Romi getting owned by a human on ICC

[Tony Thomas]

Finally the match has started.. Oh well your engine is strong enough that it will win against romi most of the time, in the endgame. I remember seeing your engine with a rating of more than 2800. Even with selective playing, Romi hasnt crossed the 2600 barrier on ICC. Your engine is getting a search depth of around 18-20 where as poor Romi is stuck at 14-15.

[Tony Thomas]

Romi loses the first game. Almost a pablo type position, and Romi does something weird, in five moves everything is over.

[Tony Thomas]

Second game is a draw, Lemming had slight advantage all along, and it ended up in a KPP v KN endgame.

[Tony Thomas]

Another loss for Romi, it simply doesnt understand endgames with pawns.

KRN VS KRPP..should have somehow exchanged the knight for the two pawns.

[BubbaTough]

Another loss for Romi, it simply doesnt understand endgames with pawns.

KRN VS KRPP..should have somehow exchanged the knight for the two pawns.

Do not be too intimidated by the depth, the search is quite selective, and Romi definitely has chances to catch it in a blunder despite lower search depth. I saw that last game, very interesting. Lemming sacrificed early and I thought had a fun game, but was slightly worse most of the time. By the time it got to the endgame you mention though, I thought LL had squirmed into decent winning chances because the king was too far out of the action. It will be interesting to see how Romi's learning effects things.

-Sam

[Tony Thomas]

Another loss again..I didnt see what happend this time around. None of the losses are due to bad opening. Learning Lemming also learns, so I doubt that a better learning ( I am not sure who has better learning) will show an effect unless we play thousands of games.

[Tony Thomas]

I am pretty sure the current game is lost as well. Darn Lemming..you surely have improved. I shall secretly log off and come back with Rybka 4.

[BubbaTough]

Another loss again..I didnt see what happend this time around. None of the losses are due to bad opening. Learning Lemming also learns, so I doubt that a better learning ( I am not sure who has better learning) wont have an effect unless we play thousands of games.

LearningLemming tries to learn in a more general way, instead of rewarding wins and losses, so I would expect Romi's learning to be more effective in a head to head match. How much, I don't know. Or maybe not. Anyway it will be interesting. Openings have been kind of funky to start with, will also be interesting to see if the two computers muddle there way toward more established theory during learning or are playing something bizarre by morning.

-Sam

[Tony Thomas]

Another loss again..I didnt see what happend this time around. None of the losses are due to bad opening. Learning Lemming also learns, so I doubt that a better learning ( I am not sure who has better learning) wont have an effect unless we play thousands of games.

LearningLemming tries to learn in a more general way, instead of rewarding wins and losses, so I would expect Romi's learning to be more effective in a head to head match. How much, I don't know. Or maybe not. Anyway it will be interesting. Openings have been kind of funky to start with, will also be interesting to see if the two computers muddle there way toward more established theory during learning or are playing something bizarre by morning.

-Sam

Your program is quite stronger than the old public version of Romi. That along with the fact that Romi is playing ponder off should mean that it would take Romi at least a million games to get any advantage against Lemming. Also, the only time Romi wins due to learning is if Romi plays from the book for around 30-40 moves. So far, Lemming has been avoiding all opening traps.

[Tony Thomas]

Romi is crazy, it thinks that its only behind by a pawn were as Lemming thinks that its ahead by a queen. I have to agree with lemming on this position.