Impressive Stockfish 1.4 !!

[Tomcass]

I decided to test this new program against two pro's (Deep Fritz 11 and Deep Hiarcs 12) and the best amateur (IMHO) Toga II 1.4.1 SE6.

Environment:

Hiarcs12 GUI
10 minutes game.
Stockfish 1.4 against the other three programmes. 20 games each.
Ponder: Off
Books:
Stockfish 1,4: CompMaster Beta (Ok, this is for me the best free book in the world today)
Deep Fritz 11 : Fritz 11
Deep Hiarcs 12: Hiarcs 12
Toga II: Magnificient 4.
Computer: Intel i7 975

Results:

2009-07Stockfish1.4 2009

Stockfish 1.4 JA 64bit(x8) - Deep Fritz 11 13.0 - 7.0 +8/-2/=10 65.00%
Stockfish 1.4 JA 64bit(x8) - Toga II 1.4.1SE6(x8) 10.5 - 9.5 +8/-7/=5 52.50%
Stockfish 1.4 JA 64bit(x8) - Deep Hiarcs 12 13.5 - 6.5 +10/-3/=7 67.50%

Although short, this is for me a clear, convincing test. Not only for the result (I expected between 40 and 45% and STfish has got more than 60%) but also for my 'feeling' of being in front of an extremely strong program. Congratulations to the author.

Regards from Barcelona.

Tom Casanovas.

[mcostalba]

Thanks for testing, Tom

Although I would think different books played a role here.

It seems also Magnificient 4, the book used by Toga II is very good because Toga II perhaps is stronger then Stockfish but for sure weaker then Fritz 11.


For people that horrifies reading 20 games each: Of course 20 games are too few, but here we are just having fun with some talking

[Tomcass]

Thanks for your comments, Marco.

I think you are right when you say that books have some effect in my test. I agree with you. Anyway, at this moment my feeling is that Stockfish1.4 is the stronger free amateur program. I can be wrong, obviously.

I have just started to test it againt the two monsters: Rybka3 and Naum4. Sotckfish1.4 again with CompMaster Beta3 book whereas Rybka3 and Naum4 with Perfect 15 book. My guess: around 25-30% against Rybka and around 30-35% against Naum4. Let's see.

Tom.

[ArmyBridge]

Why don´t your try with a set of positions? like NoomenTestsuite2008
Regards

[ernest]

my feeling is that Stockfish1.4 is the stronger free amateur program.

Could be...
My first match Stockfish 1.4 64-bit against Thinker 5.4Di 64-bit
(2-core/2-core, no pondering, 3GHz Core2 Duo, 2'+1", Book PB5moves)
ended 54 - 46 (+36 -28 =36)

Previous Stockfishes could not win against that Thinker.

[matejst]

I ran a match with that test suite against ST 1.3 yesterday on my laptop. The elo gain was +54 after 60 games (1'2", 2t).

I am testing some strong free engines at 5'5", with the Nunn suite, and on two cores, ST 1.3 has a rating of about 3000 elos (CCRL). But the thing is that in the Nunn suite, the positions arising from the french defense, the english opening, the Ruy Lopez, are clearly in favor of slower engines with an elaborate evaluation function, therefore the difference in elos compared with the CCRL results (can send the pgn if somebody want it).

Some remarks: Stockfish 1.3 plays often positions with material imbalance (and loses them always against R 2.2); in closed positions, it beats the Fruit derivatives quite easily, Rybka 2.2 also. It is weaker in open positions with less material.

ST 1.4, in that mini match against ST 1.3 with the Noomen suite, usually manages to equalize worse positions in the endings, and clearly plays better complicated middlegames.

[ArmyBridge]

Hi Boban , could you send me the games?
Tnx in advance

[matejst]

Send you both pgn tomorrow afternoon.

[tano-urayoan]

Anyway, at this moment my feeling is that Stockfish1.4 is the stronger free amateur program. I can be wrong, obviously.

What about Thinker?

[ernest]

My second match Stockfish 1.4 64-bit against Thinker 5.4Di 64-bit
(2-core/2-core, no pondering, 3GHz Core2 Duo, 2'+1", Book PB5moves)
ended 47.5 - 52.5 (+28 -33 =39)

So, based on these matches alone, Stockfish 1.4 64-bit and Thinker 5.4Di 64-bit are about the same strength.