Sunday, November 23, 2008

Sorry, Pete


Commenting on my recent computer vs computer Jerome Gambit (1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+) matches (see "I can't seem to get the hang of these things...") Pete Banks ("blackburne"), a long-time member of the Jerome Gambit Gemeinde, posted a Comment

Hi Rick,

Have you tried matching two equally-strong computer programs? It might be interesting to see what percentage White got. I'd guess about 40%.

Pete

While I liked Pete's enthusiasm for the Jerome Gambit, my Comment reply was not as optimistic as his. After all, while I have supported the use of the Jerome when giving "odds" to a weaker player; I'm not at all confident in its prospects when the players are evenly matched.

Hi Pete,

I'll ask Rybka to play a similar match against itself. I suspect it will run something like 0-20.

But, then again, what do I know?? You see, I can't seem to get the hang of these things...

Rick

The first thing that I learned is that using the Aquarium interface I wasn't able to have Rybka 3.0 play a match against itself. However, I was able to have two similarly powerful Rybka engines face off: Rybka 3 Dynamic w32 vs Rybka 3 Human w32.

I set up a 20-game Jerome Gambit match, starting from the position after 4.Bxf7+. Each computer had 5 minutes per game. I gave Dynamic the White pieces first.

The match finished 0 - 20.

I then switched engines, giving Human the White pieces.

That match finished 0 - 20.

That's zero wins for the Jerome Gambit, against forty losses!

I've really got to stop running these experiments...

(I won't post the games, but they will be available in the 4th PGN file of games presented or referred to on this blog, when it becomes available.)

2 comments:

  1. Rick,
    OK, so you were right. I suppose everybody has to be once in a while!
    The result is still very interesting though. I think you'll agree that even if you haven't finished your analysis, the percentage between human players is very different.
    For instance, in the three Jerome Gambit tournaments I have organised on Chessworld so far, I think the percentage for White was around 40%? Correct me if I'm wrong. My own percentage with White was well above 50%, and most players were roughly the same strength.
    In rated OTB games I've won 4 and drawn 1 out of 5 over the last few years.
    I think it shows that the 'shock and awe' factor is important between humans. Computers are too stupid to be intimidated!

    Pete

    ReplyDelete
  2. Pete,

    Oh, those "experiments"... I knew they were going to get me into trouble.

    As a practical weapon against the proper human opponent, I think the Jerome Gambit can be successful.

    In fact, I know that it can be successful - as you mention, some players (such as yourself) have had impressive results with it, over-the-board and in the Chessworld thematic tournaments.

    Indeed, as you say, "computers are too stupid to be intimidated" - although I still hope that one day I can figure out how to get them to emulate that "shock and awe" phenomenon.

    But for now - no more experiments.

    Rick

    ReplyDelete