
i love chess. it is such a beautiful study of life.
On May 12 2011 03:10 AsianEcksDragon wrote: What if I play Grubby in War3 a million times? Wouldn't I eventually take a game when I get really lucky with RNG and loots?
Odds are, he'd probably lag out of a game after 1 million games and you'd receive a victory.
EDIT: On topic, I think this is related to the BW vs. SC2 debate a little bit...it just seems to me that too often any advantage you take in SC2 instantly becomes way more than marginal.
Yeah, I found this via searching for Day9, and this is an absolutely fantastic article, it really applies to Guild Wars as well. Really brilliant analysis of game design.
On May 12 2011 03:10 AsianEcksDragon wrote: What if I play Grubby in War3 a million times? Wouldn't I eventually take a game when I get really lucky with RNG and loots?
After 1 million games you would most likely be a very very very good player. Which could take something like 2/10 or 1/10 games from him. Especially since you have been playing (training) with a that good player.
I love this article. One of the things I frequently do when thinking about what makes a game like Starcraft great is try to draw parallels to other different games and real life situations. Most people will be familiar with the comparison of starcraft and poker, and I think this article articulated really well what one of the common uniting factors is that makes these things endlessly interesting and what makes for a consistent winner instead of a flash in the pan.
Wow the commenters on the first page of this article is almost as impressive as the article itself T.T
Great read tho, loved the mancala analogy to other games. My question would be, is there a way to incorporate that mancala strategy into real life? (mancala games or elsewhere )
thinking poker, exploiting every advantage EV > 0, may be what you're talking about, but in the case of mixed strategy games such as starcraft what constitutes a marginal advantage may be ambiguous. some players aim for a small advantage in each game, while other players aim to win games by huge margins. i would expect the most successful player to be one with both strong standard play and capable of loose aggressive play, that is, capable of winning or losing games by huge margins.
for instance, i think incontrol's brood war play style could be described as 'greedy, winning or losing by huge margins', but this proved effective in competitive play to the effect that he won a wcg usa finals.
while probably implied by what you wrote, i'm not sure the above is completely obvious
man sean, i dont usually read shit thats this long on forums, usually because it is full of dumb shit that doesnt intrust me. This caught my attention thought and entertained my short attention span. awesome read thanks.
This was an excellent article, aside from the MvC2 reference. In games like that, teams will usually have a solid idea of what works, and what doesn't. They rarely just say, "You play your favourite character, and I'll use this guy because he is cool." If they want to go anywhere, they'll have to think, "Hey, X goes with Y because of passive-ability Z." Apart from that, the article was well written and thought out.
Fantastic article Day[9] I found it a very interesting read and completely agree with everything you say. However, I also think there is a 4th dynamic that you have not included and it is just as important as the three you previously mentioned and that is of aesthetics.
Allow me to explain my thinking in simple terms.
Take chess as an example.
We currently play chess with 2 sets of pieces which are easy to differentiate because of the way they are coloured. If those 2 sets of pieces were of similar colour it would make the game incredibly frustrating to play because it would be much harder to determine your position and consequently your strategy.
So whilst I totally agree with everything you say I also do not believe it to be the be all and end all when it comes to competitive game design.
Of course I could give many more examples and would be happy to flesh out my thoughts should you find this avenue of conversation appealing.
Ahhhh can't believe I missed this thread the first time around!
Fantastic read... and I think a lot of these things apply to SC2.
The entire section about counters and a fancy multimedia rock-paper-scissors really rings true, and I always find myself wishing I had more options in sc2 to counter whatever my opponent is doing, but it seems you have to have the right unit composition to win and must follow the game's built-in counters instead of having these choices
I don't know how so many people could've missed this, including me! This really needs to be shown to the masses, even if it's some years old now! I found it by browsing reddit. A really interesting and well-written article.
Hm, Day's criticism of specialized counters falling into rock-paper-scissors (the second main argument) sounds pretty familiar...iirc, there was a lot of ire on the SC2beta forums about SC2 being a game of hard-counters, whereas BW was a game of soft(er)-counters. The premise here is that a game of soft-counters allows for greater ambiguity with respect to optimal play.
His third main argument here sounds shockingly like Idra's argument during the now-infamous day9/idra SotG discussion.
This is basically my own thoughts on competetive gaming written down on. I specifically liked the part that explains how there are no degrees of winning and that a riskless small victory is usually the most guaranteed victory. Too few people understand this in my opinion, and this is equally the case in most forms of competition. Really good article, I love what you do
A good post, and I mostly agree. Something you reference but don't quite focus on is the concept of the "big win"--strategy aside, it's human nature to want to finish strong. Even pros are not immune to the allure that comes with an ego
This gives a great perspective and understanding of games and the theory behind them! You're giving me more material to write another song Sean ^_^. Thank you for sharing this, it gave me a lot more interesting theories to discuss.
Great article, but I think there is one thing to keep in mind that looking at computers play that makes things different. The psychological impact of winning big.
