Arasan 20.0

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

Re: Arasan 20.0

Post by MikeB »

jdart wrote:Arasan 20.0 is available from http://www.arasanchess.org.

Changes in Arasan 20.0:
1) Fix to handling of UCI_LimitStrength and UCI_Elo options. These can now be processed in either order and will set the search strength correctly.
2) Revised tuning program. Only the "Texel" method is supported now. Tuner uses a "mean-squared error" objective and takes a labeled EPD file
as input. Various optimization methods can be selected. Optimization steps are scaled appropriately based on parameter ranges.
3) Changes to king cover and king safety scoring.
4) Bug fix: Ensure hash move is always checked for validity (was not being done for evasions).
5) Fix possible race condition updating root PV.
6) Add some utility programs and Python scripts to source package.
7) Considerable code cleanup and fixing warnings and possible bugs, notably in Syzygy tb code.

This version scores quite a bit better than 19.2 against a gauntlet of opponents in my fast time-control testing, but actually did not win a 2400 game blitz match directly against version 19.2. So I don't know quite what to make of that. Anyway I am releasing it, partly because of the bug fixes, and hopefully more testing will make clearer what the relative strength is.

--Jon
Looks pretty good to me:

Very unscientific match below, since I also let them use their own book and of course not very many games.

Code: Select all

32 rounds and 96 games completed...
time control: 1500+25
Date: 02/01/17 : 00:03:34
96 game(s) loaded
Rank Name            Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 Arasan 20.0      3171   0.0   43   43    64   42.0  65.6   30   10   24  46.9  37.5  3064 
   2 Crafty-25.2      3083  88.0   43   43    64   29.5  46.1   17   22   25  26.6  39.1  3108 
   3 EXchess v7.96b   3046  37.7   43   43    64   24.5  38.3   13   28   23  20.3  35.9  3127 
--------------------------------------------------------------------------------------------------------- 
User avatar
Werner
Posts: 2872
Joined: Wed Mar 08, 2006 10:09 pm
Location: Germany
Full name: Werner Schüle

Re: Arasan 20.0

Post by Werner »

Thanks Jon,
do you think syzygy bases work correct?
Werner

[d]4k3/8/8/8/8/8/4N3/4KB2 w - -

Engine: Arasan 20.0 x64 1CPU (256 MB)
von Jon Dart

19.10 0:01 +7.27 1.Sd4 Kf7 2.Lc4+ Kf6 3.Kd2 Ke7 4.Ke3 Kd6
5.Ke4 Kc5 6.Lg8 Kd6 7.Sf5+ Kc6
8.Ld5+ Kc5 9.Ke5 Kb4 10.Kd4 Ka3
11.Kc3 Ka4 12.Lc6+ Ka3 (4.927.580) 3391 TB:1

20.10 0:02 +7.27 1.Sd4 Kf7 2.Lc4+ Kf6 3.Kd2 Ke7 4.Ke3 Kf6
5.Ke4 Kg7 6.Ke5 Kh7 7.Kf6 Kh8 8.La6 Kh7
9.Sf5 Kg8 10.Kg6 Kh8 11.Kf7 (7.047.664) 3417 TB:1

21.10 0:02 +7.27 1.Sd4 Kf7 2.Lc4+ Kf6 3.Kd2 Ke7 4.Ke3 Kf6
5.Ke4 Kg7 6.Ke5 Kh7 7.Kf6 Kh8 8.La6 Kh7
9.Sf5 Kg8 10.Lc4+ Kh7 11.Se3 Kh8
12.Sd5 (9.670.842) 3439 TB:1

22.10 0:03 +7.33 1.Sd4 Kf7 2.Lc4+ Kf6 3.Kd2 Ke7 4.Ke3 Kf6
5.Ke4 Kg7 6.Ke5 Kh7 7.Kf6 Kh8 8.La6 Kh7
9.Kf7 Kh8 10.Sf3 Kh7 11.Sg5+ Kh6
12.Se4 Kh5 (13.762.353) 3467 TB:1

23.10 0:05 +7.33 1.Sd4 Kf7 2.Lc4+ Kf6 3.Kd2 Ke7 4.Ke3 Kd6
5.Ke4 Kc5 6.Lg8 Kb4 7.Kd3 Kc5 8.Sf5 Kb4
9.Kc2 Kc5 10.Kc3 Kc6 11.Kb4 Kc7
12.Kc5 Kd7 (19.245.330) 3469 TB:1

24.02 0:08 +7.35 1.Lg2 Kf8 2.Kd2 Kg7 3.Ke3 Kf6 4.Sf4 Kg7
5.Lh3 Kh6 6.Lg4 Kg5 7.Le6 Kh4 8.Ke4 Kg5
9.Ke5 Kh4 10.Sg6+ Kg5 11.Lf5 Kh6
12.Kf6 Kh7 13.Se5+ Kg8 14.Le6+ (31.122.890) 3476 TB:1

24.10 0:09 +7.35 1.Lg2 Kf8 2.Kd2 Kg7 3.Ke3 Kf6 4.Sf4 Kg7
5.Lh3 Kh6 6.Lg4 Kg5 7.Le6 Kh4 8.Ke4 Kg5
9.Ke5 Kh4 10.Sg6+ Kg5 11.Lf5 Kh6
12.Kf6 Kh7 13.Se5+ Kg8 14.Le6+ (31.868.206) 3480 TB:1

