An idea for different type of competition between engines

[Uri Blass]

The idea is that instead of games from the initial position the sides start from a position when one side is obviously winning and play 2 games when the winner is the side who mate in less moves.


For example this may be the starting position of all the games
[d]rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNB1KBNR w KQkq - 0 1

black of course is going to win every game so when engine A play against engine B they play 2 games.
If A mates in less moves then A wins,if they need the same number of moves then it is a draw and otherwise B wins.

I think that the problem is that there is no interface that support automatic competition between engines when you can see the results(looking at a big pgn to find who won in pair of games between every pair of engines is no fun) and this is the reason that we do not see competitions like this but
I think that when games between top engines are usually draw competition who need less moves to win in won positions may be interesting and it may be interesting to see if there are engines that are significantly stronger in this type of competition relative to normal chess.

[Daniel Shawul]

MCTS engines have inherent difficulty making progress when the game is "technically won" i.e. score >= 300?, so this competition
disadvantages them. Lc0 had to actually train a net with "move left head" to try and win in shortest moves possible but is cetainly
not comparable to an AB engine.

[Ras]

It would also basically mistreat two equal strategies as sign of different playing strength: either use the material advantage for a direct king attack, or trade down as much as possible and go for the endgame. The rules of chess don't care how victory is scored as long as the result is checkmate.

[Uri Blass]

The rules do not care how victory is scored but for analysis I am interested in the fastest way so I would like to know which engine is faster for the choice of engine to analyze positions when one side is winning.

[Chessqueen]

The idea is that instead of games from the initial position the sides start from a position when one side is obviously winning and play 2 games when the winner is the side who mate in less moves.


For example this may be the starting position of all the games
[d]rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNB1KBNR w KQkq - 0 1

black of course is going to win every game so when engine A play against engine B they play 2 games.
If A mates in less moves then A wins,if they need the same number of moves then it is a draw and otherwise B wins.

I think that the problem is that there is no interface that support automatic competition between engines when you can see the results(looking at a big pgn to find who won in pair of games between every pair of engines is no fun) and this is the reason that we do not see competitions like this but
I think that when games between top engines are usually draw competition who need less moves to win in won positions may be interesting and it may be interesting to see if there are engines that are significantly stronger in this type of competition relative to normal chess.

Uri, that is an excellent idea, but it also depend what engine plays as Black, since if the difference is more than 400 rating points for instance Komodo Dragon as White playing versus Stash as Black that pawn handicap would NOT make a difference at all, but it would rather keep the game more interesting than if both start with 16 pieces. As a matter of fact all engines rated over > 400 points than the opponent engine should give a pawn odds no matter which side they are playing

Note: Also a great idea would be to remove the Opening and endgame EGTB to the stronger Engine