TalkChess.com

This is an incredibly cool project! Could get close to a soft-solve. Although engines prune so much, there are still positions they make serious mistakes, but they are fewer and fewer over time. Congrats, it's really well put together.

Input of a training example consists of position value plus probability for each legal move And how do you know the probability for each legal move from a PGN? You don't. In the normal training of AlphaZero and LC0, your training samples are written at the end of a search, and the visit count of ea...

That self play graph :shock: Almost total fiction. They even have some inversions on long spans of self-Elo versus real-Elo. Nonetheless, the progress overall is real, on very long spans. Since I am following it, 1.5 days, the self-Elo improved by 650 (!!!) Elo points, but against SF8 (real-Elo) by...

Are you telling me that this is not a problem for L0 or A0, and that it can be solved with bigger net and more training !? The purpose of the experiment it to find out... We are only at the beginning. Seriously. I think you should give LC0 some time. I know you have shown that MCTS is a disaster in...

Are you telling me that this is not a problem for L0 or A0, and that it can be solved with bigger net and more training !? The purpose of the experiment it to find out... We are only at the beginning. Seriously. I think you should give LC0 some time. I know you have shown that MCTS is a disaster in...

LC0 plays quite reasonable chess using only its evaluation I was talking about a minimax searcher, as standard Stockfish is. But even LC0 doesn't just evaluate a non-quiescent position and use that eval to choose moves. My understanding is it does a form of Monte Carlo tree search and that does not...

The strategy you advocate is fundamentally flawed. The point is that different network topologies might have wildly different performance. In such a situation it pays to try things in parallel. Suppose there are two methods, A requiring effort 1, and B requiring effort 3, but I don't know which is ...

But using the GPU would be not fair for the opponent engines, which are running on CPU only, of course. Lol. What is so funny about the fact, that a valid testing and rating means, each engine should have access on the same resources?!? LC Zero can use the GPU for its learning and improving. But no...

But using the GPU would be not fair for the opponent engines, which are running on CPU only, of course. Lol. What is so funny about the fact, that a valid testing and rating means, each engine should have access on the same resources?!? LC Zero can use the GPU for its learning and improving. But no...

Thanks to all contributors! Huge effort by Error323 and KillerDucky especially. There will be a slight drop in game speed on GPUs, so you should run a full tune with the new network. CPUs take more of a hit unfortunately. But the new network should have a pretty high ceiling hopefully :). 10x128 mea...