25.10 0:10 +7.35 1.Lg2 Kf8 2.Kd2 Kg7 3.Ke3 Kf6 4.Sf4 Kg7
5.Kd4 Kf6 6.Lh3 Kg7 7.Ke5 Kf7 8.Le6+ Kg7
9.Ld5 Kh6 10.Kf6 Kh7 11.Lc4 Kh8
12.Sd3 Kh7 13.Sc1 Kh8 (37.779.728) 3493 TB:1

26.10 0:14 +7.36 1.Lg2 Kf8 2.Kd2 Kg7 3.Ke3 Kf6 4.Sf4 Kg7
5.Kd4 Kf6 6.Lh3 Kg7 7.Ke5 Kf7 8.Le6+ Kg7
9.Ld5 Kh6 10.Kf6 Kh7 11.Lc4 Kh8
12.Sg6+ Kh7 13.Lf1 Kg8 14.La6 (52.324.439) 3528 TB:1

27.00 0:15 +7.17 1.Lg2 Kf8 2.Kd2 Kg7 3.Ke3 Kf6 4.Sf4 Kg7
5.Kd4 Kf6 6.Lh3 Kg7 7.Ke5 Kf7 8.Le6+ Kg7
9.Ld5 Kh6 10.Kf6 Kh7 11.Lc4 Kh8
12.Sg6+ Kh7 13.Lf1 Kg8 14.La6 (52.911.058) 3527 TB:1

27.00 0:15 +7.17 1.Lg2 Kf8 2.Kd2 Kg7 3.Ke3 Kf6 4.Sf4 Kg7
5.Kd4 Kf6 6.Lh3 Kg7 7.Ke5 Kf7 8.Le6+ Kg7
9.Ld5 Kh6 10.Kf6 Kh7 11.Lc4 Kh8
12.Sg6+ Kh7 13.Lf1 Kg8 14.La6 (52.911.058) 3527 TB:1

Bester Zug: Lf1-g2 Zeit: 0:15.015 min K/s: 3.527.403 Knoten: 52.911.058 TB: 1
jdart
Posts: 4367
Joined: Fri Mar 10, 2006 5:23 am
Location: http://www.arasanchess.org

Re: Arasan 20.0

Post by jdart »

The way I have implemented it, if there is a hit at the root, the engine is allowed to search but the search moves are restricted to the set that will preserve the best value found from the tablebases.

It may not play optimally (shortest distance to mate) because that is not how Syzygy works.

Arasan also has some internal code for this endgame so it can mate with B+N vs K even without tablebases.

--Jon
User avatar
Werner
Posts: 2872
Joined: Wed Mar 08, 2006 10:09 pm
Location: Germany
Full name: Werner Schüle

Re: Arasan 20.0

Post by Werner »

It will be not easy to catch version 19.2:

1 Nemorino 2.0 x64 1CPU +35 +37/=36/-27 55.00% 55.0/100
2 Arasan 19.2 x64 1CPU -35 +27/=36/-37 45.00% 45.0/100

1 Nemorino 2.0 x64 1CPU +31 +36/=37/-27 54.50% 54.5/100
2 Arasan 20.0 x64 1CPU -31 +27/=37/-36 45.50% 45.5/100

of course, we need a lot of more games...
Werner
jdart
Posts: 4367
Joined: Fri Mar 10, 2006 5:23 am
Location: http://www.arasanchess.org

Re: Arasan 20.0

Post by jdart »

Blitz gauntlet shows them pretty close:

Code: Select all

Rank Name          Elo    +    - games score oppo. draws
   1 Texel-1.05   2512   18   18   600   59%  2455   40%
   2 Arasan 19.2  2455   10   10  2400   60%  2386   36%
   3 Crafty-25.2  2387   19   19   600   40%  2455   37%
   4 Gaviota-1.0  2338   19   19   600   32%  2455   36%
   5 Toga-3.0     2308   20   20   600   29%  2455   30%

Rank Name          Elo    +    - games score oppo. draws
   1 Texel-1.05   2516   19   18   600   59%  2457   42%
   2 Arasan 20.0  2457   10   10  2400   61%  2386   37%
   3 Crafty-25.2  2375   19   19   600   37%  2457   38%
   4 Gaviota-1.0  2333   19   19   600   32%  2457   34%
   5 Toga-3.0     2318   19   20   600   29%  2457   36%

Adam Hair
Posts: 3226
Joined: Wed May 06, 2009 10:31 pm
Location: Fuquay-Varina, North Carolina

Re: Arasan 20.0

Post by Adam Hair »

So far, 20.0 is ahead of 19.2 @ 40/4 against a gauntlet of opponents (the opponents played are not identical)

CCRL 40/4 computed with Ordo:

Code: Select all

   # PLAYER                                :  RATING  POINTS  PLAYED   (%)
 225 Arasan 20.0 64-bit                    :  2993.1   830.0    1381    60
 241 Arasan 19.2 64-bit                    :  2981.0   561.0    1150    49
Meanwhile, self testing shows 19.2 ahead of 20.0 @40/4 at the moment:

Code: Select all

   # PLAYER         :  RATING  ERROR  POINTS  PLAYED   (%)
   1 Arasan 19.2    :     0.0   ----   367.5     697    53
   2 Arasan 20.0    :   -20.1   22.6   329.5     697    47
User avatar
Sylwy
Posts: 4468
Joined: Fri Apr 21, 2006 4:19 pm
Location: IASI - the historical capital of MOLDOVA
Full name: SilvianR

Re: My test

Post by Sylwy »