No +100 Elo for SF on Windows for me

[Dann Corbit]

Since there are no windows builds, I made one using the current code.
The NUMA stuff was compiled in, since I had the NUMA option for NUMA policy, which was set to auto in the Arena GUI.
I ran 2000 games between the last released Windows version which has no NUMA and the current source code base.
The result was +10 Elo +/- 5 Elo.
They did mention that Windows 10 has "partial" implementation of NUMA.

[Albert Silver]

Since there are no windows builds, I made one using the current code.
The NUMA stuff was compiled in, since I had the NUMA option for NUMA policy, which was set to auto in the Arena GUI.
I ran 2000 games between the last released Windows version which has no NUMA and the current source code base.
The result was +10 Elo +/- 5 Elo.
They did mention that Windows 10 has "partial" implementation of NUMA.

It is worth mentioning that the 104 Elo calculation is based on 278 games, not tens of thousands.

Total: 278 W: 110 L: 29 D: 139 Elo +104.25

So it may not be exact...

[Dann Corbit]

Reminds me of when I came up with "smooth scaling stockfish"
After a few hundred games I had +150 Elo.
After a couple thousand it was +50 Elo, IIRC.

[Viz]

This is really hardware dependent, can we whatever number, just not less than 0.

[Uri Blass]

Since there are no windows builds, I made one using the current code.
The NUMA stuff was compiled in, since I had the NUMA option for NUMA policy, which was set to auto in the Arena GUI.
I ran 2000 games between the last released Windows version which has no NUMA and the current source code base.
The result was +10 Elo +/- 5 Elo.
They did mention that Windows 10 has "partial" implementation of NUMA.

It is worth mentioning that the 104 Elo calculation is based on 278 games, not tens of thousands.

Total: 278 W: 110 L: 29 D: 139 Elo +104.25

So it may not be exact...

I do not understand what you are talking about.

I never read about the 104 elo calculation based on 278 games and I do not know what is the source for it.

I only read a thread with the title no +100 elo for SF when I never expected +100 elo for stockfish.
I think it is better to explain first what is the reason to expect this type of result.

If the claim is that people got
Total: 278 W: 110 L: 29 D: 139 Elo +104.25 then it is better to give a link to the result and not only to give numbers here when I need to search for the source.

Edit:I now read the second thread that gives this link

https://tests.stockfishchess.org/tests/ ... d5e27db0d8

[Uri Blass]

I see that they use 256 threads at 60+1 and 16000 mbytes hash

I think nobody here has the relevant hardware to try to reproduce the results and I wonder how many nodes per move they searched.

[Jouni]

I was wondering, if this NUMA patch needs cluster SF to run!?

[Dann Corbit]

I did another build with today's code, and the Elo gain is 21:

Score of stockfish-x64-avx2 vs stockfish_24052816_x64_avx2: 214 - 94 - 1692 [0.530]
... stockfish-x64-avx2 playing White: 203 - 30 - 767 [0.587] 1000
... stockfish-x64-avx2 playing Black: 11 - 64 - 925 [0.473] 1000
... White vs Black: 267 - 41 - 1692 [0.556] 2000
Elo difference: 20.9 +/- 5.9, LOS: 100.0 %, DrawRatio: 84.6 %
SPRT: llr 0 (0.0%), lbound -inf, ubound inf
2000 of 2000 games finished.

Cutechess-cli with the following time control:
40 moves in ten seconds with .05 second increment.
It is possible that the time control difference is the reason for the discrepancy.

[Viz]

I did another build with today's code, and the Elo gain is 21:

Score of stockfish-x64-avx2 vs stockfish_24052816_x64_avx2: 214 - 94 - 1692 [0.530]
... stockfish-x64-avx2 playing White: 203 - 30 - 767 [0.587] 1000
... stockfish-x64-avx2 playing Black: 11 - 64 - 925 [0.473] 1000
... White vs Black: 267 - 41 - 1692 [0.556] 2000
Elo difference: 20.9 +/- 5.9, LOS: 100.0 %, DrawRatio: 84.6 %
SPRT: llr 0 (0.0%), lbound -inf, ubound inf
2000 of 2000 games finished.

Cutechess-cli with the following time control:
40 moves in ten seconds with .05 second increment.
It is possible that the time control difference is the reason for the discrepancy.

what kind of book did you use to get this draw rate 85% with 40 moves in ten seconds?
Ofc on drawish books you will get much less elo gain.

[Dann Corbit]

No book