Cliff notes: Is near-perfect play achievable by the best players, specifically within several years? Or do the psychological dynamics of StarCraft prevent such perfect, calculating play from being achieved?
Poll: Is perfection of SC ability within feasible grasp? (Vote): Yes (Vote): No
The Premise First and foremost, being the undisputed best player in the world does not mean you play perfectly. Perfection, for the sake of this thread, refers to play that is near the level of what programmed AIs can do (aka never leave idle peons at all, never miss a small dot on the minimap, etc.). It's probably physically impossible to play on such a mathematically-precise level, so don't nitpick that. What I am referring to is the ability to never leave a single building unproductive when you should be macroing, never accidentally leaving a tank here and there uncontrolled when retreating, full grasp of the SC AI, etc. Obviously miscontrols are granted, since we are human, but those are miniscule and don't have a huge effect on the game (losing units here and there as opposed to accidentally pressing "i" when trying to burrow 12 lurkers, and then watching them all die).
Think at the height of BoxeR's reign, years ago. He amazed us with brilliant strategies, godlike micro control, and so on. Compare that play with the play of a top Terran player today, and you'll notice huge differences - the players today are obviously, and expectedly, on a different level. An example would be July's Mutalisk control that allows him to minimize damage yet maximize efficiency, which is a huge boost to Zerg players who are able to use this technique efficiently.
Yes: The Feasibility of Attaining Perfect Skill Older players, namely GiYoM, cite that they cannot keep up with the newer breeds of gamers because they are simply too good. These players are noticeably quicker and sharper. I believe that the game evolves due to these new players, displaying fresh talent and exhibiting game-changing techniques and strategy.
Is it feasible, then, that a level of play close to "as optimal as humanly possible" is within reach (in a few years)? Being able to carry out micro and macro controls perfectly may require being much quicker, but is this not possible? NaDa was averaging at least 350 APM per game at the OSL Finals. Four hundred apm is not an unforeseen occurrence, could players be branching into 500+ later on? And before anyone flames the usage of APM, such speed is obviously required in order to move your units, queue up more units at your factories, move your units again, rally the new scv towards a mineral field, etc. - in other words speed is required for perfect play.
StarCraft is 8 years old now, but most of the game-changing revelations have only started to appear since 2001. Builds have become faster while still maintaining economic stability, people have been discovering ways to use certain units to an exponentially higher degree of efficiency.
In 2010, will we be seeing two players at the OSL Finals sponsored by Google, with hundred k payouts, playing SC at blazing fast speeds, making absolutely no mistake in their macro, their micro at the pinnacle of excellence, never making an economic mistake like leaving 1-2 scvs idle?
No: The Dynamics of StarCraft Deny this Theory While there are many cookie-cutter builds and, for the most part you'll see the same units used in each specific matchup (i.e. Terran Mechanic vs Protoss), what has kept SC interesting over the past few years is that it's completely dynamic. Players find builds that deviate from the norm (proxy builds), the leagues constantly refresh their map pool, the entire metagame changes with each player and period of time.
StarCraft is an intensely psychological game. A few years ago, it was rare to scout some obscure corner near our main for a hidden Barracks or Pylon. Nowadays, that threat of proxy cheese is present - players are trained to scout for such builds, especially if they are inclined to believe that specific player is willing to make such a move on whatever specific map that proxying is feasible on. BoxeR successfully bunkered YellOw three times in their matchup at a league (the specific one I forget at the moment) - the very next time they met, YellOw planned and prepared and successfully countered BoxeR's attempts at winning with rush-oriented manuevers.
The maps change every so often, but I don't think this has that large of a bearing on the perfect play theory since this is under the decision of the respective leagues, not part of StarCraft the game itself.
It may also be implausible to even achieve that sort of superhuman play. Of course we can't play mathematically precise like computers, doing so would require more than 2 hands, but where do you draw the line between what is humanly possible and what isn't? (I suppose this is measured by the APM a human can sustain over several games, consistently, and in games that last more than 15 minutes)
StarCraft and Poker: Measuring Your Ability To those familiar with poker, you'll know that it's a game that includes many unknowns - specifically what your opponent's cards are, and the cards to come. However, top players still have the ability (in some to most cases) to play mathematically optimal based on many factors. In fixed limit hold'em, a person can judge how much money they could have really won by going over their session and counting the amount of bets (money wagered) they could have earned or saved (combined is the total amount of money they could have ended up with had they played optimally). This comes from looking at your play in one hand, deducing that had you put out a wager on the final street, your opponent would have called (beyond a reasonable doubt), then you would have acquired one more "value" bet.
