DESKTOP-CORSAIR, Blitz 0m+20s Ponder Off. Book Perfect 2019 to 8 moves.
Stockfish
4Gb HT
32 threads
Contempt = 0
Sergio 512x40-50
CS = 2000000
Policy Temp = 2.0
Live Stream:
Re: LIVE Stockfish 130120 vs Sergio 512x40-50
Posted: Mon Jan 13, 2020 3:46 pm
by Hai
At this moment there is a really nice game going on.
Stockfish as white said it's 0.00 and then realized in the next move it's -5.00.
I think 30 seconds per move would be a much better time control and 40x512 would play much stronger.
Re: LIVE Stockfish 130120 vs Sergio 512x40-50
Posted: Mon Jan 13, 2020 3:54 pm
by mwyoung
Hai wrote: ↑Mon Jan 13, 2020 3:46 pm
At this moment there is a really nice game going on.
Stockfish as white said it's 0.00 and then realized in the next move it's -5.00.
I think 30 seconds per move would be a much better time control and 40x512 would play much stronger.
We can try at 30s. But needs some games first at 20s so when have some data. But if stockfish stomps it at 20s I don't know what good 10s more would make. I would be inclined to try 60s. And This is only the first 512x40 and it should improve.
Re: LIVE Stockfish 130120 vs Sergio 512x40-50
Posted: Mon Jan 13, 2020 7:57 pm
by Alayan
Fixed time per move gets you much lower quality games for the same total time spent.
I guess the point is to guarantee the huge net always has enough nodes ?
Re: LIVE Stockfish 130120 vs Sergio 512x40-50
Posted: Mon Jan 13, 2020 8:23 pm
by mwyoung
Alayan wrote: ↑Mon Jan 13, 2020 7:57 pm
Fixed time per move gets you much lower quality games for the same total time spent.
I guess the point is to guarantee the huge net always has enough nodes ?
The TC is not fixed time. The engines can think longer or shorter as time permits and the position.