Petir 4.39: #18 free single-CPU engine in CCRL 40/4

[Kirill Kryukov]

Petir 4.39 has just finished some gauntlets and has 991 games now in CCRL 40/4 time control. It is now rated 2673 ELO points, which makes it #18 free single-CPU engine in our blitz list.

Free single-CPU rating list (my version which includes only stable public releases and only default settings)

All results of Petir 4.39

Its closest friends:

Code: Select all

 Rank                 Engine                  ELO   +    -   Score  AvOp  Games
   15 SOS 5.1                                2684  +14  -14  47.1%  +20.4  1865
   16 Pseudo 0.7c                            2678  +13  -13  51.9%  -18.1  2329
   17 Ruffian 1.0.5                          2675  +15  -15  49.1%   +7.2  1603
=> 18 Petir 4.39                             2673  +19  -19  43.7%  +45.9   991
   19 Aristarch 4.50                         2666  +14  -14  45.7%  +32.5  1929
   20 The Baron 1.8.1                        2653  +21  -21  45.5%  +32.4   789
   21 Smarthink 0.17a                        2650  +22  -22  45.9%  +28.2   702

Based on the games to current date, Petir 4.39 is a certain improvement over version 3.99d, with 53 ELO points difference and estimated LOS (Likelihood of superiority) of 100.0%.

Comparison of all Petir versions we tested

Good work by Peter Alloysius! I hope this speed of improvement continues with future versions!

Best,
Kirill

[Kirill Kryukov]

Petir was pushed to #20 position in CCRL 40/4 Free Single-CPU list, after expected addition of Strelka 1.0b and less expected improvement in Colossus 2007a.

Updated CCRL 40/4 Free Single-CPU list (My version which includes stable public releases and default settings only).

[Tony Thomas]

Wow, Strelka's rating is phenomenal. Since it is a bitboard engine, it should be able to gain additional 30 or so points just from a 64bit compile. Also, since there are many obvious problems with the program, fixing them should help Yuri challenge the position of Rybka.

[Marc MP]

Wow, Strelka's rating is phenomenal. Since it is a bitboard engine, it should be able to gain additional 30 or so points just from a 64bit compile. Also, since there are many obvious problems with the program, fixing them should help Yuri challenge the position of Rybka.

Fully agree with Tony. Strelka is losing an awful lots of (rating) points due to is lack of basic endgame knowledge. Fixing these plus the hash, would yield Strelka at the level of Toga (my guess). That is about 20 elo behind Rybka beta. If Strelka can benefits 64 bits, then when have a clear second for free 64-32 bits, ahead of Toga. Lets see. For me Brian Hofmann compile already rocks compared to the one from sdchesss.ru.