Thus, a person is able to calculate that they have "missed x number of bets this session" (of course this has no bearing on when extreme variables are put into place - suck outs and whatnot). StarCraft lacks these "unknowns", but it's also hard to mathematically judge how optimal you have played. However, by viewing a replay, you can still judge whether or not optimal play was achieved, where mistakes (In Poker, a mistake would be failing to value bet. In SC, it would be poor control of your Dragoons and their positioning, etc.) are minimized as much as humanly possible.
Each time you can eliminate a mistake found in your game, you come closer to the level of optimal play. For the usual player, this might be hundreds of mistakes, but it must certainly be smaller for the best pro gamers on their best days, in their best matchups and maps. There is an obvious linear correlation between eliminating your mistakes and becoming a better player - thus if pro gamers are able to eliminate the mistakes they still make, they are coming that much closer to that pinnacle of perfect play.
Conclusion Today's Pro Gamers have been stepping over bounds of strategic ingenuity and efficiency in management. As players begin to get better, there certainly must be a point where their play becomes perfect. While computer-like mathematical perfection is unobtainable, there is a point where you can play as optimal as humanly possible. Is this point within reach of our best players (and even some unknown rising stars...) by 2010? Or do the dynamics of the greatest RTS game prevent this because of certain intangible factors, like play based on psychological reasoning, situational developments brought upon by the specific map, or the human ability to err, that cannot be accounted for?
dont think so cuz players are continuously pushing the boundaries of what is humanely possible so everytime something is achieved you aim a little higher and so on. again, how do you define that fine line between human and machine? even the pathing units take when moving from a to b CAN be manually moved by right clicking madly so that the unit doesnt make any stupid round abouts and such and just omfg so much, but will it be achieved, and when we DO achieve this we will look upon it as already conquered and then move on to new heights like split second timing or something.
I have been thinking about this a lot, especially in the past year to 18 months where many of the current build orders have been developed. Progamers are planning so much further into the game with their orders now that a lot of what is happening seems very formulaic. For instance, fast nexus pvz becoming as common as 12 pool in zvt. I think as that trend continues, it fossilizes the path to "perfect" play.
That being said, the current progamers are nowhere near perfect. The real way to study this is to watch FPVods. There are still many idle workers, high money counts, and sloppy mistakes. There is a lot of room for improvement, but even with the fastest players I dont know physically if they would be capable of doing it. If they do, it will not be with terran, because nobody can be that fast. I believe it would have to be with zerg or protoss.
I hope that the answer to your poll is no, because it is these flaws, these necessary flaws, that give sc games their characters. It is the decisions that each player must make, sacrificing micro for macro or visa versa, that give games their entertainment value. You talk about apm, well people who say apm doesnt matter are really wrong. The faster your meaningful clicks, the closer to the perfection idea you come to.
I dont know if I expressed myself well enough for what i was thinking about, but I really think this was well written, and follows a train of thought I have had many times. Nice job.
if it WAS possible, there would be super strats. but because the opponent can always react to what you do in many different ways, there is no such thing as perfection.
Poker.... yeah... everything I wanted to say has been said. Just like there is no perfect poker player, there will never be a perfect SC player. Plus, even if perfection is reached.. what happens when two perfect players collide? an Imperfect game.
I previously wrote a post maybe a year ago about my theory of what perfection would be, let me dig it up..
To sum it up briefly though: no, humans are NOWHERE near Starcraft perfection, progamers are lightyears away, and a "perfect" AI would require an estimated 1000APM.
