StockFish-1.8-w32 vs (Houdini + Fire + Deep_Shredder) 40/30

[IGarcia]

Hello.

I'm new here, besides I'm a forum reader for long time. So... HI!

I will posting soon a pgn file with a gaunlet tournament, almost finished. The idea was to test SF1.8 w32 JA against the best I had here:

Houdini 1.02 w32 2cores
Fire 1.31 KLO compile w32
Deep Sredder 12 UCI w32

Games are 40 moves in 30 minutes repeating, so we have some games over 2 or 3 hours.

Conditions: Hash: 512MB, Tablebases: On, Ponder Off
Harware: Intel core 2 Duo E8500 @ 3.16 GHz
Windows XP 32 - 4Gb Ram (windows uses only 3+ Gb)
Benchmark: ArenaMark: 26,67 AM, 1,32 EF

Tournament start: 2010.07.05, 21:25:00

29 of 30 games played (only 1 to go)

Code: Select all

Rank	 Engine                      Score         %  St
1       Stockfish-18-32-ja     16.0/29  55.1  ----------	
2       Fire_131_w32_KLO        5.0/10  50.0  0===1==1=0	
2       Houdini_w32_2CPU        5.0/10  50.0  ==========	
4       Deep_Shredder_12_UCI    3.0/ 9  33.3  0==0==100	

So this are long time controls, but engines must have rhythm to keep in the 30 minutes limit per each 40 moves.

[S.Taylor]

Thanks!
So what is this seeming to indicate about SF 1.8?

[IGarcia]

Code: Select all

Rank    Engine                      Score         %  St
1       Stockfish-18-32-ja     16.5/30  55.0  ----------   
2       Fire_131_w32_KLO        5.0/10  50.0  0===1==1=0   
2       Houdini_w32_2CPU        5.0/10  50.0  ==========   
4       Deep_Shredder_12_UCI    3.5/10  35.0  0==0==100=  

There was a lot of draws but some games still very nice.

ziped pgn file 40 kb: http://www.sendspace.com/file/3d5al8

SF18 and Houdini 1.02 was a complete draw.
This make me think positive on SF1.8 in the upcoming match will host Martin Thoresen in long TC and in a super i7 computer.


This little amount of games are not enought for ELO calculations, but using Deep Shredder 12 UCI 32b as a 2800 mark, SF1.8 is about 2900. This is in-line with other tests (for example IPON and SWCR)

[S.Taylor]

Code: Select all

Rank    Engine                      Score         %  St
1       Stockfish-18-32-ja     16.5/30  55.0  ----------   
2       Fire_131_w32_KLO        5.0/10  50.0  0===1==1=0   
2       Houdini_w32_2CPU        5.0/10  50.0  ==========   
4       Deep_Shredder_12_UCI    3.5/10  35.0  0==0==100=  

There was a lot of draws but some games still very nice.

ziped pgn file 40 kb: http://www.sendspace.com/file/3d5al8

SF18 and Houdini 1.02 was a complete draw.
This make me think positive on SF1.8 in the upcoming match will host Martin Thoresen in long TC and in a super i7 computer.

Was Houdini on stronger hardware than Stockfish in your test now? Or the same?

[IGarcia]

All in same hardware, all engines using both cores 100% when it was their turn. Same conditions for all.

What I did not mention was opening: Used Arena Mainbook up to move 5 per side, this is 10 half moves.

It can be very frustrating to do a match between SF18 and Houdini 1.02, say 48 games with very long times, to only see 3 or 4 games not draws.

[S.Taylor]

All in same hardware, all engines using both cores 100% when it was their turn. Same conditions for all.

What I did not mention was opening: Used Arena Mainbook up to move 5 per side, this is 10 half moves.

It can be very frustrating to do a match between SF18 and Houdini 1.02, say 48 games with very long times, to only see 3 or 4 games not draws.

So you say that the tie with Houdini 1.2 makes you feel positive about the upcoming Thoresen match vs Rybka 4, but it only makes me feel it might be similar, but not necesarily do better than Houdini did.

In a few days, there will be a Houdini 1.3, also. We know nothing about what this will be like.

Perhaps the only great thing about SF might be that it might be the highest rated, below Rybka 4, that no one suggests might be a clone/derivative.

[IGarcia]

So you say that the tie with Houdini 1.2 makes you feel positive about the upcoming Thoresen match vs Rybka 4, but it only makes me feel it might be similar, but not necesarily do better than Houdini did.

In a few days, there will be a Houdini 1.3, also. We know nothing about what this will be like.

Perhaps the only great thing about SF might be that it might be the highest rated, below Rybka 4, that no one suggests might be a clone/derivative.

I see Houdini 1.02 has some small advantage over StockFish 1.8 in fast games. That advantage seems to disappear with a longer time control. And that is what makes me think SF1.8 can get better on longer times.
Not suggesting SF will beat R4, but it will probably do well.

See, On other post I read something like : "Why a match SF1.8 vs R4 if stockfish engine is weaker tha Ippos" "Better a SF1.8 vs Houdini 1.02"

And I cant see that. I'm confident with Houdini future. Also with SF future!