Long TC matches with Houdini 3 Beta

[Houdini]

I'm currently running the first long Time Control (TC) matches between Houdini 3 Beta and three other top engines: Houdini 2.0c, Stockfish 2.3.1 and Komodo 5. Against each opponent a 120-game match will be played at 90 min+30 sec/move TC, starting from 60 Noomen opening positions played from both sides.
I've never run any serious match slower than 2 min+2 sec/move with Houdini 3, so it's quite interesting to see how well the Houdini 3 strength improvement scales at about 30 times longer TC.

Detailed match conditions are:
- 120 games against Houdini 2.0c, Stockfish 2.3.1, Komodo 5
- 90+30 TC, single-core, 512 MB hash, 60 positions from Noomen 2006/2008 test suites played from both sides.
- No table bases, resign after 3 successive scores larger than 5.00, draw after move 120 if both evaluations below 0.10.
- Hardware: AMD Opteron 6274 Server (SSE4) producing approx. 1 MN/sec per core with Houdini.
- The Houdini 3 version that is played can be considered a Houdini 3 "Beta": close to, but not exactly the final release.
- The error margin on the individual 120-game match results will be about ± 45 Elo.

The first match has ended earlier today, the final result was:
Houdini 3 - Houdini 2.0c: +48 -15 =57 or 76.5-43.5 (+94 Elo ± 42 Elo).
You can download the games.

The result is better than I expected, which was about 60-65 Elo (based on 2'+1.2" results). Clearly Houdini 3 has had a favorable run.
I'm also pleased with the high number of decided games, producing less than 50% draws is very good at long TC.

The Houdini 3 - Stockfish 2.3.1 match is now running, current standing is +10 -5 =11 after 26 games.

Cheers,
Robert

[beram]

hi Robert, quite impressive, please release this beast

[Vinvin]

hi Robert, quite impressive, please release this beast

As he wrote in another thread : around 10th of October ! Only 2 weeks to wait

[Uri Blass]

Here is a position from the match(houdini3 beta-houdini2c) when houdini3 beta blundered

[D]k1n4r/2q5/b2b1p2/N1pPpPnB/1pP1P1P1/p2Q4/P2N2KR/B7 w - - 23 77

After Nab3 I suspect that even pablo can draw with white against houdini2 but houdini3 played Ndb3 and lost(I am not sure if objectively white is losing after Ndb3 but there is no reason to play this move)

[MM]

Here is a position from the match(houdini3 beta-houdini2c) when houdini3 beta blundered

[D]k1n4r/2q5/b2b1p2/N1pPpPnB/1pP1P1P1/p2Q4/P2N2KR/B7 w - - 23 77

After Nab3 I suspect that even pablo can draw with white against houdini2 but houdini3 played Ndb3 and lost(I am not sure if objectively white is losing after Ndb3 but there is no reason to play this move)

Interesting, i tend to agree, without computer assistance i can say that i think that c4 seems to be very weak and perhaps with the pressure on e4, the hole in a4 for a black piece and the possible threat of the push in b3 black could have some chances. If you look at white pieces, they are totally passive, especially the poor bishop a1.

Best Regards

[Matthias Gemuh]

I'm currently running the first long Time Control (TC) matches between Houdini 3 Beta and three other top engines: Houdini 2.0c, Stockfish 2.3.1 and Komodo 5. Against each opponent a 120-game match will be played at 90 min+30 sec/move TC, starting from 60 Noomen opening positions played from both sides.
...
Robert

At 90 min+30 sec/move TC, Stockfish can lose on time, depending on how the GUI defines the time increment:

a) if it is a bonus given only as a reward for playing a move before time-out, Stockfish may fail. Winboard and ChessGUI define an increment this way.
TCs like 0+y therefore not possible.

b) if it is a free gift given up-front before each move, Stockfish may be fine.
Fritz-GUI defines an increment this way, I think.
TCs like 0+y are possible.


Matthias.

[Modern Times]

For Stockfish I set "emergency base time" as 1000 instead of the default 200 for my repeating time control matches. Stockfish has in the past lost on time sometimes otherwise. Once I changed this to 1000, I have not had a problem since.

[Houdini]

At 90 min+30 sec/move TC, Stockfish can lose on time, depending on how the GUI defines the time increment:

I've played many games with Stockfish at around 10"+0.1" with cutechess-cli and observed no time losses, so I guess cutechess-cli does it right.

The current match standing is Houdini 3 - Stockfish 2.3.1: +26 -10 =31 after 67 games. No time losses so far.

[sainzlei]

Hurry to release H3 , and all we will test it for you !

[Matthias Gemuh]

At 90 min+30 sec/move TC, Stockfish can lose on time, depending on how the GUI defines the time increment:

I've played many games with Stockfish at around 10"+0.1" with cutechess-cli and observed no time losses, so I guess cutechess-cli does it right.

...

You misunderstood me totally. 10'+0.1" would not expose the Stockfish problem under Winboard/ChessGUI either. 1'+10" is much more likely to expose it, perhaps under cutechess-cli also.