Everyone on mersenneforum vs. Stockfish

[MooMoo2]

Now that the latest vote chess game has ended, I'm wondering if there's any interest in having a humans vs. machine game. It'll be structured similar to the vote chess games, except that everyone here will be playing as one team against the chess engine Stockfish. I'll make the moves for Stockfish.

There can be two games - in one game, everyone plays as white and Stockfish plays as black, but no engine assistance is allowed. In the other game, Stockfish plays as white, and engine assistance is allowed, as long as you're using an engine other than Stockfish to analyze the position.

I'm wondering what the results will be. Anyone want to create a new sub-forum for discussion and join in?

[Brian-E]

Interesting idea!

A couple of practical points initially spring to my mind:

(1) Do we need a subforum? Considering that Stockfish won't be reading the forum, can't we organise it all and actually play the games from this single thread?

(2) We will need a clear protocol for discussion and how and when to make our moves, and in particular what those moves should be when there is no general agreement as there usually won't be. It was difficult enough to do this with teams of just three or four (plenty of general bickering, sometimes disinclination to get on with the discussion, sometimes misunderstandings about whether people were finalised on the move to be played). I think with "everyone" playing, those problems will be greatly magnified and we will need a rather brutal, clearcut method of deciding on the move and playing it.

What kind of ELO's are acceptable?
Assuming normally distributed responses, would there be access to the tail contribution containing the most devious moves?
I once played a fellow by the name of "Arkady" in Washington Sq. years ago who said he was from Belarus and said he was an international grand master (very broken English and I could never verify this), however, he was soon teaching me how to play. Russian Paul in NYC is also a very nice fellow. I am always interested in improving my game so chalk me up please.

[Batalov]

ELO of 3100 will probably do.

Originally Posted by Batalov

ELO of 3100 will probably do.

Cute. Pawn sacrifice..

[MooMoo2]

Originally Posted by Brian-E

Interesting idea!

A couple of practical points initially spring to my mind:

(1) Do we need a subforum? Considering that Stockfish won't be reading the forum, can't we organise it all and actually play the games from this single thread?

(2) We will need a clear protocol for discussion and how and when to make our moves, and in particular what those moves should be when there is no general agreement as there usually won't be. It was difficult enough to do this with teams of just three or four (plenty of general bickering, sometimes disinclination to get on with the discussion, sometimes misunderstandings about whether people were finalised on the move to be played). I think with "everyone" playing, those problems will be greatly magnified and we will need a rather brutal, clearcut method of deciding on the move and playing it.

(1) We can certainly play the games from this thread. However, I was thinking of getting a (public) subforum because discussing the moves, voting on them, and playing the game in one thread might get confusing.

(2) One method is creating a poll for each move, which would close in a few days. Once the poll is closed, the most popular move is played, Stockfish plays its response, and the next poll is opened. If there's a tie, another poll would be created that only has the moves with the same number of votes, and if the vote is still drawn, a random move (out of the tied ones) would be made.

Originally Posted by jwaltos

What kind of ELO's are acceptable?
Assuming normally distributed responses, would there be access to the tail contribution containing the most devious moves?
I once played a fellow by the name of "Arkady" in Washington Sq. years ago who said he was from Belarus and said he was an international grand master (very broken English and I could never verify this), however, he was soon teaching me how to play. Russian Paul in NYC is also a very nice fellow. I am always interested in improving my game so chalk me up please.

All ELOs would be acceptable, but one of the things we could try would be putting in an (honor-based) system for chess competency and move confidence. Votes from good players with high confidence could be given greater points than votes from average players who are not too sure of their move choice.

[Brian-E]

Originally Posted by MooMoo2

(1) We can certainly play the games from this thread. However, I was thinking of getting a (public) subforum because discussing the moves, voting on them, and playing the game in one thread might get confusing.

Yes, I see. Maybe two threads would be adequate, though, one for the moves of the game and the other for discussion.

(2) One method is creating a poll for each move, which would close in a few days. Once the poll is closed, the most popular move is played, Stockfish plays its response, and the next poll is opened. If there's a tie, another poll would be created that only has the moves with the same number of votes, and if the vote is still drawn, a random move (out of the tied ones) would be made.

That sounds good for the move selection itself. I also think we would need to have a clear period of discussion before the poll even opens. The reason I think that is because if we allow polling right from the start when each move has been received, then we may get too many people voting for what is actually a blunder (maybe overlooking a reply which is difficult to see). The fact that it is a blunder needs to be highlighted by the players in the discussion before the voting starts to make sure it doesn't get played despite it having become clear that it's a mistake.