Simple tasks like these would completely imbalance the game:
a) Goliath / Tank + Dropship (sunkens become useless and become practically invulnerable to most units) b) Dropships + Units in them become way too strong c) Lurkers become useless becuase Marines + Medic dodge every lurker shot d) Dragoons / Reavers / Shuttle become useless due to Ghost + Lockdown e) High Templar goes way down in value vs computers becuase units can spread out in all directions easily (same goes for vs Lurker f) The Macro of computers would be incredible (around 1.4x iloveoov, at least), and they can battle in 4 locations easily (can have one dropship cosntantly flying around and doing damage while moving flawlessly around the map with a split force perfectly microed and moving out with another, while using 1 dropship + 1-2 goliath to destroy a zerg expansion), whilst macroing perfectly at home with 5 scvs patroling the map (not patroling, but actually actively scouting) for counterattacks and potential expansion sites.
oh yeah i remember reading exalteds post on hat a while ago it's kinda true it's impossible and it REALLY would change the whole balance and dynamics of the game
On November 21 2006 05:17 exalted wrote: Simple tasks like these would completely imbalance the game:
a) Goliath / Tank + Dropship (sunkens become useless and become practically invulnerable to most units) b) Dropships + Units in them become way too strong c) Lurkers become useless becuase Marines + Medic dodge every lurker shot d) Dragoons / Reavers / Shuttle become useless due to Ghost + Lockdown e) High Templar goes way down in value vs computers becuase units can spread out in all directions easily (same goes for vs Lurker f) The Macro of computers would be incredible (around 1.4x iloveoov, at least), and they can battle in 4 locations easily (can have one dropship cosntantly flying around and doing damage while moving flawlessly around the map with a split force perfectly microed and moving out with another, while using 1 dropship + 1-2 goliath to destroy a zerg expansion), whilst macroing perfectly at home with 5 scvs patroling the map (not patroling, but actually actively scouting) for counterattacks and potential expansion sites.
It would be ridiculous, to put it lightly.
Yeah, that's assuming it even goes past the 5 minute mark with perfect micro. I can't even imagine how ridiculous it'll be
"In 2010 (or whatever), will we be seeing two players at the OSL Finals sponsored by Google, with hundred k payouts, playing SC at blazing fast speeds, making absolutely no mistake in their macro, their micro at the pinnacle of excellence, never making an economic mistake like leaving 1-2 scvs idle?"
Cute.. lol, mind you between the players that are actually earning decent money it is already getting like this, so it might not be 2010, but earlier - expect for the word "never". Errare est Humanun - to err is human. Also important is that in any competition sport, the great mistakes, the great blunders are equally important when it comes to making games memorable and matches exciting. Pure perfection is beautiful.. but perfection mixed in with the uncertainty of blunders makes watching a top StarCraft game more fun!
Constructive criticism: You should have left the poll out though (it is very biased given that you leave a well thought out opinion!). Your conclusion section is weak i.e. it wonders off too much the main point of your write up. Fix it. Make the current conclusion with a different title "StarCraft & Poker: unknowns" or something in those lines plus elaborate more on the critical StarCraft uncertainty factors - and then add a final title "Conclusions" with a synthesis of the points you are trying to make with your article. Do it in honor of perfection!
An interesting thing to try is playing at slowest speed vs comp or human. This takes a very long time, and isn't actually very fun during the playing part, but you get to see what you are ultimately capable of given a MUCH higher action per (game) minute possibility.
This does not make your strategy any stronger and in many cases your sense of timing with strategy would have to be relearned in order to "perfect" your game sensability. Also patience and persistance in this setting will ultimately do more to help win than anything. Single Player vs computers lets you dynamicly change the speed, so this is a bit more bearable during the beginning parts of the game.
You can do some really bizzare moves with shuttles (making them spin in circles while moving sideways around the map) and other accelerate/decelerate units (vulture, wraith, etc). Play a game at slowest and then sit back and watch the gosu first person replay as you manage the entire game at 700+ apm and extreme control. Then show your friends and dont tell them what you did.
GoaLD50 - I did the same thing : p Funny how weak lings can get when you can perfectly control the healing of firebats with ur medics, and dancing with single marines : )
When I tried to watch replay, it kind of screwed tho, as it changed speed from time to time.
As for perfection, I like the post by exalted. Perfect micro could change unit efficiency dramatically. As for the human level. I think players like NaDa are doing pretty good in their prime, but dunno how far they are from human limit .... Would be interesting to see if Starcraft was as popular in the world as soccer ...... how far from NaDa today would the best players be?
