The 1.7% chance of win for Black... that's hilarious. I acknowledge the possibility but like... the idea that opening moves like e4, Nf3, d4, c4, e3, g3 are losing is just so unlikely. It would basically mean all our best engine evals are systematically off in favor of black by at least a whole point simultaneously and that following chess principles in the opening is losing, and NOT following them is also losing. I would give that outcome something like 1 in 10^1000.

@PeterSmythe Maybe your estimate is right but I don't think your reasons for it amount to anything. Engines are not anywhere near the ballpark of evaluating solved chess; what they think about opening positions has literally no bearing on what an eventual solve may say. And "principles" heuristics, even less.

lol, 6% that white starts the game in zugswang...

@daniel do you have any reason to believe that isn't the case?

@bluerat white starts with an advantage according to the most powerful chess engines, and wins much more often than black in the highest level play between both humans and AIs.

That black has a forced-win, would mean that an as-of-yet undiscovered advantage exists for black, no matter which move white starts with. So basically, every chess engine is wrong in calculating an advantage for white, despite having checked every position to a depth of >50 (>25 moves).

I'd expect to see current best estimate advantage of 0.3 pawn advantage for white either widen as the game progresses, turning into a forced-win for white, or not be enough and be a forced-draw for black.

@daniel chess engines currently use opening books because the game isn't solved. They've already refuted many popular openings that were used at the highest level such as the French and the kings Indian

@TheeDuelist yes, when chess engines actually play the game they use opening books. Otherwise they'd select the same moves every time (if processing time doesn't vary).

I meant running a chess engine evaluation on the starting position of chess. Engines can then scan the game tree very deeply and judge evaluations for resulting positions, and for all engines there are moves that give white a slight advantage at the start. This is how engines changed how humans play chess openings, not because engines use opening books, but because they can evaluate the game tree better than us.

A forced win for black would mean the starting position's correct evaluation is -Infinity (or mate in X moves). Even if today's engines can't solve chess, a winning forced sequence for black will at some point result in a negative evaluation, no matter where you are in the game tree. So to believe "black has forced win in chess" means that if we just can run the engines for longer and deeper, the evaluation will at some point turn negative.

White having an advantage is weak evidence, but not strong evidence, for white being able to win in perfect play. For tic tac toe, the first player has a big advantage when the play is imperfect*, but the second player is still able to draw.

*Start in a corner, not the middle.

@Mqrius It wouldn't be terribly surprising if chess was a draw with perfect play. In fact it's probably the most likely outcome. The surprise would be if there turns out to be a forced win for black.

@Sophia Ah right, though that doesn't seem impossible at all, it does seem unlikely.