Your move!
Link to the game in lichess:
d4 d5
c4 c6
Nc3 Nf6
Nf3 e6
Bg5 dxc4
e4 b5
e5 h6
Bh4 g5
You (Manifold) have been challenged to a game of chess! You'll be white playing me (Alex) a ~2000 lichess player who will not use any outside assistance.
After the market close I will generate a random number and the move corresponding to that percentage will be picked. For example, order the moves from highest percentage to lowest. Then to find a moves range you sum the percentages that are greater which give you the lower end of the range than add the percentage of the move itself to get the higher end of the range.
Here's the previous move and a market for the result:
Oct 6, 6:54am: Random number between 1-998 is.. 925 so Nxg5 is the move.
1,000
3.00@ms We have discussed this in the will white win market, this is potentially better, but potentially much worse, because people who buy 'no' on previous move markets are incentivized to manipulate future move markets into a loss (which can be done with a last minute bet).
This is a KBS, not a proper market. The way to use the invisible hand to play chess is to create for every possible move a market of “will white win if this move is made”, to pick the move predicted to lead to the highest chance of winning, resolve the market when the game ends, resolve other markets to N/A.
@JonathanNankivell I've put in my two cents in this move, but I still vastly prefer Nxg5. Note that if you filter Lichess games by time control (classical+correspondence) and rating (>=2000), exf6 performs worse, and this is effectively correspondence chess.
Sadly there don't seem to be enough data to draw meaningful conclusions looking at just correspondence games, except that people still play Nxg5 an OoM more often.
@Mateon1 It's not the engines favourite, sure, but in both the database of masters games and the database of all games played on Lichess it's exf6 that has the best track record
@Mateon1 Actually, I have re-analyzed this position, it's not losing - stockfish variants give less than one pawn advantage for black - it's just a poor move. It's probably okay if this gets picked since 2000 range still definitely blunders vs engines and can't convert such a small advantage.
@JonathanNankivell While capturing the knight in exchange for the bishop immediately seems like the intuitive move, since the material difference is only one pawn in favor of black after the move and we have a very advanced pawn, it's actually a losing position. The pawn is not protected in any way so the queen can just take it and we're down two pawns with no advantage. That is not bad enough to have no winning chances versus a lichess 2000, but it's a very uncomfortable position.