Who would be the Pelè of Starcraft? edit: or Garry Kasparov : p
- Mani, I must admit I haven't seen any FPVods in a while (I quit SC 4 years ago so I only watch recreationally), but I had assumed the careless mistakes I've seen made back then weren't as typical nowadays, but I guess I'm wrong .
- Exalted, Didn't see that post the first time around, but you do bring up some good points. I would like to say that SC gameplay at its current level is far beyond what anyone could have imagined plausible 4 years ago, but I have no solid evidence or reference to games to back that up.
- Physician, thanks for the great constructive criticism. I'll keep it in mind but I had no plans for this to be a big, in-depth research article or anything of that sort. This was just something I had in the back of my head for a week or so, and decided to finally write it up last night at 4 AM under terrible conditions. My conclusion was simply the last idea I wanted to get across, I didn't really think of wrapping it up as a whole since I thought it was getting long and I wanted to hurry up and finish. You're right though, and I'll change it, along with editing out some of the less-eloquent parts of the article . Thanks a lot, that was good feedback.
One point you do bring up also reminds me of another reason why perfection is not attainable: Players in any game environment strive to capitalize on their opponents' mistakes. I can gain a huge advantage placing hidden expansions around the map against an opponent whom I know is bad at scouting. If it appears as if my opponent makes huge blunders in his micromanagement, I might try to capitalize on this by attacking and harassing more often, knowing that each manuever on my part has an even greater effect than it should.
In a "perfect" game, mistakes are inevitable (a concession I gave in the OP since miscontrols are part of the game), so their opponents will tend to open up those wounds. However, miscontrols are random, unforeseeable, and thus cannot be accounted for in a mathematically optimal game - if one perfect player were to make suddenly make 20 (random arbitrary number) small blunders in one game, but average 5 in many other games, he may lose that 20-mistake game because of this unaccountable variance that wasn't necessarily due to his inferiority as a player.
I think I'm rambling here and I'm not as concise as I should be, but I hope you guys get the point.
- GoaLD50, I believe [KiD]ReD made a thread about such a game, I think it was a huge thread with a lot of results. Two people played for over an hour in a 1v1 slowest game in order to capitalize on lessening mistakes in micro and macro due to inefficient speed.
I think most of us are going to agree that near-perfect play is unattainable. For the sake of keeping this discussion going, how do you think StarCraft will evolve in the future? So far we've seen an evolution in quickness and efficiency, calculating precision based on timing, and also many proxy/rush-oriented builds. The proxies of today are part of the metagame in order to counter most tech-oriented builds that weren't prepared for such a rush, in a few years how will this scene change?
Because starcraft is a "real time" game, it would seem APM will always be the limiting factor in terms of the perfection you describe. This very dynamic aspect is what makes the game so fun. I guess one could argue that a turn-based game offers a more realistic "perfect" game, but its static play can be somewhat dull. Chess comes to mind... Kasparov prob comes closest to being that "perfect gamer". Having as much as 1+ hour per move would be an interesting apm.
Chess players nowadays often play fast games (called blitz?). So guess they got some CPM (Cognitions per minute) restrictions as well : )
Still hanging on to my soccer thought ... think a billion Indians or Chinese playing sc. What lvl and apm would the best players reach to? Guess we will never know
What is a more interesting and real question is (but that's perhaps what you meant):
Will one player become good enough to break the race balance for all time? That is, come up with builds and counters that will thwart anything provided nothing but very small mistakes are made?
These would probably still be mapspecific but still.
Take like 1 starved and 1 hungry route diversion from a default build based on what the other player started out with, that lets him counter whatever the other guy can do with close to 100% success.
Edit: So like, when will racial imbalance/maps be much more important than individual skill? It's a large part of the outcome already today.
I think it will be imposible for a human to reach what you call "perfection". Its like everybody else said, physically impossible. Plus, under pressure, sooner or later *somebody* is going to make a mistake, maybe not micro/macro, but thinking wise; i dont know if you get what i mean.
You should take in mind that ultimately, DECISIONS make winners, and losers.