I was wondering if lichess could make an official bot tournament.
Could, but won't. One of the 100+ stockfish clones would win, which isn't all that interesting. And sadly that's still true even for variants, we may not know the winners name but we know what engine they will be running. It would be somewhat interesting if only one bot could enter for each engine, but would take a lot of work to check that people were not just renaming stockfish and trying to claim that it was something else.
@ProgrammerAngrim it kind of depends. Someone with a powerful computer can beat the other sfs with an intel pentium 2.
I've written a chess engine and have been looking into the lichess API to make a bot, but I agree with ProgrammerAngrim. Anybody could simply fork stockfish and make a bot from it and call it their own engine.
@ProgrammerAngrim @DavidChopin
True, that's the downside. But @XXIstHuman is correct, and there are still Bots out there neither running SF or Leela. Unofficial tournaments are unreliable.
Besides, I would like to see how well NNUE (github.com/official-stockfish/stockfish#a-note-on-classical-and-nnue-evaluation) plays against Leela.
True, that's the downside. But @XXIstHuman is correct, and there are still Bots out there neither running SF or Leela. Unofficial tournaments are unreliable.
Besides, I would like to see how well NNUE (github.com/official-stockfish/stockfish#a-note-on-classical-and-nnue-evaluation) plays against Leela.
It is kind of interesting to see the rise of the AI engines.
May there needs to have dedicated chess engine teams. Put all the clones in one team.
Then do Team Battles, Team Tournaments & Swiss Tournaments.
If you discover a bot that does not belong in that team, you kick it of the team.
I assume most engines are still have problems with chess fortresses.
So having a team that only holds engines that can solve all these known fortresses might be very interesting.
Then again how would a bot user prove the engine solved the fortress problem?
If an engine evaluates it a fortress, then it should say that it is a fortress, like when it says mate in 3 or draw.
An evaluation other than zero proves it does not see the fortress.
An evaluation past 50 moves deep proves it does not use the 50 move rule.
Try this fortress out on your bot:
lichess.org/analysis/6Q1/3kp3/5r2/4K3/8/8/8/8_w_-_-_0_1
en.wikipedia.org/wiki/Fortress_(chess)
It shows every move the Queen can go to and shows which ones are draws or losing moves, but still continues to evaluate. It's evaluating what at 73/99 moves deep?
I would assume it would stop evaluating once the exact main lines for each move is complete.
Then do Team Battles, Team Tournaments & Swiss Tournaments.
If you discover a bot that does not belong in that team, you kick it of the team.
I assume most engines are still have problems with chess fortresses.
So having a team that only holds engines that can solve all these known fortresses might be very interesting.
Then again how would a bot user prove the engine solved the fortress problem?
If an engine evaluates it a fortress, then it should say that it is a fortress, like when it says mate in 3 or draw.
An evaluation other than zero proves it does not see the fortress.
An evaluation past 50 moves deep proves it does not use the 50 move rule.
Try this fortress out on your bot:
lichess.org/analysis/6Q1/3kp3/5r2/4K3/8/8/8/8_w_-_-_0_1
en.wikipedia.org/wiki/Fortress_(chess)
It shows every move the Queen can go to and shows which ones are draws or losing moves, but still continues to evaluate. It's evaluating what at 73/99 moves deep?
I would assume it would stop evaluating once the exact main lines for each move is complete.
@CalKK10 nnue is stockfish. It is stockfish 12 which is depending on nnue evaluation.
@XXIstHuman yes, that's what I mean.
yup
This topic has been archived and can no longer be replied to.