Humans have emotions, computers do not. A crushing defeat is definitely more strongly felt than a close one, and getting demoralized can seriously affect ones performance in later parts. Obviously part of being a pro is dealing with psychological effects (this is why top pros, in tennis for example, are pretty calm in terms of never being too angry/negative, but they are always there)
Great read for someone trying to learn to be more competitive like me. This article really shows the decision making difference between a pro and a newbie. I am defiantly going to try and become more comfortable playing with a marginal advantage rather than trying to all in an opponent once I get a lead. As always day9 you are a beacon of knowledge to us newbies.
While I mostly agree with the article I would put a greater emphasis upon foresight than a marginal advantage. While in some sc2 games the match is decided from the very beginning of the game I would argue its more important from a strategical point of view to analyze the games in which someone was behind and yet still won. This is based upon a tradeoff of advantage early on for advantage later (ie having an army *just* big enough to hold of the opponents push but droning like a boss). A marginal advantage is important, but being able to predict when one is actually necessary, is vital.
Wow. What a gem of an article. This is a concept I've been trying to be aware of in my play, and you've expressed it so much more clearly than I ever could have.
Some people can articulate their thoughts so well! Not me. I try to explain some simple little thing, such as the big bang theory, and all that comes out is a bunch of "Um," "Like," and "Yeah's". Naaaaaaah, It's not that bad. teehee.
"There is no best race, no best strategy, and certainly no best way to win."
BRB 1/1/1 vs 99.9% of protoss just sayin.
I would argue that this is indeed a "best strategy" in almost any given situation against virtually all builds protoss has at their disposal at this point.
btw I was a bronze protoss for a while but I just got promoted to silver 2 weeks ago and have experience with the game.
I myself have played a wide variety of competitive games and have found myself wondering what makes a great competitve game quite often. I've come to the same conclusions you have, except marginal advantage. Very good read. To pull from an experience in my own life where I was beaten by an expert employing marginal advantage, I was playing against arguably the best super smash brothers brawl (which is an exeptional competitive game) player in the world at a tournament last year. His "tag" is ally. He barely beat me, but he beat me every time. I played him multiple times for fun after he knocked me out of the tournament and he beat me literally every time by a pretty small margin. Facinating that I am only now realizing how he was beating me, from the most unlikely source.
Hum ... If I'm not mistaken, this idea of game theory has already been stated and used in economics. I think it goes by the name of the Nash Equilibrium! (http://en.wikipedia.org/wiki/Nash_equilibrium)
Anyway thanks for the nice read and looking forward as usual to your excellent casting
This is not nash equilibrium - but it is closley related.
Nash equilibra is about stability, marginal advantage is about constantly pushing a dynamic situation ... eg by getting through lights on amber by not being a sunday driver .... gets a marginal advantage that many say is insignificant yet get me into work on time ;p there is also the problem of omniscience in the nash equilibrium. they both meet, but there is a lot of unknown data in sc2 ... maybe you can argue differently for bw as its been around for a long time.
really what this is is a description of game that consists of many different games but moves that can take effct in many of those games. you get marginal advantage by prioritising some games over others.
so by playing a move you effect some of the games but not others, the assumption is that you are always making good moves and thereby are always keeping the idea of equilibrium as Nash stated it, however you can focus on some games over others and so bias yourself one way or another ... these small adjustments in where to put the effort are what marginal advantage is.
It almost made it to the top of hacker news the other day which was kinda cool to see.
What an amazing article that applies as much today as it did when day9 first wrote it. Please mods, let it stay on the front page for a bit. I think a lot of us could use the read .
On August 16 2011 09:52 Losacio wrote: "There is no best race, no best strategy, and certainly no best way to win."
BRB 1/1/1 vs 99.9% of protoss just sayin.
I would argue that this is indeed a "best strategy" in almost any given situation against virtually all builds protoss has at their disposal at this point.
btw I was a bronze protoss for a while but I just got promoted to silver 2 weeks ago and have experience with the game.
1/1/1 isn't the best. Protoss just aren't sure how to handle it. Not trolling.
For any given strategy, there is an effective way to beat it. There is no strategy which is so much better than all possible strategies, known and unknown, proven and unproven, likely and unlikely, that it can not be beat.
For any given strategy, there is an effective way to beat it. There is no strategy which is so much better than all possible strategies, known and unknown, proven and unproven, likely and unlikely, that it can not be beat.
In a game that is closed to balanced, yes. But imagine that marines cost 1 mineral and no supply. The best strategy would be to only make marines, no exceptions.
I'm not saying 1/1/1 is imbalanced either, just that there can exist universally optimal strategies in badly designed games and 1/1/1 has the potential to be one if we see no metagame movement in a year+.
The funny thing is that if you take this article as an authority (which I agree with it fully), it shows that SC2 is not as strong a competitive game as BW, for example.
One of the major factors is that a weaker player will almost never be able to beat a stronger one. This happens quite often in SC2, and almost never in BW.
Using my own experience playing in the CSL in BW, as a B- level player, I won almost 100% of my CSL games (average skill level in the CSL was probably C- / C). Looking at the top foreigners in tournaments, the same X people (Ret, IdrA, Draco, IefNaij, Mondragon, White-Ra) at the end of the game would be Top 3 in some order almost EVERY single time.
