Binaries

[Carbec]

Hello,

I am looking for binaries to test my engine. I could find some directly on github, but not all.
Some home sites are even dead. So, is there somewhere a repositery with binaries that testers use ?

Thanks
Philippe

[jabolcni]

Hi,

I just use the binaries that are available on github. Usually you can find exe binaries for at least 6 engines which are around testing elo +/- 50 points.

[Carbec]

Yes, I do the same. And I find also a lot of engines, sometimes older versions of these engines, as mine is not so strong.
But I wonder where are the others used in the matches I see here.
The important thing is that I can make my own matches for testing.
I notices something strange, some engines have a score very different of their ccrl elo. Is it something natural ?

Thanks

Philippe

[Gabor Szots]

Hello,

I am looking for binaries to test my engine. I could find some directly on github, but not all.
Some home sites are even dead. So, is there somewhere a repositery with binaries that testers use ?

Thanks
Philippe

Hi Philippe,

These links may help you:
http://computer-chess.org/doku.php?id=c ... ngine_list
http://computer-chess.org/doku.php?id=c ... nload_list

[Gabor Szots]

I notices something strange, some engines have a score very different of their ccrl elo. Is it something natural ?

Which are they?

[Carbec]

Hi,

Thanks for the links.
To answer your question, this is 2 matches I did recently with cutechess-cli:
I added the engine's elo.
> GOOB should be higher.
> Weiss scores very well.

Rank Name Elo +/- Games Score Draw
0 Zangdar 2.26.07.101 24 7 6500 53.4% 31.4%
1 Weiss 1.2 3062 35 26 464 55.0% 34.9%
2 Princhess 0.15.1 3075 29 29 465 54.2% 17.2%
3 Drofa 3.2.0 3098 28 26 465 54.0% 34.8%
4 Pedantic 0.6.0 3077 25 25 464 53.6% 35.6%
5 Wahoo v4.0.0 3080 17 25 466 52.5% 38.0%
6 Polaris 1.8.1 3052 15 26 464 52.2% 31.0%
7 Ethereal 9.30 3055 14 25 466 52.0% 39.3%
8 4ku 4.0 3032 -39 25 464 44.4% 35.8%
9 Peacekeeper v1.60 3044 -44 27 464 43.6% 27.8%
10 PeSTO 2.210 BMI 2993 -57 27 463 41.9% 29.4%
11 Halogen 8.1 3014 -60 28 463 41.5% 25.9%
12 Koivisto 64 3.0 3027 -71 27 464 39.9% 30.6%
13 GOOB 1.8.9 3072 -85 26 464 38.0% 33.8%
14 Vengeance 3.0.0 2999 -148 29 464 29.8% 24.8%

Rank Name Elo +/- Games Score Draw
0 Zangdar 2.26.07.101 -14 5 12000 47.9% 33.7%
1 Berserk 4.1.0 3140 88 20 801 62.4% 33.1%
2 Senpai 2.0 3120 74 19 800 60.5% 37.3%
3 Koivisto 64 4.0 3140 45 19 801 56.5% 38.3%
4 Stash v28.0 3089 41 19 800 55.8% 37.9%
5 Ethereal 9.30 3055 35 19 800 55.0% 36.8%
6 Vajolet2 2.6 3098 35 20 800 55.0% 34.5%
7 Wahoo v4.0.0 3080 31 20 800 54.4% 34.4%
8 Drofa 3.2.0 3098 23 19 800 53.4% 37.3%
9 Pedantic 0.6.0 3077 21 19 800 53.1% 34.9%
10 Weiss 1.2 3061 18 19 800 52.6% 34.8%
11 Princhess 0.15.1 3075 16 22 800 52.3% 17.9%
12 Polaris 1.8.1 3052 -19 20 799 47.2% 34.2%
13 4ku 4.0 3032 -37 19 799 44.7% 38.9%
14 Peacekeeper v1.60 3044 -64 20 800 40.9% 30.5%
15 Protej 0.6.5 3074 -90 21 800 37.3% 25.4%

Here, Protej is the last one;
I am not a statistician, and don't have an opinion why. Its perhaps the number of games that is important.

Philippe

[Gabor Szots]

I am not a statistician, and don't have an opinion why. Its perhaps the number of games that is important.

I am not one either. But to use only one opponent against each of the engines may cause distortions. That's why we test each engine against a wide selection of engines. Also, we try to select opponents so that their average strength be close to that of the actual engine tested.

I can give no better explanation.