Search found 160 matches

by nionita
Wed Apr 26, 2017 7:19 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: How to simulate a game outcome given Elo difference?
Replies: 6
Views: 1338

Re: How to simulate a game outcome given Elo difference?

I think the correct approach is to determine C statistically and solve the equation for w (then d and l) given e, which comes from Elo difference. Many thanks, I needed it to test an algorithm for parameter optimization, and before I try long chess matches I wanted to be sure it works well on some k...
by nionita
Tue Apr 25, 2017 4:21 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: How to simulate a game outcome given Elo difference?
Replies: 6
Views: 1338

Re: How to simulate a game outcome given Elo difference?

Can I do it in 2 steps, like considering draw probability constant, and then using Elo formula to "predict" the result, then take that as a probability for the stronger engine to win? The problem with this approach is that the draw probability also depends on the Elo difference, at least when this ...
by nionita
Tue Apr 25, 2017 12:29 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: How to simulate a game outcome given Elo difference?
Replies: 6
Views: 1338

How to simulate a game outcome given Elo difference?

Can I do it in 2 steps, like considering draw probability constant, and then using Elo formula to "predict" the result, then take that as a probability for the stronger engine to win? The problem with this approach is that the draw probability also depends on the Elo difference, at least when this i...
by nionita
Sat Apr 08, 2017 12:37 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: OICS new monthly tournament
Replies: 14
Views: 3370

Re: OICS new monthly tournament

My point is, anyone can request and run. If you cannot but want to be included, I do have hardware to spare and will run at you behalf @ your config if you can point me to what you want. All my cluster is Linux. Admit it's a lil outdated, but I have a lot ofwhat was nice like 2-3x years ago -Josh H...
by nionita
Sun Mar 19, 2017 9:16 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: OICS Reborn?
Replies: 8
Views: 1283

Re: OICS Reborn?

I think HGM's one is so successful because it is a shorter one, at a defined (announced) time. 24/7 would be probably most of the time not visited. But to speak for myself, I think another monthly tournament would be welcome (so we have 2 per month).
by nionita
Mon Mar 13, 2017 10:33 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: improved evaluation function
Replies: 5
Views: 2695

Re: improved evaluation function

Said that, my training data is far from perfect, and if you have working suggestions please let me know. This was not meant as a critique of your data, just a remark. When I collected a few million quiet positions from QS and tried SGD on them to improve the eval in Barbarossa I got exclusively ELO...
by nionita
Sun Mar 12, 2017 7:59 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: improved evaluation function
Replies: 5
Views: 2695

Re: improved evaluation function

After a bit of cleanup and speed up the Elo drop is only ~30 Elo. I also published this post here https://medium.com/@brtzsnr after cleaning it up a bit. Thanks for the insight. Are you doing the network operations in int(64) or float? There is a point with testing (or even training) only with quie...
by nionita
Sun Jan 29, 2017 6:40 pm
Forum: Computer Chess Club: General Topics
Topic: Chess Engines with problems on different hard-/software
Replies: 12
Views: 1736

Re: Chess Engines with problems on different hard-/software

cdani wrote:Barbarossa compiled in a windows 7:

www.andscacs.com/barbarossa/barbarossa0.4.0.zip
Thanks Daniel!

Norbert, Graham, Ruxy: can you check if this Win 7 version works at normal speed? I suppose you all have Win 7 installed.

Thanks,
Nicu
by nionita
Mon Jan 23, 2017 11:47 am
Forum: Computer Chess Club: General Topics
Topic: Barbarossa-0.4.0 release
Replies: 9
Views: 1977

Re: Barbarossa-0.4.0 - I stopped my test !!!

The windows compiles have problems! I did it under Win 10, and could not test under Win 7 until now.

Take a look at CCRL 40/4, which runs under linux. The new version is ~120 Elo better.

I stil had not possibility to see what is wrong with the windows compiles until now.