TalkChess.com

I didn't say anything about whether something should or should not be allowed, so all that's a bit irrelevant; anyone is of course free to play matches in whatever conditions they'd like :) My point was that it is unsurprising that a 2300ish CCRL engine at fast time controls plays some bad opening l...

The existence of lines that Leela completely misplays is unsurprising at this point. It's maybe somewhere in the 2300s CCRL these days (on strong hardware), which, while impressive given the "zero" start just over a month ago and all the bugs they've dealt with, is still rather weak in the grand sch...

Just to be clear, the change to get more variety in openings (the temperature decay) is only for the match games used for gating (determining whether the new network passes or not); it is not used in self-play games. There's already a high variety of openings in self-play because a temperature of 1 ...

Thanks for the update and that position, Kai! The biggest problem in that position is just that black's mate (in the line 136.Kh4 g5+ 137.Kxh5 Qf3+ 138.Kh6 Qh1+ 139.Qh3 Qxh3#) has a series of only winning moves, and the net assigns all of them but the last a very low probability of being played (4%,...

Leela does not do playouts to the end of the game and then tabulate results, so the expected scores returned don't have to somehow be evenly divisible into the number of playouts/visits. Even if the MCTS only searches one line of play, and that line of play is only one ply deep, the NN will return a...

Probably that "winrate" (expected score) is the average at the root; I'd have to look at the code to be sure, but that (or something similar) is almost certainly the case. It knows that Qh4 is 100%: Using 2 thread(s). Generated 1924 moves Detecting residual layers...v1...64 channels...6 bloc...

Just to be clear, Leela does not REQUIRE different hardware than normal engines. It's just that as the network gets bigger, the performance penalty of running a CPU-only version will grow. For now, though, if you have a low-to-middle-end GPU, you might actually get better performance from the CPU-on...

Thanks for the update Kai! On the one hand, it's quite possible that a fundamental change to its MCTS implementation will be required at some point if it wants to compete at the highest level, and the work Daniel Shawul has done with Scorpio could prove quite useful in that case (well, it's fantasti...

Indeed. Leela's output is an expected score, not actually a win%. The wording on the site has since been changed, it seems, although now it includes this "50%=draw" bit in the legend, which caused some debate in the discord. Probably should just say "50%=equal chances" or some such thing ("expected ...

With the current smallish network, you'd actually need at least a couple really expensive GPUs to match the performance the cpu-only version would get on 43 CPU cores. So, if the requirement for fairness was something like cost of hardware, it would currently be better off playing on the 43 CPU core...