Chess Perfect Game
Desmond Hutchins
A perfect game of chess is always a draw. This is what I have discovered when matching Houdini against Houdini. If you match a very strong player against an equally strong player, the result will likely be a draw.
Chess is a game of errors. A good chess player recognizes and takes advantage of the errors. White, by having the first move, probably makes the first error. Therefore, the perfect game of chess would have black recognizing white's error first and capitalizing upon it:)
In my opinion, a perect game of chess is impossible to know for sure, at least right now. In order for a perfect game of chess to be figured out, we need a computer that can analize every single possible outcome in a game of chess, and find the best move order untill to end of the game. But yet again, finding the "perfect game" or best move order is really an opinion that theoretically cannot be proven. Even though almost anyone can agree that e4 or d4 is a much better first move than a4 or h4, it gets alot more complicated then that. This is where many computers flaw, and why humans can beat computers. computers can make mistakes along the line, but usually their miskates are very far down the line (20 move or more) and besides Super GM's, humans cannot see 20+ moves. And they can make positional mistakes as well, based on the way they were programed. basically, since a computer was originally programed by a human, it too can make mistakes (yet, a lot less then humans do).
Another major thing, is that if chess was, at perfect play, a win for white, then chess would be invalidated as a game, at least at the GM level, which in turn would make chess lessly played in the lower levels, and chess in turn would become less popular professonally (yet still played alot for fun at lower levels). Yet again, the same thing would happen if GM's could memorize and perfect a forced draw from move one.
In perfect chess game where each player plays best move given the situation, will result in white winning. White has advantage of making the first move and black will always respond to it, even though response will be best.
At present there is no way to determine what makes "a perfect game". The very best players often win/lose games between them, and many super-GM draws have multiple inferior moves--enough on each side to neutralize each other. So you can't take human games as examples of perfection.
Engines can outplay the best humans but we can't take their word for what is perfect either. All the best present-day engines will easily defeat the top engines from early this century, and all of today's engines will be handily beaten by the top mid-century engines. So you can't get a proper evaluation of perfection from engines either.
Chess is exactly the same because it's a board game which can't carry on forever - there's a limit to the number of moves players can play. And that set of moves is even less if you only picked the best possible moves that would lead to a draw for each side.
In my opinion, it's almost impossible to say that a game is perfect if you include the opening. If you take the first few moves as a given (for example 1. e4 e5), then it is possible to say that a series of moves is perfect.
One example of a so-called perfect game is the Immortal Draw. Carl Hamppe and Philipp Meitner played to a draw in 1872, and in the next century, no improvements have been found for the game. It has stood up to all sorts of computer analysis and GM analysis.
Obviously the argument can be made that there is a "flaw" in the first 3 moves - maybe 1. e4 is not best, or 3. Na4 was an error, but if we ignore those 3 moves, every move starting with move 4 has held up as arguably the best move. For instance, 7. Qe1 is often given as an option for white, but black can once again sac his queen for a perpetual check.
What you're really asking is whether grandmasters have played a game where a computer would have played the same moves (for both sides)? Assumptions about computer and the analysis time for the computer need to be made. I think it's possible to find such a game, but as you let the computer think longer and longer the number of results will dwindle to 0. If you limit it to only one side, you might find more games.
It's similar to playing a drawn tablebase endgame position against an engine that uses the tablebase. It's usually incredibly easy to draw, because the engine doesn't care about making life hard for you. E.g., in a drawn R + pawn v R position, it will often just give away the pawn. That's not worse than other drawing moves, according to the tablebase. It's flawless chess.
You could define a game as perfect when the engine evaluation of the position stays within some boundary (e.g. between +0.3 and -0.3) the entire game. It should be possible to find at least one game that satisfies this condition.
The question has been updated since last time: But is there a game that no computer has ever been able to improve as of this date? To answer this question, someone needs to automate the process of evaluating a game in this manner and then run this script on e.g. all GM games from some freely available database. I guess such games should be few in number and the shorter the game the higher the chance that it meets the requirement!
