On-line engine blitz tourney April

[Rémi Coulom]

Thanks for organizing the tournament. CrazyZero is my generic AlphaZero implementation (I currently have networks for Go, shogi, gomoku, Othello, renju, and chess). I spent only a few days training a network. It is 20 layers of 128 units. It was running with two Volta GPU (V100 + TitanV). It has good positional evaluation, but big tactical blindspots. I hope it will have made progress for the next tournament.

[Joost Buijs]

LeelaRB - for a 7+ year old AMD Phenom II X6 and a sub £100 GTX 1050, that was a nice performance ! A bit of luck too for sure, but I'll take it

This month I used it without a book (which I think is probably best) and with a stronger net, JH.T6.532.

It is quite scary going up against 24 core Xeon, Threadripper, and an 180 core EC2 instance !

Indeed, it did a lot better than last month, it is amazing that it performs like this with a mere GTX1050. If you want to use a book it is probably best to use a book build with LC0 games, otherwise it will get into positions that it doesn't like.

The other engines are progressing very fast too. The version of the engine I played with this month performs in self-tests ~30 Elo higher as the one from last month. I had hoped for a somewhat better result, but with just 9 games it is like a tombola, one or two unlucky games and you can shake it. My hardware gets a bit outdated too, all the engines at the top (besides LC0 of course) were running at 16 cores or more, at least that is what I think when looking at their n/s. This summer both Intel and AMD will announce new processors (Cascade Lake and Zen 2), then I will decide on building a more competitive system.

[George Tsavdaris]

Thanks for organizing the tournament. CrazyZero is my generic AlphaZero implementation (I currently have networks for Go, shogi, gomoku, Othello, renju, and chess). I spent only a few days training a network. It is 20 layers of 128 units. It was running with two Volta GPU (V100 + TitanV). It has good positional evaluation, but big tactical blindspots. I hope it will have made progress for the next tournament.

Othello(Reversi)? Nice, how does it do against the big guys, e.g NTest, Edax etc?
Also are you gonna release any of them?
And will the whole framework be open sourced? I would be interested in a Capablanca Random Chess NN network so I keep dreaming.

[Rémi Coulom]

Othello(Reversi)? Nice, how does it do against the big guys, e.g NTest, Edax etc?
Also are you gonna release any of them?
And will the whole framework be open sourced? I would be interested in a Capablanca Random Chess NN network so I keep dreaming.

I did Othello very rapidly. It is probably not competitive at all with the strong engines.

I won't open-source my code. I licence it to my publisher (Unbalance). They will make cell-phone and console apps.