Rybka 4 Fritz 12 Hiarcs 13 amp future


Back to forum

Thibault de Vassal    (2008-12-06)
Rybka 4, Fritz 12, Hiarcs 13 & future...

What about a small point on chess engines ?

A few months ago, it was quite predictable to me that Chessbase engines (well, Rybka 3 actually also is a Chessbase engine now) like Fritz, Shredder & Hiarcs were dedicated to catch Rybka in the computer chess rating lists.

The current CEGT 40/20 (AMD 4200+) rating list shows :

Rybka 3 x64 4CPU 3202
Rybka 2.3.2a x64 4CPU 3079
Deep Fritz 11 4CPU 3031
Zappa Mexico II x64 4CPU 3022
Shredder WM Edition Bonn 4CPU 3011
Naum 3.1 x64 4CPU 3011
Hiarcs 12 MP 4CPU 2968

In other words, Rybka 3 always has a quite large advance, but all other ones filled the major part of the gap with Rybka 2.3.2, including free chess engines. What future for chess engines now ? What kind of improvement can we expect from Rybka 4, Fritz 12, Shredder 12, Hiarcs 13 and co. ?

IMO one future version in theory may reach 3600 or more in such rating lists (which probably doesn't mean anything compared to the human rating list) but the interest of programmers may now decrease in this race where efforts/money can be compared to the grandmasters involvement to enter the elo top 100... What do you think ?

Andrew Stephenson    (2008-12-06 21:51:23)
Reaching a peak

My gut feeling is that rating improvements will tail off and we will not see any program crack 3400. I dont know the sales figures but looking for example at New In chess analysis by Carlsen he seems to use only Rybka and perhaps this program is becoming completely dominant among GM's. Perhaps the biggest impact will be hardware improvements allowing faster deeper analysis. This will mean fewer points missed and quicker conclusions as the time needed for the program to dig into the position shortens. You can still see examples of theoretical analysis in recent New in Chess Year books where misjudgements have been made because they needed to keep the program running a bit longer to see the evaluation flip but I think this will decrease ........

Thibault de Vassal    (2009-01-04 14:14:38)
Rybka 4

Vasik Rajlich wrote a few comments on what can be expected from the Rybka engine/team in 2009 :


Three engine releases are planned in 2009:

* Rybka 3+ - Rybka 3 playing strength, with bugfixes and cosmetic improvements, for Rybka 3 customers
* Pocket Rybka 3 - published by Convekta/ChessOK, packaged with their Pocket Champion interface, conforms to S. Tsukrov's Pocket-UCI protocol
* Rybka 4 - better search, better eval, new analysis features

"Our tournament goal for 2009 is to win a top freestyle event in 100% automated mode. (...)"

Very interesting, a new challenge for Eros :)

Wayne Lowrance    (2009-01-05 18:06:24)
Rybka 4 Fritz 12, Hiarcs 13 & future

It is my thought that Vas is running close to empty on improvements from Rybka 3 to Rybka 4. Rybka 3 was a huge, huge improvment.Other programmers are getting closer, specifically Naum. I sorta think Vas is looking a ways in the future when cluster comps are taking hold. Before Rybka 3 release there was a lot of excitement about Monte Carlo, but Rybka 3 Monte Carlo is not effective overall, It is just a novelity as far as my evaluation is concerned. Maybe too, like cluster MC will have a place. Correspondence chess players sh ould be delighted with Rybka three. A few minor improvements can be made, but wont improve ELO much, such as deep evaluation pv listing. Right no for CC player the current PV thing is crazy. Well nice topic Wayne

Wayne Lowrance    (2009-01-05 18:09:14)
Nutz, sorry

my last post I hit "post" when I wanted to review first to cull my typo's Wayne

Thibault de Vassal    (2009-01-18 18:20:34)
SSDF rating list

Current SSDF ratings for the best chess engines running on 2GB Q6600 2.4 GHz are :

Deep Rybka 3 x64 - 3226
Deep Fritz 11 - 3086
Zappa Mexico II x64 - 3064
Naum 3.1 x64 - 3046
Deep Shredder 11 - 3043
Deep Hiarcs 12 x64 - 3033
Hiarcs 11.2 MP - 3008
Deep Junior 10.1 - 2981

By the way, how long before a new Deep Junior version...