TalkChess.com

Nice, though I think several of those titles go to Claude Shannon.

maxdeg wrote: ↑

Mon Jul 15, 2019 1:51 pm

Waiting for Bad Byoy

feral, aggressive ……

No boy’s, just gyals.

Even against the new LD3, it’s not all one way traffic. [Event "T15072019 3"] [Site "Germany, isle of Rügen"] [Date "2019.07.15"] [Round "3"] [White "Lc0 Little Demon 3 173k"] [Black "Lc0 Bad Gyal 4a"] [Result "0-1"] [BlackElo "2200"] [ECO "A10"] [Opening "Englisch"] [Time "18:20:27"] [Variation "1....

Ferdy wrote: ↑

Mon Jul 15, 2019 12:58 am

Tried it against UCI_Elo 1500 engines at TC 40/2min, it is too strong at this level.

You’ll have to reduce to 100-300 nodes per move.

Bad Gyal 4a

The endgame data was problematic. This is a rerun without that data at 128x10. This should be the strongest net tactically, including T40, LD2.

ScorpioNN is good and might make it to DivP. The weakest part is probably the NN. It needed another 10-20 million self-play games to catch up to t30, but I wasn’t willing to shell out on vast.ai resources to get there. So until someone cooks up a better net, it’s likely to be a bit weaker than lc0/A...

I'm producing 128x10 nets for now to avoid the BLAS issue in the release version of lc0.

https://github.com/dkappe/leela-chess-w ... i/Bad-Gyal

Based on your work with Lc0, what would have to be done to make Lc0 play handicap chess well, other than revert to an old network like 11248? Can you explain why 11248 could beat GM Naroditsky at piece odds, while the new networks just blunder left and right when trying to give knight or more odds?...

nobody cares how many shots of vodka Mariya Muzychuk would need to drink to give me an even chess game. That I would pay to see. Yes, the chess engines are very good, but there’s also an element to these contests that’s like having an axe murderer spellcheck your emails. One little slip up and it’s...

I thought 64x6 would un on regular lc0, but it looks like you’ll need the development version. I’ll release 128x10 from now on.