Pawn and Move

[lkaufman]

Years ago, I ran many widely publicized matches of Rybka giving various handicaps, particularly Pawn and Move (Rybka played Black with no f7 pawn), against grandmasters and also against other strong engines of the day. Just for fun, I decided to try the same with Komodo 5.
I picked Fritz 12 as the opponent, partly because I felt the elo gap was about right for the handicap, and also because, since a much older version defeated World Champion Kramnik in a match, it is likely that Fritz 12 is a stronger player than any human on earth. Time limit was 3' + 2", the same used for the World (human) Blitz Championship. I used a testsuite I made of ten typical openings used at the handicap.
Result was a 5.5 to 4.5 victory for Komodo. It's really hard to believe that it is possible to win a match from a program as strong as Fritz 12 giving it such a large handicap (just check the eval Fritz 12 gives for the start position!), even with good luck. I'm not saying that some other superstrong engine might not do the same, in fact I invite others to run similar matches with other engines on either end. Just look for elo gaps in the 250 ballpark.
The implication is that Komodo could successfully give pawn and move in blitz to any human. While this is not necessarily so from the above data (because humans know about the handicap and can adapt their play), I believe it is true, primarily because at blitz I think Fritz 12 will destroy any human.
Maybe soon we'll try some Komodo vs. human master/grandmaster handicap matches.

[gleperlier]

Years ago, I ran many widely publicized matches of Rybka giving various handicaps, particularly Pawn and Move (Rybka played Black with no f7 pawn), against grandmasters and also against other strong engines of the day. Just for fun, I decided to try the same with Komodo 5.
I picked Fritz 12 as the opponent, partly because I felt the elo gap was about right for the handicap, and also because, since a much older version defeated World Champion Kramnik in a match, it is likely that Fritz 12 is a stronger player than any human on earth. Time limit was 3' + 2", the same used for the World (human) Blitz Championship. I used a testsuite I made of ten typical openings used at the handicap.
Result was a 5.5 to 4.5 victory for Komodo. It's really hard to believe that it is possible to win a match from a program as strong as Fritz 12 giving it such a large handicap (just check the eval Fritz 12 gives for the start position!), even with good luck. I'm not saying that some other superstrong engine might not do the same, in fact I invite others to run similar matches with other engines on either end. Just look for elo gaps in the 250 ballpark.
The implication is that Komodo could successfully give pawn and move in blitz to any human. While this is not necessarily so from the above data (because humans know about the handicap and can adapt their play), I believe it is true, primarily because at blitz I think Fritz 12 will destroy any human.
Maybe soon we'll try some Komodo vs. human master/grandmaster handicap matches.

Nice idea to give Human versus Machine match a new interest.

Gab

[lkaufman]

Overnight I reran the same match at the much slower time control of 15' + 10". I expected that at the longer time control Fritz would be better able to exploit the handicap. But the result was a more decisive 6 to 4 victory for Komodo 5 over Fritz 12! So the combined results are 11.5 to 8.5 for Komodo.