Better Results with 1.7.1

[AdminX]

I just did another 100 Game match using the new 1.7.1 version. The results vs Rybka 3 were much better.

Blitz 5m

Q9000
Fritz 12 GUI
Mybook 21.2
2 Cores Each
1024 Hash
Ponder On

Stockfish 1.7.1 64bit
+37/=27/-36 50.50% 50.5 / 100

Rybka 3 64bit
+36/=27/-37 49.50% 49.5 / 100

(100 Games)

Games Download

[jpqy]

Hi Ted, was is with default settings or with Zugzwang detection.

[AdminX]

Hi Ted, was is with default settings or with Zugzwang detection.

I used the default Stockfish settings.

[Vinvin]

Defaut SF 1.7.1 = SF 1.7
SF 1.7.1 with zugzwang detection = same strength than SF 1.7
So, your results depends on randomness

my best,
Vincent

Hi Ted, was is with default settings or with Zugzwang detection.

I used the default Stockfish settings.

[Zagalo]

Defaut SF 1.7.1 = SF 1.7
SF 1.7.1 with zugzwang detection = same strength than SF 1.7

Actually, SF 1.7.1 without zugzwang detection = same strength than SF 1.7

[Vinvin]

To be more clear :
SF 1.7.1 with zugzwang detection : no measurable strength difference compare to SF 1.7
SF 1.7.1 without zugzwang detection : no measurable strength difference compare to SF 1.7

Defaut SF 1.7.1 = SF 1.7
SF 1.7.1 with zugzwang detection = same strength than SF 1.7

Actually, SF 1.7.1 without zugzwang detection = same strength than SF 1.7

[mcostalba]

To be more clear :
SF 1.7.1 with zugzwang detection : no mesurable strength difference compare to SF 1.7
SF 1.7.1 without zugzwang detection : no mesurable strength difference compare to SF 1.7

Sorry Vincent, but this is not more clear, this is wrong.

The only _real_ and proven rule is the one below:

Code: Select all

SF 1.7.1 without zugzwang detection  = SF 1.7

Sorry to be annoying but I would like to stop the mess while it is growing up as it, sadly, seems to happen.

I am sorry if the announcement was not clear enough, to me it seems clear, but I start to understund it was not, so I am trying to be as much clear as possible even if it could seems a bit rude.

Sorry in advance for this.

[BBauer]

My comparisons show different behavior between Sf1.7 and line 1391 commented out vs Sf1.7.1.
I can see this different behavior in positions played over the board and in zugzwang positions. I used only one proc.
So, if Sf1.7.1 would give a higher elo rating than Sf1.7, I wouldn't be surprised. Sf1.7.1 is strong and it has IMHO a lot of additional potential.
It may become or allready be the strongest analysis tool running on one proc. And the best is:
SF1.7.1 comes with sources!!
kind regards
Bernhard

[Ralph Stoesser]

I use the zugzwang detection patch from Marco Costalba for version 1.7.
Additionally I have added a condition to not try null move in case TT entry indicate it would be pointless to do so.

I also think the zugzwang detection version plays rather differently. I have no game statistics but from watching a couple of 5min/40moves bitz games against robbo* it seems zz-detection version plays slightly stronger.