Fruit & Toga comparison with Rybka

[Mike S.]

4m+2s, AthlonXP@1200 MHz
Noomen 2006 openings (30)
96 MB hash each, 32 MB tbs. cache
ponder off
3-/4- and R vs. R 5-piece Nalimovs
(no Bitbases)
Windows 98SE, Chessprogram8-GUI

Typical search depths ~11...13 ply. 60 games each.

Code: Select all

Rybka 2.3.2a 32-bit   - Toga II 1.3.1           44.0 - 16.0    +33/-5/=22    73.33%
Rybka 2.3.2a 32-bit   - Toga II 1.3.4           44.5 - 15.5    +36/-7/=17    74.17%
Rybka 2.3.2a 32-bit   - Fruit 51103             43.0 - 17.0    +30/-4/=26    71.67%
Rybka 2.3.2a 32-bit   - Fruit 2.3.1             42.5 - 17.5    +32/-7/=21    70.83%
Rybka 2.3.2a 32-bit   - Fruit 2.3.1 xch(*)      43.0 - 17.0    +33/-7/=20    71.67%

*) Extend Checks ON

The results are within a narrow margin and do not allow any conclusion about which of these "Fruitoga" engines is stronger. - I think the check extensions did not harm the game performance of Fruit 2.3.1 (at least not against Rybka under these conditions) and I think they are an improvement for analysis.

The Elo performance difference to Rybka 2.3.2a is approx. -170 (here).