Search found 294 matches

by Alayan
Mon Jul 13, 2020 10:15 pm
Forum: Computer Chess Club: General Topics
Topic: Ethereal & Schooner seems to be "Big Monsters"
Replies: 32
Views: 1016

Re: Ethereal & Schooner seems to be "Big Monsters"

Ethereal "fast wins" results are much less impressive now after 6 rounds than they were after 2, aren't they ? But I'm glad this outlier 11 fast wins out of 80 games made you look at Ethereal games and enjoy them. Even though during development, strength is the main focus, it's rewarding to have the...
by Alayan
Sun Jul 12, 2020 1:16 pm
Forum: Computer Chess Club: General Topics
Topic: Cfish is back!
Replies: 43
Views: 1984

Re: Cfish is back!

When comparing two versions a few patches away, you'll have plenty of positions one version "gets" and the other doesn't. Minor seemingly unrelated changes can end up having a big impact on specific positions. But chess strength is about finding good moves in general positions, not in one specific t...
by Alayan
Sat Jul 11, 2020 5:39 pm
Forum: Computer Chess Club: General Topics
Topic: StockFiNN Release and Scaling
Replies: 109
Views: 6223

Re: StockFiNN Release and Scaling

Josh isn't doing anything illegal and he's abiding by the software licenses while producing something valuable. But my personal policy is that I would never pay a cent for any Stockfish derivative. Most of the work come from the original SF devs, plus here significant work by the japanese NNUE autho...
by Alayan
Thu Jul 09, 2020 2:59 pm
Forum: Computer Chess Club: General Topics
Topic: More than 2 years of Leela trolling is over! Introduction of MLH and major changes in version 26 engine!
Replies: 9
Views: 941

Re: More than 2 years of Leela trolling is over! Introduction of MLH and major changes in version 26 engine!

The issue with S18 openings is that more than a quarter of them are busted openings. This is incorrect. Double 1-0 doesn't mean busted opening. An opening is busted if it can't be defended, not if both engines fail to defend it. There were two clearly busted opening lines : the Najdorf, and the Lat...
by Alayan
Thu Jul 09, 2020 1:58 am
Forum: Computer Chess Club: General Topics
Topic: Stockfish sewing machine
Replies: 28
Views: 967

Re: Stockfish sewing machine

If there is a single cause for an easily reproducible defect, binary search allows to track down the responsible commit extremely quickly.

However if the defect is "random" and unlikely to happen, or if it has no clear single cause, then it's hard to impossible to fix.
by Alayan
Wed Jul 08, 2020 7:07 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Stockfish has included WDL stats in engine output
Replies: 43
Views: 1467

Re: Stockfish has included WDL stats in engine output

The most obvious one is that one needs a mechanism to introduce variety and this may influence the result. Another difficulty is that it requires an unreasonable amount of ressources to measure wdl probabilities in this way. Running the engine with multiple threads should introduce suitable variety...
by Alayan
Tue Jul 07, 2020 2:03 am
Forum: Computer Chess Club: General Topics
Topic: How did this happen? SF says +154 and draw.
Replies: 22
Views: 654

Re: How did this happen? SF says +154 and draw.

Probably Graph History Interaction with the hash table unaware of the 50 move rule.
by Alayan
Mon Jul 06, 2020 4:31 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Stockfish has included WDL stats in engine output
Replies: 43
Views: 1467

Re: Stockfish has included WDL stats in engine output

That was not my point however. My point is that it is simple to convert an existing alpha beta engine to work with win probabilities (or should I say win or draw probabilities to satisfy you). Using my way of probabilities in alpha beta has many obvious advantages as I mentioned in my previous post...
by Alayan
Mon Jul 06, 2020 1:08 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Throwing out draws to calculate Elo
Replies: 214
Views: 2729

Re: Throwing out draws to calculate Elo

LoS should not be used in patch testing for engines because the uniform prior is wrong. A correct prior is a distribution curve that says "your patch most likely lose some elo, if it gains any, it's most likely a small value". The prior tells us that if your test shows +10 elo and 99% LoS, this perf...