Random Posts

Thursday, December 30, 2010

Houdini Revisited

Apparently, from what I’ve been reading, GM’s are now using engines to discover theoretical novelties in the openings and I wonder how this will affect what is being played, or even the popularity of certain openings themselves.

Apparently there’s controversy over whether or not Houdini is a clone of Rybka or an original engine, but unless you are a computer engine guru, who cares? I came across the position below in which it was claimed is beyond the ability of most popular engines to solve but Houdini 1.03a found 1.Rb1 in 46 seconds on the poster’s slow laptop. A discussion of the following position (White to move) is discussed HERE.

Marjan Semyrl, a Correspondence GM, wrote a Squidoo page discussing the four elements of good engine analysis HERE. If you really want to know how top level engine analysis should be done, check out his page. As he points out, good analysis requires a powerful computer with good chess programs (note the plural form!), a lot of time, well versed in how to perform computer analysis, and finally, you must be a strong and experienced player.

All that disqualifies most of us! As I’ve pointed out in the past, that explains why there is more to playing top level CC than just buying Fritz or Rybka and it explains why players who just rely on engine generated moves will not do well against a top level CC player. Even playing at lower levels in engine tournaments on Lechenicher SchachServer I never had very good results because I was using a 10 year old desktop, Fritz 6 and and never let it run more than maybe 2-3 minutes before selecting a move; longer only if I disagreed with the engine's evaluation. I started at LSS several years ago at my official CC rating of 2060 and in my first tournament there didn't realize engine use was allowed. Even after I found out I could use engines, it didn't help much because I ended up with only a 50% score and still managed (rather Fritz 6 managed) to lose over 200 rating points!

1 comment:

  1. Houdini has recently come out with a new version which makes it equal or in some cases stronger than Rybka 4.

    As far as the controversy goes, I would not call Houdini a direct clone of Rybka. The programmer was honest and said he looked at all open source programs at the time like Stockfish, Crafty, Fruit and the Ippolit/Robbolito/Igorrit/Ivanhoe group.

    The Ippolit/Robbolito/Igorrit/Fires/Ivanhoes were probably derived from a reverse engineered Rybka so some of Houdini's code is from that. But he used much more of that and added his own ideas to push it maybe ahead of Rybka.

    Sort of a hybrid from everything available even if one (of four main ones) source has questions about it.

    Houdini really is the best of all modern chess programmer ideas to date.