All ELOs would be acceptable, but one of the things we could try would be putting in an (honor-based) system for chess competency and move confidence. Votes from good players with high confidence could be given greater points than votes from average players who are not too sure of their move choice.

Hmmm. How would we quantify such confidence and identify the "good players" to start with? Most people won't have an active rating from genuine chess competition. Unless we can do this objectively somehow, there could be some bad feeling generated.

[MooMoo2]

Originally Posted by Brian-E

Yes, I see. Maybe two threads would be adequate, though, one for the moves of the game and the other for discussion.

Two threads would be enough for the game and the discussion, but the polls would have their own threads.

That sounds good for the move selection itself. I also think we would need to have a clear period of discussion before the poll even opens. The reason I think that is because if we allow polling right from the start when each move has been received, then we may get too many people voting for what is actually a blunder (maybe overlooking a reply which is difficult to see). The fact that it is a blunder needs to be highlighted by the players in the discussion before the voting starts to make sure it doesn't get played despite it having become clear that it's a mistake.

Agreed; the poll will be created 2-3 days (or maybe more, depending on the complexity of the position) after the discussion starts.

Hmmm. How would we quantify such confidence and identify the "good players" to start with? Most people won't have an active rating from genuine chess competition. Unless we can do this objectively somehow, there could be some bad feeling generated.

Good point. We'll just stick to the "one man, one vote" rule for now. In the discussion thread, people can say "I have a rating of 2500, I know move X is the right move to make, and here's why", but it's up to each individual to make the decision and vote.

[LaurV]

Very interesting idea! I proposed something like that in the past, and I even considered that we (collectively) can win against ANY chess engine, given a week per move to think and discuss. I was criticized for that opinion (oh-pinion!) the discussion is here around, somewhere.

I am in for both games. For mid-game and end-game phase, I am a good player and an excellent "combinationist", very strong on game finals. My weak part is the openings (never learned more than a narrow path which I usually play, and almost all games I lose, I lose because I meet an opening I don't know and I make opening mistakes). I usually win against most of the engines if I use some "assisted opening" and if I have enough time to think. I regularly beat win7's default "chess titans", at max level (it didn't won any game for ages!) but general opinion is that "chess titans" is a weak engine (and I almost learned its 'always the same' opening moves).

I hope that this "combined front" will be able to discern weaknesses within Stockfish
and possibly help improve the engine.
Convekta advertises a complete 7 piece endgame tablebase and I am not aware of anything else
commercially available beyond this. "Aussie rules" chess - as polling progresses for various moves
(as this seems to be a team effort) everyone's relative strengths/weaknesses should become apparent
during the transitions from opening/middle/end and a tenuous local ELO index may be established provided
the particants have submitted a sufficient number of moves to profile.

[Brian-E]

Originally Posted by jwaltos

I hope that this "combined front" will be able to discern weaknesses within Stockfish
and possibly help improve the engine.

You share LaurV's opinion then. I would just like to voice my own opinion that we don't have a hope of avoiding defeat unless we have someone here who is at least International Master strength (roughly 2400-2550), or unless the program plays on an articially weak setting. As long as everyone involved is prepared for that reality, we shouldn't get too discouraged when our position inevitably deteriorates.

Convekta advertises a complete 7 piece endgame tablebase and I am not aware of anything else
commercially available beyond this.

I would suggest that the use of tablebases by the human team (should the game reduce to a 7- or fewer piece endgame) should be outlawed in the same way that we are not allowed any other sort of chess-playing software use. Tablebases guarantee perfect play with no skill required, and that's not quite the idea, I would suggest.

"Aussie rules" chess - as polling progresses for various moves
(as this seems to be a team effort) everyone's relative strengths/weaknesses should become apparent
during the transitions from opening/middle/end and a tenuous local ELO index may be established provided
the particants have submitted a sufficient number of moves to profile.

I certainly think those with the loudest mouth will quickly become apparent. (Guilty here!).
I can't really believe that the contribution towards the early moves of a single game can possibly provide any objective evaluation of someone's playing strength, though.
I am in agreement with MooMoo2 that it's best to stick to "one player = one vote", if only to avoid ill feeling amongst the team if people feel their contribution is being unfairly downgraded.