Nalimov and the E-E-T...

[Glarean]

.

See an interesting comparison:
The E-E-T (100 Endgame Positions) with and without Tablebases

Code: Select all

Engine                               Solved      Difference


Rybka3(2CPU)__Mit EGTB               75
Rybka3(2CPU)__Ohne EGTB              74           1

Zappa MexicoII(2CPU)__Ohne EGTB      71
Zappa MexicoII(2CPU)__Mit EGTB       71           0

Colossus2008b__Mit EGTB              69
Colossus2008b__Ohne EGTB             63           6

Hiarcs11.2__Mit EGTB                 68
Hiarcs11.2__Ohne EGTB                59           9

Movei00.8.438__Mit EGTB              66
Movei00.8.438__Ohne EGTB             62           4

Fritz11__Mit EGTB                    64
Fritz11__Ohne EGTB                   58           6

Naum3.1(2CPU)__Mit EGTB              66
Naum3.1(2CPU)__Ohne EGTB             62           4

SmarThink1.00__Mit EGTB              61
SmarThink1.00__Ohne EGTB             55           6

ChessTiger2007.1__Mit EGTB           60
ChessTiger2007.1__Ohne EGTB          52           8

DeepShredder11(2CPU)__Mit EGTB       58
DeepShredder11(2CPU)__Ohne EGTB      57           1

Fruit05/11/03__Mit EGTB              57
Fruit05/11/03__Ohne EGTB             55           2

WildCat8__Ohne EGTB                  56
WildCat8__Mit EGTB                   48           8

Yace0.99.87__Mit EGTB                56
Yace0.99.87__Ohne EGTB               53           3

Pharaon3.5.1(2CPU)__Mit EGTB         55
Pharaon3.5.1(2CPU)__Ohne EGTB        53           2

The Baron1.8.1__Mit EGTB             53
The Baron1.8.1__Ohne EGTB            50           3

Gandalf6.0__Ohne EGTB                51
Gandalf6.0__Mit EGTB                 47           4

Crafty23.00(2CPU)__Mit EGTB          48
Crafty23.00(2CPU)__Ohne EGTB         47           1

Junior10.1__Mit EGTB                 47
Junior10.1__Ohne EGTB                43           4

Anaconda2.0.1__Mit EGTB              46
Anaconda2.0.1__Ohne EGTB             43           3

Nimzo8__Ohne EGTB                    45
Nimzo8__Mit EGTB                     42           3

Quark2.35__Mit EGTB                  39
Quark2.35__Ohne EGTB                 37           2

SOS5__Mit EGTB                       36
SOS5__Ohne EGTB                      35           1

AnMon5.60__Ohne EGTB                 29
AnMon5.60__Mit EGTB                  28           1

Delphil1.9__Mit EGTB                 26
Delphil1.9__Ohne EGTB                23           3

Homer2.0__Ohne EGTB                  24
Homer2.0__Mit EGTB                   22           2

More details (german):
http://glareanverlag.wordpress.com/2009 ... -tabellen/

The E-E-T (german):
http://glareanverlag.wordpress.com/2007 ... endspiele/

Regards: Walter

.

[Norm Pollock]

Very interesting. Thank you Walter. I would also like to see a third column indicating the number of endings solved by both mit & ohne (with & without), and a fourth column indicating the number of endings NOT solved by either.

-Norm

[PauloSoare]

Try StockFish 1.3 64 bits in your test, Walter. On my Q6600, with 512 Mb,
it do more points than Rybka and Zappa. I don´t tryed Stockfish 1.4 yet.

[Glarean]

Try StockFish 1.3 64 bits in your test, Walter. On my Q6600, with 512 Mb,
it do more points than Rybka and Zappa. I don´t tryed Stockfish 1.4 yet.

Stockfish doesnt know the EGTB...

Walter

[PauloSoare]

I tested without EGTB. What amazed me was the number of points, 86.

[Werner]

Here are some results with egbb 5-men:

Fruit 2.3.1 with egbb: 54 von 100 Zügen gefunden
Fruit 2.3.1 without egbb: 61 von 100 Zügen gefunden (fruit can use too Nalimov egtb! - here disabled)

Bright-0.4a mit egbb: 74 von 100 Zügen gefunden
Bright-0.4a ohne egbb: 67 von 100 Zügen gefunden

TogaII142JD mit egbb: 61 von 100 Zügen gefunden
TogaII142JD oh. egbb: 58 von 100 Zügen gefunden

Scorpio21 mit egbb: 56 von 100 Zügen gefunden (egbb im RAM)
Scorpio21 ohne egbb: 56 von 100 Zügen gefunden

Sloppy022x64 m. egbb: 28 von 100 Zügen gefunden (egbb im RAM)
Sloppy022x64 o. egbb: 30 von 100 Zügen gefunden

...with some surprises

Werner

[BBauer]

Hi Walter,

thanks for sharing your results.
However, you say nothing about the way the tablebases are accessed.
Assume tablebases on a very slow medium. Use may hurt.
Assume tablebases on a very fast medium. Now its a question of implementation. If tablebase access is as fast as memory access tablebases should be usefull. Btw. Stockfish1.4 gives for me 75 solutions - without tablebase access.
Kind regards
Bernhard

[Norm Pollock]

Hi Walter,

thanks for sharing your results.
However, you say nothing about the way the tablebases are accessed.
Assume tablebases on a very slow medium. Use may hurt.
Assume tablebases on a very fast medium. Now its a question of implementation. If tablebase access is as fast as memory access tablebases should be usefull. Btw. Stockfish1.4 gives for me 75 solutions - without tablebase access.
Kind regards
Bernhard

Good point Bernhard. I access EGTB on an 8G flashdrive. Fast and it doesn't wear out any hard drive.

[Glarean]

Hi Walter,

thanks for sharing your results.
However, you say nothing about the way the tablebases are accessed.

Read the report please:
http://glareanverlag.wordpress.com/2009 ... -tabellen/

Regards: Walter

.

[BBauer]

Hi Walter,
I read your report but I'm perhaps too stupid to understand why I should have read it. Anyway, I ran your test with Rybkav2.2n2mp.x64 under Arena
60 sec 74/100 37:53 with egtb used frequently, 5 men egtbs on USB stick
60 sec 74/100 38:10 no egtb used
Computer: AMD X2 2.6 GHz 2 procs used

kind regards
Bernhard