iCE 3 gets released

[tpetzke]

Thanks, I hoped for something in the 3 digit area as in fast selfplay iCE v3 did really well against v2.

So if it holds that would be cool.

[elcabesa]

congratulations for your hard work, Ice seems very very strong

[cdani]

Congratulations!!

[Steve Maughan]

Brilliant progress Thomas - congrats!

Steve

[Wolfgang]

http://cegt.forumieren.com/t503-testing ... 8-x64#1178

Now ca. +136 after 1300 games!!

[tpetzke]

Hi Frank,

just had a look at your site. You put iCE 3 656 in your queue. Sure that it is this sub-version? This version has an error (illegal move) in the internal book and should not be used. Did I accidentally put this version in one of my archives?

The official release is iCE 3 v658. I cross checked my website and it seems 658 is here online. So if you have a 656 please throw it away and switch to 658.

In this version the error is fixed.

Thomas...

[tpetzke]

Lars Hallerström has finished its initial tests and it looks like iCE gained 140 ELO. The early values from CEGT seem to confirm that.

A more beautiful table is here

http://macechess.blogspot.de/2016/01/ic ... -gain.html

Note that most of the opponents are using 4 cores, where as iCE is using only 1.

Code: Select all

33 Ice 3.0 (64)                   : 3054   18  18  1000    60.8 %   2978   31.3 %
73 Ice 2.0 (64)                   : 2911   14  14  1640    53.0 %   2891   30.8 %


33 Ice 3.0 (64)              : 3054  1000 (+451,=313,-236), 60.8 %


Fruit 2.3.1                   :  40 (+ 30,=  8,-  2), 85.0 %
Ktulu 9                       :  40 (+ 29,=  9,-  2), 83.8 %
Deuterium 14.01 (64)          :  40 (+ 29,=  6,-  5), 80.0 %
Bug2 1.9 (64-4cores)          :  40 (+ 32,=  6,-  2), 87.5 %
Sjeng WC2008 (64-4cores)      :  40 (+ 21,= 11,-  8), 66.2 %
Scorpio 2.7.6 JA (64-4cores)  :  40 (+ 20,= 12,-  8), 65.0 %
Gaviota 1.0 (64-4cores)       :  40 (+ 10,= 21,-  9), 51.2 %
Rybka 4.1 SSE42 (64-4cores)   :  40 (+  4,= 12,- 24), 25.0 %
Chiron 2 (64-4cores)          :  40 (+ 12,= 10,- 18), 42.5 %
Spike 1.4 (4cores)            :  40 (+  9,= 19,- 12), 46.2 %
Senpai 1.0 sse42 (64-4cores)  :  40 (+ 12,= 15,- 13), 48.8 %
Spark 1.0 (64-4cores)         :  40 (+ 17,= 13,- 10), 58.8 %
Critter 1.6a (64-4cores)      :  40 (+  2,= 18,- 20), 27.5 %
Cheng 4.37 rc0 (64-4cores)    :  40 (+ 20,= 10,- 10), 62.5 %
Texel 1.05 (64-4cores)        :  40 (+  9,= 14,- 17), 40.0 %
Hannibal 1.5 (64-4cores)      :  40 (+  6,= 19,- 15), 38.8 %
Arasan 18.1 (64-4cores)       :  40 (+ 13,= 19,-  8), 56.2 %
Rodent 1.7 (64)               :  40 (+ 29,=  8,-  3), 82.5 %
Tornado 7 (64-4cores)         :  40 (+ 25,=  9,-  6), 73.8 %
Octo 5190 (64-4cores)         :  40 (+ 30,=  7,-  3), 83.8 %
Pedone 1.3 (64-4cores)        :  40 (+ 30,=  5,-  5), 81.2 %
BobCat 7.1 (64-4cores)        :  40 (+ 16,= 15,-  9), 58.8 %
Crafty 25.0 JA (64-4cores)    :  40 (+ 18,= 14,-  8), 62.5 %
Ex 7.88b (64-4cores)          :  40 (+ 12,= 19,-  9), 53.8 %
SmarThink 1.80 (64)           :  40 (+ 16,= 14,- 10), 57.5 %

[Sylwy]

Lars Hallerström has finished its initial tests and it looks like iCE gained 140 ELO. The early values from CEGT seem to confirm that.

I guess the next CCRL 40/4 update is going to show similar improvement.

Hi Gabor !

The improvement is absolutely impressive . My 32-bit (unfortunately not a 64-bit one) test result will be posted here in 3-4 hours !

[Sylwy]

http://talkchess.com/forum/viewtopic.ph ... 87&t=50060

[Manza]

Good news!!