I honestly don't think you will ever find a "perfect" game just due to the fact that we are all human and we make mistakes. Bobby Fischer was once asked to name his best game and said it was his game with Donald Byrne, but it wasn't perfect.
To me, it is still unclear what one would call perfect, but in terms of wins, Anderssen struck twice with the most unbelievable wins in the history of chess, which I'm going to paste here, the Immortal Game.
I have come across the idea among several threads lately that with perfect play a draw should in theory result. I would think white has the advantage if for no other reason then white would likely be able to choose the first exchange and thus control play.
It is unknown if "perfect play" yields a draw, a win for black, or a win for white. Those that say it "should", or "more than likely" have any of the three results, are guessing. Chess is far from being solved and therefore, no one knows the result of perfect play.
Ohgod. After a few weeks on the site, I think I've realized this is one of those "turns up once a week/fortnight" threads, along with the "favourite move/opening/piece/grandmaster/kitten", "should we resign in a poor position" and "drawing by repetition is cheap" threads. Ah well, another one to toss in the "threads to avoid" list.
I apprecaite the input and tyziefits on this subject. For me it is a psycological benefit in understanding this. If the best black can hope for is a draw then a more defensive game is in order. I find it impossible to think that any advantage confered to black by moving second, thus resulting in a win, could not be equally assumed by white. As far as analasis goes, I would think calculating what side wins more often in Grand Master play would be a good gauge. IN any case it will be sometime before I can aspire to "perfect" play.
I too suspect it is a draw -- if you consider that K+N vs K or K+B vs K (or even K+N+N vs K) are draws and that they present a much larger imbalance (especially when considered as a percentage of remaining material) than is offered by the first move advantage it seems like it would be hard, impossible really, to leverage that first move advantage into a sufficient imbalance to force the win against a perfect opponent.
ahh, thanks TheGrobe. I suppose sacrificing a piece for a drawn position would be a perfectly reasonable outcome. ..but still I wonder, dream perhaps,, does white not have the perfect attack? No draw, no chance, just the unrelenting onslaught of white wood!
Yes I have a source. Remember, I said that it would be a draw "according to the analysis that has been done". Of course, not all analysis has been done (i.e. chess has not been 'solved' as a mathematical exercise - i believe that draughts [checkers] has been solved).
Isnt the point in most grandmaster games getting a small advantage and turning it into a slightly bigger advantage over time. Is it not possible that small advantage of moving first could be turned into a material advantage at some point with perfect play.
Untill 'perfect play' has been worked out there can be no way of knowing who would win or wether it would be draw by repetition after three moves because going first is a disadvantage and so both black and white just move their knights back and forth.
Thank you Amnesiac, sound thinking, people keep assuming that with perfect play black can stop white turning his small advantage into a game winning advantage, and whilst there is not really any proof he/she can, there is no proof he/she can't either. Apart from "common sense" which does not require logic
Using statistics to try to decide what happens with perfect play isn't a good idea. In connect four, going second is often a big advantage for a human player because it's just easier to win that way. However, connect four has been solved and with perfect play the first player will win. So, why does the second player statistically do better?
It's because playing perfectly is extraordinarily difficult, beyond the capability of any human. I don't remember the exact numbers, but in order to play perfectly in connect four the first player only has one correct reply to each of black's moves during the first 15 moves of the game. If we assume there are roughly 5 possible moves each turn, in order to play perfectly the first player would have to memorize 5^15 opening sequences. A computer can do that, but a human can't.
What I'm saying is this: in chess it could be that in reality there are 1 million ways for white to win, and only one way for black to win, but black's way happens to be a forced win and black will always win with perfect play. Even if that were so, playing as white would still be much EASIER for a human being.
Just because white wins statistically more often is absolutely no reason to think that in reality the game is biased towards white, or that it would be a draw. Basically, we can't infer anything, since the game isn't solved.