Deep Junior 13 Welcome Test

[Jouni]

Stockfish is showing the value of ICGA Wch title: +4, = 1 against Junior (soon +5) Nor has Junior scored a single win against Critter/Rybka.

Jouni

[Laskos]

Stockfish is showing the value of ICGA Wch title: +4, = 1 against Junior (soon +5) Nor has Junior scored a single win against Critter/Rybka.

Jouni

Yes, the score will be 4:12 against SF, Rybka, Critter, a 25% performance in 16 games, in line with short TC with more draws. Against Houdini it was some 42% in 20 games. The statistics is still weak, but it is possible that Houdini match, played on a different machine, had some problems. I don't believe in tuning, as Critter plays a similar chess to Houdini.

Kai

[Jouni]

Houdini match used Hiarcs book / 10 moves - maybe that is big difference?

Jouni

[Laskos]

Houdini match used Hiarcs book / 10 moves - maybe that is big difference?

Jouni

What book is used now? Some lines are short, some are longer. The exiting scores out of book are not that bad, and I could check the positions for fairness using many fast games, but I would need Houdini-DJ match PGN.

Kai

[Wolfgang]

- Noomen Opening Suite 2012 testset (30 openings), every engine has to play every opening with white and black

as Timo mentioned in his initial posting...

Best
Wolfgang

[Jouni]

I have feeling, that SF performance so far (6-1) is may be the strongest play in CCC history.

Jouni

[Damir]

Yeah, in the second game against Critter there was a break after move 39 which caused Critter to lose on time.

Now the matches are restarted with latest GUI version (F13).

Again the links to live broadcasts:
Deep Junior 13 vs. Critter 1.4a: http://www.team-oh.de/livegames1/
Deep Junior 13 vs. Stockfish 2.2.2: http://www.team-oh.de/livegames2/
Deep Junior 13 vs Deep Rybka 4.1: http://www.team-oh.de/livegames3/

Best regards
Timo

Hi Timo

Based on your results where Junior is facing several strong opponents, and is not able to win a single game, I have serious doubts about Clemens test, where Junior is playing VS Houdini, who appears to be stronger, than the engines you are testing it against. Either Houdini is playing with a handicap, or with a less cores than Junior, or something else which prevents it to play in its full strength.

PS: maybe you should have a talk with Clemens regarding this ?

[FriedmannC]

Hi. Please can you test the new Deep Junior 13 against Deep Shredder 12 and Naum 4.2 ? In this way we will see a lot more balanced games. Deep Junior 13 is being destroyed by Rybka, Stockfish and Critter. It would also be interesting to see Deep Junior 13 vs. Fritz 13 both running on a single core machine but at long time controls Also a Stockfish 2.2.2 - Houdini 2.c Pro at LTC will be exciting, as Stockfish seems to be one of Houdini's killers. Regards!

[TimoK]

I have feeling, that SF performance so far (6-1) is may be the strongest play in CCC history.
Jouni

Yeah, I'm very impressed, too. This "new" Stockfish seems to be a real killer...

Best regards
Timo

[TimoK]

Hi Timo

Based on your results where Junior is facing several strong opponents, and is not able to win a single game, I have serious doubts about Clemens test, where Junior is playing VS Houdini, who appears to be stronger, than the engines you are testing it against. Either Houdini is playing with a handicap, or with a less cores than Junior, or something else which prevents it to play in its full strength.

PS: maybe you should have a talk with Clemens regarding this ?

Hi Damir,

since I know Clemens as one of the best specialists in both computer hardware and computer chess tests I believe that his test was made under 100% fair conditions (no penalty for Houdini or something like that).

In the german CSS-Forum we are currently discussing the situation. There are many possibilties, why Junior scored much better in Clemens' test compared to mine (e.g. better book lines, statistical abberation, tuned against Houdini, luck?!). But there is one thesis I prefer: Clemens used a different time control: 120/40+120/40+120. That means time for each engine is doubled compared to my tests AND Junior is forced to ration its time more careful because there are time controls after the 40th and the 80th move. I watched Junior lose some games in my matches because it had too less time left in the endgame and so had to play very fast.

I'm thinking about restarting my matches with a different time control (not 6h/game for each player like in Clemens' test, but maybe 4h/game and divided more granularly into 3 time controls).

Best regards
Timo