In SC2 you have much more variability. Could be because the game is young and hasn't developed fully yet, or it could be a real flaw in the game....
On October 06 2011 03:35 Xeris wrote: The funny thing is that if you take this article as an authority (which I agree with it fully), it shows that SC2 is not as strong a competitive game as BW, for example.
One of the major factors is that a weaker player will almost never be able to beat a stronger one. This happens quite often in SC2, and almost never in BW.
Using my own experience playing in the CSL in BW, as a B- level player, I won almost 100% of my CSL games (average skill level in the CSL was probably C- / C). Looking at the top foreigners in tournaments, the same X people (Ret, IdrA, Draco, IefNaij, Mondragon, White-Ra) at the end of the game would be Top 3 in some order almost EVERY single time.
In SC2 you have much more variability. Could be because the game is young and hasn't developed fully yet, or it could be a real flaw in the game....
I happen to think it's the latter.
I think you're right and SC2 has flaws in game design. But i don't think we should be pessimistic. With such excellent articles and the voice of top players the community has the ability to influence Blizzard into making a much better game with upcoming patches, but more so by removing and adding units with the expansions.
So the question now is, is SC2 at the level that the marginal advantage exists, and there is no "broken beetle" strategy? I'm inclined to say yes, but regardless, excellent read.
PS-TY to Hacker Monthly for the necro, else I'd never have noticed this.
an interesting read however i would caution against applying this to starcraft 2 at this time versus others (such as BW cs etc.) as it is not finished yet.
On October 06 2011 03:35 Xeris wrote: The funny thing is that if you take this article as an authority (which I agree with it fully), it shows that SC2 is not as strong a competitive game as BW, for example.
One of the major factors is that a weaker player will almost never be able to beat a stronger one. This happens quite often in SC2, and almost never in BW.
Using my own experience playing in the CSL in BW, as a B- level player, I won almost 100% of my CSL games (average skill level in the CSL was probably C- / C). Looking at the top foreigners in tournaments, the same X people (Ret, IdrA, Draco, IefNaij, Mondragon, White-Ra) at the end of the game would be Top 3 in some order almost EVERY single time.
In SC2 you have much more variability. Could be because the game is young and hasn't developed fully yet, or it could be a real flaw in the game....
I happen to think it's the latter.
i think that these sort of comparisions are actually useless a better comparison would be to look at the tournament results or multiplayer results for the time that starcraft vanilla was out as the game in itself had a number of issues both in design as well as in balance WOL is the same there are design flaws and imbalances when it was released and a good number of balance changes have been made since release as well as a massivee change in the map pool.
In any case once legacy of the void is released and there can be no more design changes then we can have accurate and valid discussion about the game and how a player gets and maintains an advantage. This is not to say that having discussions about how to preserve or seek an advantage shouldnt be done but they must be taken with a big bucket of salt when comparing them against or games, tests, scenarios that are either complete or are only meant to test games that are finished.
I will say one thing though, in a game as complex as Starcraft playing the minimal advantage strategy is very, VERY hard. Allow me to elaborate.
The minimal advantage strategy is only the best strategy if a few truths hold.
1. The game must be balanced. (I don't claim SC, or SC2 are balanced, or unbalanced. I'm not good enough to say.) 2. The players must have very, very good intuition and foresight of how a move will play out. 2. The players must have flawless execution.
Explaining: 1. This is obvious, if the game isn't balanced/fair you cannot opt to use a minimal advantage strategy. 2. If you can't see near perfectly how a move will play out then you cannot predict who will take the advantage. Pretty straightforward.
3. The players must have flawless execution. - Here is where it gets messy. Firstly, the theory is absolutely sound, winning no matter how close it was, is winning. But, a minimal advantage strategy makes for a longer game, naturally. It naturally leans toward a defensive follow up posture to gaining any lead. The longer a game is, the more chance you get of firstly making your own mistakes and throwing away the lead/game. And secondly, being presented with something you were either unprepared for, or have not seen before. As such you need to be flawless, because a minimal advantage is the easiest advantage to lose by mistakes in execution. Not micro-ing those 3 stalkers, overextending your marines/tanks, forgetting to inject. Any small mistake, or accumulation of very small mistakes will forfeit your minimal advantage.
As such, in some cases (no one is perfect and no one can maintain perfection throughout a series) it is actually the better choice to go for the win big strategy. Especially if you think your opponent can't counter it quick enough.
This another old thread reopened by Beyonder? Really like these. Awesome read, really emphasizes something i read Zelniq say about not taking unnecessary economic risks because a good player will punish you for overexpanding.
I really liked reading this, it was all put together very well and explained your point nicely. I am a huge fan day[9] and your incites are just wonderful to listen to and read about, keep it up
Very good article Day[9] and very true. I haven't played SC2 in a while but I remember that when I was in bronze I would try for the big marginal wins (Didn't help i had no micro and macro lol) but when I went into season 2 I went for just keeping the lead and started to cream platinum players. Most of the time I noticed they would try to go for the same type of game execution you described above. Again excelent article and love the daily. It taught me a lot and helped me back in season 2!
EPT
I'm not sure how they do it, but I guess it's all about reading people.
