Page 1 of 1

FGRL 10 min + 6 sec Rating list - Komodo 13.01 MCTS

Posted: Tue May 21, 2019 2:53 pm
by fastgm
FGRL Rating list - 10 minutes + 6 seconds

Komodo 13.01 MCTS (+53 to Komodo 12.3 MCTS)

http://www.fastgm.de

Re: FGRL 10 min + 6 sec Rating list - Komodo 13.01 MCTS

Posted: Tue May 21, 2019 6:04 pm
by lkaufman
fastgm wrote: Tue May 21, 2019 2:53 pm FGRL Rating list - 10 minutes + 6 seconds

Komodo 13.01 MCTS (+53 to Komodo 12.3 MCTS)

http://www.fastgm.de
So just as with CEGT 40/20, the gain is far more than the 30 elo we estimated for blitz, so the improved scaling worked. Now only 70 elo between the two modes of Komodo 13 on your list. I imagine the gap will be even narrower if you have the time to run K mcts on your one hour list.

Re: FGRL 10 min + 6 sec Rating list - Komodo 13.01 MCTS

Posted: Tue May 21, 2019 8:08 pm
by Modern Times
lkaufman wrote: Tue May 21, 2019 6:04 pm
So just as with CEGT 40/20, the gain is far more than the 30 elo we estimated for blitz, so the improved scaling worked. Now only 70 elo between the two modes of Komodo 13 on your list. I imagine the gap will be even narrower if you have the time to run K mcts on your one hour list.
We have +34 at blitz which is probably what you were expecting.

Re: FGRL 10 min + 6 sec Rating list - Komodo 13.01 MCTS

Posted: Thu May 23, 2019 6:38 am
by Gary Internet
fastgm wrote: Tue May 21, 2019 2:53 pm FGRL Rating list - 10 minutes + 6 seconds

Komodo 13.01 MCTS (+53 to Komodo 12.3 MCTS)
On the FastGM 60+0.6 list it only gained 8 Elo over Komdo 12.3 MCTS. When I saw that, I scoffed at MCTS and began to wonder what the fuss was about. But seeing this gain of 53 Elo completely changed my mind. It's very impressive. With 10 times the amount of time you get 6.625 times the Elo gain.

I don't know what the exact release date of 13.01 was, but I remember that 12.3 was released on 17 December 2018. If 13.01 was released on 17 May 2019, that's 5 months, giving an average of 10.6 Elo per month. If you keep that up, the 70 Elo gap between MCTS and regular Komodo will be all but closed by the end of the year. :)

Re: FGRL 10 min + 6 sec Rating list - Komodo 13.01 MCTS

Posted: Thu May 23, 2019 9:52 am
by JJJ
Gary Internet wrote: Thu May 23, 2019 6:38 am
fastgm wrote: Tue May 21, 2019 2:53 pm FGRL Rating list - 10 minutes + 6 seconds

Komodo 13.01 MCTS (+53 to Komodo 12.3 MCTS)
On the FastGM 60+0.6 list it only gained 8 Elo over Komdo 12.3 MCTS. When I saw that, I scoffed at MCTS and began to wonder what the fuss was about. But seeing this gain of 53 Elo completely changed my mind. It's very impressive. With 10 times the amount of time you get 6.625 times the Elo gain.

I don't know what the exact release date of 13.01 was, but I remember that 12.3 was released on 17 December 2018. If 13.01 was released on 17 May 2019, that's 5 months, giving an average of 10.6 Elo per month. If you keep that up, the 70 Elo gap between MCTS and regular Komodo will be all but closed by the end of the year. :)
And how many time until it become better than Stockfish 10 ? Then Stockfish dev ?

Re: FGRL 10 min + 6 sec Rating list - Komodo 13.01 MCTS

Posted: Sat May 25, 2019 3:18 pm
by mjlef
JJJ wrote: Thu May 23, 2019 9:52 am
Gary Internet wrote: Thu May 23, 2019 6:38 am
fastgm wrote: Tue May 21, 2019 2:53 pm FGRL Rating list - 10 minutes + 6 seconds

Komodo 13.01 MCTS (+53 to Komodo 12.3 MCTS)
On the FastGM 60+0.6 list it only gained 8 Elo over Komdo 12.3 MCTS. When I saw that, I scoffed at MCTS and began to wonder what the fuss was about. But seeing this gain of 53 Elo completely changed my mind. It's very impressive. With 10 times the amount of time you get 6.625 times the Elo gain.

I don't know what the exact release date of 13.01 was, but I remember that 12.3 was released on 17 December 2018. If 13.01 was released on 17 May 2019, that's 5 months, giving an average of 10.6 Elo per month. If you keep that up, the 70 Elo gap between MCTS and regular Komodo will be all but closed by the end of the year. :)
And how many time until it become better than Stockfish 10 ? Then Stockfish dev ?
Anyone who has invested in mutual funds knows the phrase "past performance is not an indicator of future outcomes". But we are hopeful.