Search found 1635 matches

by yanquis1972
Fri Dec 21, 2018 7:10 am
Forum: Computer Chess Club: General Topics
Topic: Alphazero news
Replies: 381
Views: 98343

Re: Alphazero news

theres a link and summary here: https://sites.google.com/site/computers ... t2017books

It’s meant for engine-engine matches, seems to be a good compromise of quality and diversity.
by yanquis1972
Thu Dec 20, 2018 8:36 pm
Forum: Computer Chess Club: General Topics
Topic: Smallnet (128x10) run1 progresses remarkably well
Replies: 24
Views: 5854

Re: Smallnet (128x10) run1 progresses remarkably well

But kais main point, I think, was that experimental runs should be as efficient as possible. Could eg tests20+30, and again test40 (if the aim is truly experimental as the name implies) be performed entirely on 128x10 (or smaller) nets, or does it not translate? test30 is an experimental run, and s...
by yanquis1972
Thu Dec 20, 2018 1:15 am
Forum: Computer Chess Club: General Topics
Topic: Smallnet (128x10) run1 progresses remarkably well
Replies: 24
Views: 5854

Re: Smallnet (128x10) run1 progresses remarkably well

But kais main point, I think, was that experimental runs should be as efficient as possible. Could eg tests20+30, and again test40 (if the aim is truly experimental as the name implies) be performed entirely on 128x10 (or smaller) nets, or does it not translate?
by yanquis1972
Thu Dec 20, 2018 12:06 am
Forum: Computer Chess Club: General Topics
Topic: Smallnet (128x10) run1 progresses remarkably well
Replies: 24
Views: 5854

Re: Smallnet (128x10) run1 progresses remarkably well

I’ve assumed the bignet has extended its schedule markedly after publication of several A0 parameters that have been implemented into the training. It’s also only had one LR drop and provided they are aiming for SF10+ strength with it I don’t see the issue with expense.
by yanquis1972
Wed Dec 19, 2018 8:01 pm
Forum: Computer Chess Club: General Topics
Topic: Stockfish 10 Strikes Back
Replies: 6
Views: 1428

Re: Stockfish 10 Strikes Back

nice test, the other thing that stands out to me is test30s results vs test10. it's really remarkably close in strength if it's LR can be taken at more or less face value (only one drop so far). SF has reigned supreme in KIDs among AB engines for years but to completely flip the results would seem t...
by yanquis1972
Wed Dec 19, 2018 5:22 pm
Forum: Computer Chess Club: General Topics
Topic: Alphazero news
Replies: 381
Views: 98343

Re: Alphazero news

after 93 games at 5+5, test30 is stable at -70 elo vs SF10 & -21 elo vs SF10+book (perfect2017/2018). will attach pgn's later, going to try to get a decent sample of SF10 vs SF10+book results.
by yanquis1972
Tue Dec 18, 2018 10:09 pm
Forum: Computer Chess Club: General Topics
Topic: Alphazero news
Replies: 381
Views: 98343

Re: Alphazero news

I think the error margin for only 50 games is too large. I plan to run more, but it was pretty consistent, and has been across the board. So far SF10 + test10 (11248) are both outperforming SF10+book, albeit in small samples. (3W 16D 1L for SF10, 2W 3D 0L for id11248 — four draws and a loss against...
by yanquis1972
Tue Dec 18, 2018 5:03 pm
Forum: Computer Chess Club: General Topics
Topic: Alphazero news
Replies: 381
Views: 98343

Re: Alphazero news

But as far as I can tell there’s little reason to conclude variety is the issue. Again, at longer TC, with a far lower leela ratio, test10 is performing fine against SF with what I understand to be a high quality book (albeit one that’s likely weaker than SF10 by itself). You can see in the pgns I l...
by yanquis1972
Tue Dec 18, 2018 11:08 am
Forum: Computer Chess Club: General Topics
Topic: Alphazero news
Replies: 381
Views: 98343

Re: Alphazero news

So, again, the conclusion would be that the variety of openings hurts Lc0. Makes sense. Variety relative to what? There’s a w/l/d breakdown for A0s games against SF8 in each of the 12 openings (for each color); while it’s clearly a monster with white in many of them it scored significantly worse in...
by yanquis1972
Tue Dec 18, 2018 10:52 am
Forum: Computer Chess Club: General Topics
Topic: Alphazero news
Replies: 381
Views: 98343

Re: Alphazero news

I think the error margin for only 50 games is too large. I plan to run more, but it was pretty consistent, and has been across the board. So far SF10 + test10 (11248) are both outperforming SF10+book, albeit in small samples. (3W 16D 1L for SF10, 2W 3D 0L for id11248 — four draws and a loss against...