TalkChess.com

Just by making Leela available to Caruana, Ding and the other super-GMs fully for free instead of ad-hominem freemium coaching services, the devs have missed out on a 5 (maybe 6) figure amount. (Well, that was a consequence of Glaurung's GPL, but they could still have done the Fat Fritz cloud trick...

Thanks Ferdy for making your tool available! I like some of the plots you output for the results. I wanted to point out that Optuna uses tree-structured parzen estimators as their model, which does not model interactions between parameters. Tools based on Gaussian processes (like the chess-tuning-to...

If you are interested in a global optimizer specifically adapted to the application of chess, I recently refactored the tuning code we use for Lc0 into its own tuning library: https://github.com/kiudee/chess-tuning-tools Since I want to make it as easy to use as possible, I also started work on the ...

After reading the code very carefully, I noticed that the standard deviation shown in the result output results from passing the --simulations flag. Ordo is then replaying the games several times to obtain error bars on all quantities. So, in the case of white advantage and draw rate, the prior info...

How are you sure it is a bug not a feature :) Questions I have: a) Does Ordo even have the concept of priors/posterieors etc? You are talking as if it has it but AFAIk it is not based on bayesian approach. b) Does it take into consideration the color and draw ratio when calculating elos? a) Yes, Or...

Why are you using Ordo anyway, clearly it has inferior algorithms than bayeselo that is based on bayesian approach. Here https://www.remi-coulom.fr/Bayesian-Elo/ Remi remid discusses some of the advantages over prior alogrithm, EloStat. For the calculation of accurate standard deviations, there is ...

I was experimenting with using prior knowledge about white advantage and draw rate in Ordo, but noticed that the posterior standard deviations were much too low. Sometimes even two orders of magnitude lower than expected. To test this I created a very simple pgn file consisting of just 2 games. One ...

As a matter of fact, we already trained a Chess960 network to test the implementation. Take a look at our list of networks and download for example the network 712574. It supports both Chess960 and the new moves-left-head. If you want to see the moves-left-head in action, try these additional settin...

Yes, the current distributed tune in progress is the first one for longer time controls (with over 100k nodes / move). I will release the complete data as soon as it’s finished. Until then, the only thing I can recommend to experiment with is CPuctFactor. CPuct itself should stay low, since it has a...

Then again, I am surprised. A CPuct = 1.900 doesn't seem very intuitive to LTC, does it? How re-farctoring works with Lc0? Really, I feel I have to check and re-check every detail of my setup. Even if a self-play, an improvement of 70 Elo points with 70% draw rate at 400k npm is a bit hard to belie...