Will the things listed show up in any of the 14 games, excluding tiebreaks, in the World Chess Championship 2024, which starts sometime after the candidates tournament ends in April 24, 2024
The match will be held between Ding Liren and whoever wins the Candidates Tournament 2024
Rules
Pawns "die" when promoting
By "points" I mean the traditional (~inaccurate) point heuristic: 1 point per pawn, 3 points per bishop or knight, 5 for a rook, 9 for a queen. Because of complexity, I do not currently use anything related to engine's "points of advantage estimation" in this market.
Only the 1st 14 (or fewer) games, to make it easier to estimate. I wish we were guaranteed 14 total games; I'd rather run the risk of it ending early than prepare to do some difficult interpolation, so we'll just use as many of the nor Al 14 games as they actually play.
For the same reason we do not count tiebreak games. That would distort estimation too much until the exact time.
"Piece" means any chess piece. I don't use the meaning "a knight or a bishop" here, for simplicity
bishop "undoing" moves requires the initial move to not capture a piece; if we allow that, then it's not really an undo. undo means move from A to B, other player does something, then that same bishop moves back to A.
Knights on the rim means a knight on the a or h file.
I know it's very annoying, but if they shake hands on a draw 1 move before an official stalemate happens, that means the stalemate didn't happen. I wish we could resolve this "ends early" ambiguity but I don't see a way now. Open to suggestions.
"self-capture chess" aka "capture anything". https://chess.stackexchange.com/questions/28487/is-there-a-name-for-a-chess-variant-where-you-are-allowed-to-capture-your-own-pi Note that if the game mode switches, the defender can also capture anything to get out of check. So be careful when evaluating this.
A full rank has been changed to "A full rank (ranks 3-6 only)" because I couldn't think of a way to clearly distinguish meaningfully full ranks of the first and last two ranks. Even if I require a mix of your own or the other's pieces, it's not clear where it becomes interesting.
Contest
Whoever gets the most "first direction of bets" right, where they also made their first bet before the first game started, will win the official Manifold Markets Ernie Chess 2024 MegaMarket brilliancy prize. Regardless of bet size this is just if your initial direction was towards the final price in an option. Later selling is irrelevant. So yet your bets in before the match starts! You have to submit your own scores for this since the UI doesn't make it that easy to run myself
Update 2024-12-12 (PST): The final score will be calculated as 10% of the maximum material advantage seen in any final position across all games, not the sum of material advantages. (AI summary of creator comment)
Update 2024-13-12 (PST): For social distancing, the orthogonal distance refers to the manhattan distance between pieces. (AI summary of creator comment)
Thanks for playing everybody.
if you claim "chess speaks for itself" was said, please provide links otherwise I'll resolve NO at EOY
I am not sure all the evaluators are correct - they were mostly written by o1. The code is updated here: https://github.com/ernop/pawncube
and I also ran them agains a set of ~2000 earlier ding games to test them. You can see that many of them which didn't come true in the championships this year did hit there. Here's a google docs link to the output log of that run with ascii board diagrams. https://docs.google.com/document/d/1UyMNwZoZXgPNvk4JydluucO_ufM2c91R_eCtFOpHJ_0/edit?tab=t.0
I can reresolve things if you provide clear links. Ideally it'd be a link to a lichess board of one of the games, at the specific position you want me to pay attention to! Thanks
@Ernie Thanks for a fun market. I don't think you need to wait that long for that single option, a week should be fine.
source:
https://lichess.org/broadcast/fide-world-championship-2024/game-1/G2R2lPcA#boards
games:
1) black b pawn never moves
2) white h pawn never moves
3) black b pawn never moves
4) white f pawn never moves
5) white a pawn never moves
6) white b pawn never moves
7) white light bishop remains
8) white h pawn never moves
9) move 31 black h pawn moves 2 squares and doesn't move again until fewer than 8 pieces remain
10) white h pawn never moves
11) both f pawns never move
12) both b pawns never move
13) black g pawn never moves
14) move 25 white h pawn takes to stay on black
@Ernie there are 14 points so far (-3 in G1, 1 in G5, 3 in G7, -2 in G8, 2 in G11, 3 in G12)
@R2D2sm5g wait, reading this again and looking at code, do you think this means to sum up the advantage? I wrote the code to do max(material_advantage_in_all_games_final_position) * 10%
RESOLUTIONS Unpaused, looking into questions needing updates now, thanks!
@Ernie as far as I know, game 3 was lost on time. Hard to know if Ding technically resigned in time, but his time ran out before clocks were stopped in any case. Take take take reports it as a time loss.
@Ernie Game 2 started 1. e4 e5. The White and Black armies are in two clusters and the third cluster is the two pawns in the middle of the board.
@Ernie I was wondering what you were going for with this one, since clearly this wasn't what was intended!
@Calibrate yeah, faire fair though 😎 I've been trying to think of what to do for "Mexican standoff 2"