I wrote this on another forum:

Assume a computer that plays from move one and is following the instructions of the "solved" game.

If that perfect playing computer is playing another one it will either always be a win for white, always be a win for black, or always be a draw.

Experts think it is almost certain that it will always be a draw.

If it is indeed a draw, than white does not have a theoretical "advantage". But since it wins slightly more often than black in the real world, there must be a reason.




Since a perfect playing chess computer will never make a move that gives the opponent a forced checkmate (if it is indeed true that two perfect computers always draw), it is almost certainly true that the reason that black loses more often than white in the real world is because it misses all of the moves that the perfect computer could make. But why would it miss all of these moves more often than white? The reasonable explanation is that the white computer has more choices than black under the constraint that it will not make a move that gives the other guy a forced checkmate. If it has more acceptable choices than black, a fallible human is more likely to hit upon one of those acceptable choices when playing white.

(I realize that an alternative explanation for white winning slightly more often is that black is intimidated by its color and thus is more likely to play for a draw.)

Current computers, which don't even have the game solved yet, make moves that make zero sense to any human player. These computer moves, one of the key ways you catch a cheat, prepare for something many moves ahead. The importance of these moves suggests the decision space for chess is so much wider than humans really grasp, and that perfect play is so much narrower than we really imagine.

Humans use heuristics to shortcut the need to calculate every possible line. That means we generally overlook quiet preparatory moves that turn out to be essential in the best chess. If chess were shown to be a theoretical draw, while we know human players win more as white than as black, then to me that suggests a few possibilities:

- The overall game tree has more wins for white than for black. Our human deviations from perfectly play may be no better than random, after all.

- Our heuristics are biased toward white. It may be our deviations are far enough from random that they bias in favor of the first player.

- Our heuristics are biased toward decisive outcomes, and that bias hurts black more than white.

When you jump from describing computers playing a "solved" game to chess you seem to be ignoring the fact that chess is not yet solved nor do we know whether given physical constraints chess can be solved.

I don't think it's super important whether or not chess has actually been solved yet for this question. It's just a setup to assume chess is a theoretical draw to ponder why Black is losing more in the real world. I think the conclusion is correct yes, Black loses more because the development disadvantage gives Black less non-losing moves that a fallible human might play.

I just don't get the point of the question though. This is obviously the case whether or not chess is a theoretical draw.

number of atoms in the observable universe : between 10^78 to 10^82

number of possible chess games : between 10^111 to 10^123

i think the white edge means exactly what it is

computers still don't know how much edge white has , but they know it has an edge , lets say its +0.1% , that computer playing by those standards would never lose and would win 1 in 10 games over millions of games

if computers can't figure out the exact number of possible combinations for all possible chess games, its still too early to know what edge white has, but it has some , and will always translate into never losing and winning at x frequency , combinations are just too many to understand

In the simplest terms, white has an advantage because it has the first right of movment... and that's it.

Black will always be reacting to input, and White is always implementing input. This is of course reliant on optimal play by both white and black.

In the real world, as it is with poker, suboptimal play changes/influences outcomes and therefore this first mover advantage can be nullified and/or exploited.

This is the failure of gto over a smaller sample size.