Discussion of anything and everything relating to chess playing software and machines.
Moderator: Ras
-
Jouni
- Posts: 3772
- Joined: Wed Mar 08, 2006 8:15 pm
- Full name: Jouni Uski
Post
by Jouni »
new net in SF framework:
Total: 1014840 W: 274645 L: 272456 D: 467739
+0,67 ELO

Jouni
-
Ajedrecista
- Posts: 2166
- Joined: Wed Jul 13, 2011 9:04 pm
- Location: Madrid, Spain.
Post
by Ajedrecista »
Hello Jouni:
Jouni wrote: ↑Mon Apr 24, 2023 5:31 pm
new net in SF framework:
Total: 1014840 W: 274645 L: 272456 D: 467739
+0,67 ELO
Gains are hard to find nowadays. Just to give a link for when this topic will not be recent anymore:
https://tests.stockfishchess.org/tests/ ... 0e719d2042
More than a million games (TC per side: 60" + 0.6"/move) in less than six days!
Code: Select all
Start time: 2023-04-18 17:11:11
Last updated: 2023-04-24 13:44:36
Elapsed time: 506005 seconds (5 days, 20 hours, 33 minutes and 25 seconds).
Games/second: 1014840/506005 ~ 2.0056 games/second.
Regards from Spain.
Ajedrecista.
-
ImNotStockfish
- Posts: 56
- Joined: Tue Sep 14, 2021 12:29 am
- Full name: .
Post
by ImNotStockfish »
100 games was
never enough
Right after that million-game-net passed, a new one was tested against that one and it passed in under 25k games
https://tests.stockfishchess.org/tests/ ... 6529b3dc46
Code: Select all
LLR: 2.95 (100%) <0.50,2.50>
Total: 24544 W: 6817 L: 6542 D: 11185
Ptnml(0-2): 8, 2252, 7488, 2505, 19
-
smatovic
- Posts: 3480
- Joined: Wed Mar 10, 2010 10:18 pm
- Location: Hamburg, Germany
- Full name: Srdja Matovic
Post
by smatovic »
ImNotStockfish wrote: ↑Tue Apr 25, 2023 9:54 am
100 games was
never enough
[...]
I am pretty sure this was a pun to some derivatives/clones which run a 100 game match to prove ....
IIRC Bob mentioned he ran 10K games to test 10 Elo back then with Crafty, 1M for <1 is quite a effort.
--
Srdja
-
Branko Radovanovic
- Posts: 89
- Joined: Sat Sep 13, 2014 4:12 pm
- Location: Zagreb, Croatia
- Full name: Branko Radovanović
Post
by Branko Radovanovic »
smatovic wrote: ↑Tue Apr 25, 2023 10:31 am
ImNotStockfish wrote: ↑Tue Apr 25, 2023 9:54 am
100 games was
never enough
[...]
I am pretty sure this was a pun to some derivatives/clones which run a 100 game match to prove ....
I thought it referred to the TCEC SuFi.

-
syzygy
- Posts: 5824
- Joined: Tue Feb 28, 2012 11:56 pm
Post
by syzygy »
smatovic wrote: ↑Tue Apr 25, 2023 10:31 am
ImNotStockfish wrote: ↑Tue Apr 25, 2023 9:54 am
100 games was
never enough
[...]
I am pretty sure this was a pun to some derivatives/clones which run a 100 game match to prove ....
You might be too kind.
IIRC Bob mentioned he ran 10K games to test 10 Elo back then with Crafty, 1M for <1 is quite a effort.
DTS was "proven" to be superior by running tests on no less than 24 positions.
-
smatovic
- Posts: 3480
- Joined: Wed Mar 10, 2010 10:18 pm
- Location: Hamburg, Germany
- Full name: Srdja Matovic
Post
by smatovic »
syzygy wrote: ↑Wed Apr 26, 2023 12:30 am
[...]
IIRC Bob mentioned he ran 10K games to test 10 Elo back then with Crafty, 1M for <1 is quite a effort.
DTS was "proven" to be superior by running tests on no less than 24 positions.
Good old days
--
Srdja
-
Jouni
- Posts: 3772
- Joined: Wed Mar 08, 2006 8:15 pm
- Full name: Jouni Uski
Post
by Jouni »
I remember, when SSDF published ratings after 100 games. And they are still available
https://ssdf.bosjo.net/long.txt
424 Fidelity Prestige 6502 4 MHz 1568 63 -66 118 44% 1609

Jouni