A Look at Rybka 4 - Houdini 1.5a, Game 1

A while ago I reported that the free program Houdini 1.5a defeated the king of the commercial programs, Rybka 4, by a 23.5-16.5 score in a recent match. Many people, both on my blog and elsewhere, were especially impressed by Houdini's play in the first game of that match, and I can't blame them! Houdini sacrificed three pawns for play, and the end result was an overwhelming initiative. You can have a look at the game, with my comments, here.

Comments

[Jeffrey "notyetagm" Hall]

Thanks for annotating this game.

I cannot tell you how much this game impressed me. Engines have come a *long* way from the greedy pawn-grabbers of old.

[:D]

Cool game and the new board/server whatever you are using to annotate games is much nicer than the old one.

[DM: I like it better too, but the credit (and earlier blame!) go to Martin Thoresen - the TCEC page is his site. The annotations are mine, but all the formatting and everything else there is his.]

[JH]

Very impressive play by Houdini. By appearance the game is not a comp/comp encounter
but a comp/human master encounter.

[Larry Conn]

Dennis, thanks for annotating this very interesting game!

Question about the TCEC format: Under the old format, you could click through the annotations and the pieces would move accordingly. I haven't been able to figure this out here; nor can I move the pieces manually. For us patzers, visualizing the board through the entire analysis chain can be a bit challenging. Am I missing something?

[DM: As I stated two comments back, I had nothing to do with the formatting or anything else on the website; all I contributed were the annotations. If you look at the bottom of the TCEC page and click on his (Martin Thoresen's) name, you can email him with technical questions. That said, while I don't see any way to examine the variations the way you wanted (and the way that was possible on the old format), you can click on "Download all games". That gives you the PGN, which you can cut and paste into Fritz, ChessBase or whatever database program you use.]

[Larry L]

Great effort Dennis on annotating this complex game. I appreciated your insights on the game. I think the reason the game dragged on after Black clearly was winning, was Thoresen's threshold for resigning is set fairly high he notes it on the site under "Information"
It was Martin Thoresen first told me to take a look at the game a few days after it was played. I was asking him a technical question about the engines that he uses when he said look at game one its amazing. I had seen sacrifices of material for positional edges before but nothing like this. Your analysis added even more to the intrigue of how the engine found such an approach.
Having both engines I ran them simultaneously to see how they saw the game. Rybka evaluated the game in its favor upto around move 20 as giving it a small edge, Houdini thought it had equality. Move 24 seems to be where Rybka got things really wrong. It seemed to think 24 bxa5 gave it equality while Houdini saw it as a bad move immediately and prefered 24 f4 with a small edge for Black.
At least from my experiments comparing their analysis of positions, except for clear cut forcing lines, these two engines do not think alike. By that I mean they don't just give a higher or lower evaluation of the same move they often follow completely different lines of play.