The Houdini 1.5a x64 vs Deep Rybka 4.1 SSE42 x64 match will start tomorrow. The time control is 11:30:00.00 + 30 seconds per move, and there will be 30 rounds, each side playing White and Black of a specific opening. Here is the link:

http://www.westportchessclub.org/computer-chess/houdini-vs-rybka-30-games.htm

Round 1, Houdini vs Rybka, was a Sicilian Najdorf, and it ended in a draw.

FYI, I am making it so you can view each and every game and the details of them (how long each engine spent on a particular move, etc.).

View round 1 here: http://www.westportchessclub.org/computer-chess/houdini-vs-rybka-30-games-round-01.htm 

Round 2, Rybka vs Houdini, was a Sicilian Najdorf, and it ended in a draw. Rybka thought about move 37 for over 2 hours (2:08:48)!!

View round 2 here: http://www.westportchessclub.org/computer-chess/houdini-vs-rybka-30-games-round-02.htm

Round 3, Houdini vs Rybka, was a Nimzo-Indian, Main line, and it ended in a win for Rybka.

View round 3 here: http://www.westportchessclub.org/computer-chess/houdini-vs-rybka-30-games-round-03.htm

I have reservations about continuing this tournament, given the recent news of Rybka's alleged code thievery: http://www.chessvibes.com/reports/rybka-disqualified-and-banned-from-world-computer-chess-championships/

Round 4, Rybka vs Houdini, was a Nimzo-Indian, Main line, and it was a draw.

View round 4 here: http://www.westportchessclub.org/computer-chess/houdini-vs-rybka-30-games-round-04.htm

Round 5, Houdini vs Rybka, was a Ruy Lopez, Closed, Chigorin, and Houdini won.

View game 5 here: http://www.westportchessclub.org/computer-chess/houdini-vs-rybka-30-games-round-05.htm

Round 6, Rybka vs Houdini, was a Ruy Lopez: Closed, Chigorin, and it was a draw.

View game 6 here:

http://www.westportchessclub.org/computer-chess/houdini-vs-rybka-30-games-round-06.htm

The score is tied at 3 points each.

The problem with Rybka is a copyright violation.  People were claiming this right here a year or more ago, the code which forms the basis of the Rybka engines comes from a program called Fruit and another playing algorithm. 

This isn't saying the Rybka team didn't make strong improvements to the software.  But it was open source, free to use - that means you can use it freely and improve it, but you cannot then copyright the modifications.  This is what Rybka basically did.  It's the same thing as going into a national park and building a nice log cabin, then trying to claim the land because you improved it.  No banana, Koko!

But it's still a super-strong engine, not all of the unique strength is tainted by the attempt to end run the law on public domain.  I think there were possibly legal ways to do it, but the Rybka team chose not to try, but instead to deceive about the origin of the code.

There is still no getting around the sheer strength of the thing, though, so let the match proceed!

Round 7, Houdini vs Rybka, was a King's Indian, Mar del Plata. It was an obvious draw at move 53, but it dragged on for another 80 moves.

View game 7 here: http://www.westportchessclub.org/computer-chess/houdini-vs-rybka-30-games-round-07.htm

It's still tied up at 3.5 points each.

ivoryknight71, are you sure that there is any point in leaving them 24 hours as opposed to say 3 hours to go at it?

Because I would think that after a significant length of time, the amount of RAM you have comes hugely into play. Suppose Houdini is trying to decide between move a and move b... it will have to keep all of move a variations and subvariations in its memory as well as all of move b's variations and subvariations. Yeah it will rule out some of the subvariations, but then it's not searching to the same depth.

This is why even if there were some kind of quantum computer fast enough to "solve chess" we wouldn't have anything like the space to store it anyway. 7-piece tablebases take up terrabytes IIRC and 6 pieces take up something like 7 gigabytes.

good question. the engines have more than enough RAM for this tournament. the only time a [processor] core will come close to using 256 mb of RAM is if you let it think for 24 hours. i am only letting mine think for 12 hours and am giving them 1,024 MBs.

and i've already done the 3-hour thing...100 times: http://www.westportchessclub.org/computer-chess/rybka-vs-houdini-100games.htm

Round 8, Rybka vs Houdini, was a King's Indian, Mar del Plata. It ended in a draw by repetition after 66 moves.

View game 8 here: http://www.westportchessclub.org/computer-chess/houdini-vs-rybka-30-games-round-08.htm

It's still tied up (4 points each).

I was wondering how a Queen's Indian transposes into a King's Indian. The game was a King's Indian right from the start.

you know what. you are right. i don't know what i was thinking. nice catch.

Round 9, Houdini vs Rybka, was a Sicilian Dragon, Yugoslav. It ended in a draw.

View game 9 here: http://www.westportchessclub.org/computer-chess/houdini-vs-rybka-30-games-round-09.htm

It's still tied. 4.5 points each.

Round 10, Rybka vs Houdini, was a Sicilian Dragon, Yugoslav. Rybka won.

View Game 10 here: http://www.westportchessclub.org/computer-chess/houdini-vs-rybka-30-games-round-10.htm

Rybka came up w/ a novelty on move 32 that gave it the game I think, where it gave up a Rook for a Bishop and a pawn. Rybka had used up about 2 hours of the clock more than Houdini by move 27.

Rybka leads the match by a score of 5.5 - 4.5.

Thanx for keeping us updated, I dont normally comment, but I do check out the games...

If every variation of chess were solved with minimax pathways, and executed by a computer with a huge amount of fast RAM to search the database in time, how quickly do you thik it would defeat Rybka or Houdini?

How many moves do super GM's last against those programs, typically?

I don't know about the first part of your question, and not long is the answer for your last one. They even give IMs pawn or Knight odds and still get destroyed. I don't know of any GM that's taken the Knight odds challenge, however.

I was going to comment on how maybe you could edit the first post or have each new game post include the running score, but it seems you've started doing that. Thanks :D