Are we sure that Stockfish NNUE is better than the Normal Stockfish ?

Discussion of anything and everything relating to chess playing software and machines.

Moderators: Harvey Williamson, bob, hgm

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Chessqueen
Posts: 1078
Joined: Wed Sep 05, 2018 12:16 am
Full name: Nancy M Pichardo

Re: Are we sure that Stockfish NNUE is better than the Normal Stockfish ?

Post by Chessqueen » Sun Oct 18, 2020 1:42 pm

Leto wrote:
Sat Oct 17, 2020 3:41 pm
Chessqueen wrote:
Sat Oct 17, 2020 3:14 pm
Werewolf wrote:
Sat Oct 17, 2020 2:32 pm
Dann Corbit wrote:
Sat Oct 10, 2020 4:40 am
Chessqueen wrote:
Sat Oct 10, 2020 2:38 am
corres wrote:
Tue Oct 06, 2020 9:12 am
mwyoung wrote:
Mon Oct 05, 2020 11:36 pm
...
And the reason for this is TCEC uses bias openings. If a match shows a win for both engines in the same line. It is a busted opening. But we already know from TCEC own words. Wins are more important for views. So this is expected.
TCEC is not a fair competition, it is only an interesting show of chess.
You are correct to say that TCEC uses very bad openings, instead of using balanced Openings where there is no big advantage to either side to determine which engine is the best. They could use a balanced 4 or 6 opening moves.
How exactly do the bad openings make the contest unfair?
Exactly.
A strong engine must be able to win won positions.
You probably meant the strongest of the two engines competing should be able to win a won position after the opening, but if the engine is 85- 95 Elo weaker like in the case of Komodo the most that it will do is win only 33% of the games where the opening is better for White. For instance in this last game the opening is considered very balanced where White has a slight advantage of no more than 3.5, but because white or Komodo which is the weaker engine has the first move it will probably draw. https://tcec-chess.com/live.html#div=sf ... &season=19
What do you mean 3.5? As in 3 and a half pawns advantage?
If you click on the Link that I provided under the diagram wright after Stockfish Classical it will give you [3.3% w==>
https://tcec-chess.com/live.html#div=sf ... &season=19

syzygy
Posts: 4797
Joined: Tue Feb 28, 2012 10:56 pm

Re: Are we sure that Stockfish NNUE is better than the Normal Stockfish ?

Post by syzygy » Sun Oct 18, 2020 3:07 pm

Chessqueen wrote:
Sun Oct 18, 2020 1:42 pm
Leto wrote:
Sat Oct 17, 2020 3:41 pm
Chessqueen wrote:
Sat Oct 17, 2020 3:14 pm
You probably meant the strongest of the two engines competing should be able to win a won position after the opening, but if the engine is 85- 95 Elo weaker like in the case of Komodo the most that it will do is win only 33% of the games where the opening is better for White. For instance in this last game the opening is considered very balanced where White has a slight advantage of no more than 3.5, but because white or Komodo which is the weaker engine has the first move it will probably draw. https://tcec-chess.com/live.html#div=sf ... &season=19
What do you mean 3.5? As in 3 and a half pawns advantage?
If you click on the Link that I provided under the diagram wright after Stockfish Classical it will give you [3.3% w==>
https://tcec-chess.com/live.html#div=sf ... &season=19
That's a drawn ending position where both engines estimate win/draw/loss probabilities at 3.3%/93.4%/3.3%. Doesn't have much to do with the opening position?

Chessqueen
Posts: 1078
Joined: Wed Sep 05, 2018 12:16 am
Full name: Nancy M Pichardo

Re: Are we sure that Stockfish NNUE is better than the Normal Stockfish ?

Post by Chessqueen » Mon Oct 19, 2020 2:50 am

syzygy wrote:
Sun Oct 18, 2020 3:07 pm
Chessqueen wrote:
Sun Oct 18, 2020 1:42 pm
Leto wrote:
Sat Oct 17, 2020 3:41 pm
Chessqueen wrote:
Sat Oct 17, 2020 3:14 pm
You probably meant the strongest of the two engines competing should be able to win a won position after the opening, but if the engine is 85- 95 Elo weaker like in the case of Komodo the most that it will do is win only 33% of the games where the opening is better for White. For instance in this last game the opening is considered very balanced where White has a slight advantage of no more than 3.5, but because white or Komodo which is the weaker engine has the first move it will probably draw. https://tcec-chess.com/live.html#div=sf ... &season=19
What do you mean 3.5? As in 3 and a half pawns advantage?
If you click on the Link that I provided under the diagram wright after Stockfish Classical it will give you [3.3% w==>
https://tcec-chess.com/live.html#div=sf ... &season=19
That's a drawn ending position where both engines estimate win/draw/loss probabilities at 3.3%/93.4%/3.3%. Doesn't have much to do with the opening position?
I am NOT referring to the ending ,but if you look at the first 8 moves where both engines estimate win/draw/loss probabilities at 22.4% for white Komodo being the weaker engine could NOT convert that advantage all the way to the end, but in the reverse where Stockfish classical got the same opening it was a total disaster for Komodo.

