Computer Chess Club Archives


Search

Terms

Messages

Subject: New PEJ-ratinglist: Hiarcs 10 overtakes Fruit 2.2.1 and is the new no. 1

Author: Per Jørgensen

Date: 06:50:37 02/26/06


Hello Chesscomputerfriends!

As I wrote in this Forum at several occations I'm making a new ratinglist. I am
using fixed openingpositions in stead of the openingsbooks of each engine
(or a common book for all engines). Here are the 10 fixed openingpositions I am
using for my tests:

1. Queens Gambit: 1.d4, d5 2.c4
2. Kings Gambit: 1.e4, e5 2. f4
3. Stonewall: 1.d4, e6 2.c4, f5 3.Sc3, d5 4.Sf3, c6
4. Caro Kann: 1.e4, c6 2.d4, d5
5. English: 1.c4, c5 2.Sc3, Sc6
6. Russian: 1.e4, e5 2. Sf3, Sf6
7. King- or Grünfeldindian: 1.d4, Sf6 2.c4, g6 3.Sc3
8. Sicilian: 1.e4, c5 2.Sf3, d6 3.d4, cxd4
9. Queens indian: 1.d4, Sf6 2.c4, e6 3. Sf3, b6 4. g3, Lb7 5. Lg2, Le7
10. French: 1.e4, e6 2.d4, d5

Every engine has got to play each openingposition with both white and black
meaning that a match between 2 engines consists of 20 games. I find that the
chosen fixed openingpositions provide a fair mix of open, tactical and closed
positional openings. I would like to thank Klaus Wlotzka (running the
CSS-Ratinglist) for support and advices in the whole process. It's my aim to
create a ratinglist of 25 top-engines.

I'm running my tests on this Hardware:

Asus A8V Motherboard, Socket 939, FSB 1.000 MHz
AMD Athlon 64 4000+, 2.4 GHz, 1 MB Level 2 Cache, 90 nm San Diego Core
2 x 512 MB Corsair PC 3200XLPT RAM (CAS Latency 2-2-2-5)

And here are the specific testconditions:

Windows XP Pro
Hiarcs 10 GUI
10 fixed openingpositions, all openingbooks for the engines are OFF
Thinkingtime for each engine: 6 min. for the whole game + 6 bonussec. pr. move
Hash Tables for each engine: 64 MB PC 3200 Ram (CAS Latency 2-2-2-5)
8 MB Cache for TB's
TB's: Nalimov - 3 und 4-pieces
Pondern OFF

Lately I have been testing Pro Deo 1.1. Here are the testresults for the dutch
engine:

Pro Deo 1.1 – Fruit 2.2.1 6-14
Pro Deo 1.1 – Hiarcs 10 4½-15½
Pro Deo 1.1 – Toga II 1.1a 5-15
Pro Deo 1.1 – Shredder 9 7-13
Pro Deo 1.1 – Fritz 9 5½-14½
Pro Deo 1.1 – Spike 1.19½-10½
Pro Deo 1.1 – SmarThink 1.00 12½-7½
Pro Deo 1.1 – Ktulu 7.5 10½-9
Pro Deo 1.1 – The King 3.33 3-17 (!!)
Pro Deo 1.1 – Naum 1.91 15-5 (!)
Pro Deo 1.1 – Gandalf 6.0 12-8
Pro Deo 1.1 – Loop List 600 9-11
Pro Deo 1.1 – Glaurung 1.0.1 9-11
Pro Deo 1.1 – Ruffian 2.1.0 14-6

Generally the results were quite different from very good (15-5 against Naum
1.91 and 12½-7½ against SmarThink 1.00) to very bad (3-17 against The King
3.33). The total score acheived by Pro Deo 1.1 was 43,8% which is a bit below my
expectations (45%). As a result of the latest testresults the ratinglist now
looks like this:

   Program                       Elo    +   -   Games   Score   Av.Op.  Draws
 1 Hiarcs 10                   : 2760   36  36   280    63.9 %   2660   27.1 %
 2 Fruit 2.2.1                 : 2758   35  35   280    63.7 %   2660   29.6 %
 3 Toga II 1.1a                : 2744   34  34   280    61.6 %   2661   33.9 %
 4 Fritz 9                     : 2724   35  35   280    58.8 %   2663   28.2 %
 5 Shredder 9                  : 2723   35  35   280    58.6 %   2663   28.6 %
 6 Spike 1.1                   : 2704   34  34   280    55.7 %   2664   32.9 %
 7 The King 3.33               : 2662   34  34   280    49.3 %   2667   31.4 %
 8 SmarThink 1.00              : 2655   34  34   280    48.2 %   2668   31.4 %
 9 Ktulu 7.5                   : 2652   36  36   280    47.7 %   2668   23.9 %
10 Pro Deo 1.1                 : 2626   35  36   280    43.8 %   2670   25.4 %
11 Loop List 600               : 2610   36  36   280    41.2 %   2671   23.2 %
12 Gandalf 6.0                 : 2610   35  35   280    41.2 %   2671   28.9 %
13 Naum 1.91                   : 2604   33  34   280    40.4 %   2671   33.6 %
14 Glaurung 1.0.1              : 2594   37  37   280    38.9 %   2672   22.1 %
15 Ruffian 2.1.0               : 2580   35  35   280    37.0 %   2673   30.4 %

All engines are playing with their default-settings with exception of Hiarcs 10
which is being tested with Hypermodern ON.

There are several changes in the ratinglist compared to the last one. The most
interesting is probably that Hiarcs 10 has replaced Fruit 2.2.1 as the topengine
in my list. Since the start of this ratinglist Fruit 2.2.1 has been leading but
all the time Hiarcs 10 has been close to Fruit. And finally Hiarcs 10 managed to
make the last step and overtake Fruit 2.2.1 if only by a small margin. However,
Hiarcs 10 probably won't be able to enjoy this position for a long time since
the next engine to be tested is Rybka 1.1!

As soon as Rybka 1.1 is available I will start the test and this time I will not
only post the final result but also a half-way report.

Best regards
Per



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.