Wikipedia:Reference desk/Archives/Computing/2024 May 12
| Computing desk | ||
|---|---|---|
| < May 11 | << Apr | May | Jun >> | May 13 > |
| Welcome to the Wikipedia Computing Reference Desk Archives |
|---|
| The page you are currently viewing is a transcluded archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages. |
May 12[edit]
Longest computer-chess game[edit]
What's the longest chess game (i.e. the one with the most moves) on record between computers, not counting moves after one side could've claimed a threefold-repetition draw? Would the answer change if known endgame solutions retroactively replaced the fifty-move rule where available? NeonMerlin 05:10, 12 May 2024 (UTC)
- I'm not sure there's any sensible answer to this. A poorly written program could play essentially forever. It wouldn't be hard to write a program that just makes random moves without attempting to checkmate the other side's king. Playing it against itself, it could probably play for thousands of moves before one side accidentally stumbles on a checkmate. CodeTalker (talk) 18:58, 12 May 2024 (UTC)
- Okay, so what if we consider only those games where both programs have beaten the World Chess Champion in a publicly recorded match without a handicap in their favor, or have beaten in such a match a program that had done so in such a match, or so on transitively? Or only those whose FIDE-equivalent Elo ratings are at least 2800 based on human-computer matches that someone's bothered to provide enough hardware for and publish the moves from? NeonMerlin 01:46, 13 May 2024 (UTC)
- I will see the wrong direction in there as the (poor) program will easily crumble in on itself if not designed to severely enlarge repetition paths. Prospective targeting needs to be entirely missing to not arrive early to a simple tic-tac-toe configuration, if pawns are designed to look on to always forward they are often only postponing a clear view but, related to this there's also a notable level of subjectivity. As a result, and ratings being about expectations, what could be the use trying to hardcode them? --Askedonty (talk) 01:04, 17 May 2024 (UTC)
- Okay, so what if we consider only those games where both programs have beaten the World Chess Champion in a publicly recorded match without a handicap in their favor, or have beaten in such a match a program that had done so in such a match, or so on transitively? Or only those whose FIDE-equivalent Elo ratings are at least 2800 based on human-computer matches that someone's bothered to provide enough hardware for and publish the moves from? NeonMerlin 01:46, 13 May 2024 (UTC)