On a different subject I believe the LCZero and Stockfish Classical are very much equal on strength

syzygy
Posts: 4797
Joined: Tue Feb 28, 2012 10:56 pm

Re: Are we sure that Stockfish NNUE is better than the Normal Stockfish ?

Post by syzygy » Mon Oct 19, 2020 11:05 am

Chessqueen wrote:
Mon Oct 19, 2020 2:50 am
syzygy wrote:
Sun Oct 18, 2020 3:07 pm
Chessqueen wrote:
Sun Oct 18, 2020 1:42 pm
Leto wrote:
Sat Oct 17, 2020 3:41 pm
Chessqueen wrote:
Sat Oct 17, 2020 3:14 pm
You probably meant the strongest of the two engines competing should be able to win a won position after the opening, but if the engine is 85- 95 Elo weaker like in the case of Komodo the most that it will do is win only 33% of the games where the opening is better for White. For instance in this last game the opening is considered very balanced where White has a slight advantage of no more than 3.5, but because white or Komodo which is the weaker engine has the first move it will probably draw. https://tcec-chess.com/live.html#div=sf ... &season=19
What do you mean 3.5? As in 3 and a half pawns advantage?
If you click on the Link that I provided under the diagram wright after Stockfish Classical it will give you [3.3% w==>
https://tcec-chess.com/live.html#div=sf ... &season=19
That's a drawn ending position where both engines estimate win/draw/loss probabilities at 3.3%/93.4%/3.3%. Doesn't have much to do with the opening position?
I am NOT referring to the ending ,but if you look at the first 8 moves where both engines estimate win/draw/loss probabilities at 22.4% for white Komodo being the weaker engine could NOT convert that advantage all the way to the end, but in the reverse where Stockfish classical got the same opening it was a total disaster for Komodo.
OK, then I am not sure what you were referring to with "no more than 3.5" and "3.3% w".

syzygy
Posts: 4797
Joined: Tue Feb 28, 2012 10:56 pm

Re: Are we sure that Stockfish NNUE is better than the Normal Stockfish ?

Post by syzygy » Mon Oct 19, 2020 12:42 pm

Gary Internet wrote:
Sat Oct 17, 2020 7:24 am
Chessqueen wrote:
Sat Oct 17, 2020 3:23 am
They should be testing the normal Stockfish Vs Stockfish NNUE instead of versus Komodo https://tcec-chess.com/live.html
They will be. The winner is SF Classic vs Komodo then goes on to play SFNNUE. You will get the match you want. Be patient.
Next match I'd like to see is SF NNUE Hybrid vs SF NNUE Pure.

Chessqueen
Posts: 1078
Joined: Wed Sep 05, 2018 12:16 am
Full name: Nancy M Pichardo

Re: Are we sure that Stockfish NNUE is better than the Normal Stockfish ?

Post by Chessqueen » Tue Oct 20, 2020 3:24 am

syzygy wrote:
Mon Oct 19, 2020 12:42 pm
Gary Internet wrote:
Sat Oct 17, 2020 7:24 am
Chessqueen wrote:
Sat Oct 17, 2020 3:23 am
They should be testing the normal Stockfish Vs Stockfish NNUE instead of versus Komodo https://tcec-chess.com/live.html
They will be. The winner is SF Classic vs Komodo then goes on to play SFNNUE. You will get the match you want. Be patient.
Next match I'd like to see is SF NNUE Hybrid vs SF NNUE Pure.
The next match will be Stockfisk Classical Vs LCZero. I remember when Stockfish NNue was playing versus LCZero after 35 games the score was 21 to 14 in favor of SF NNUE which is exactly the same score that SF NNUE Vs SF Classical at the moment, so SF classical vs LCZero will be a very close match :shock: https://tcec-chess.com/live.html

syzygy
Posts: 4797
Joined: Tue Feb 28, 2012 10:56 pm

Re: Are we sure that Stockfish NNUE is better than the Normal Stockfish ?

Post by syzygy » Tue Oct 20, 2020 11:29 am

Chessqueen wrote:
Tue Oct 20, 2020 3:24 am
The next match will be Stockfisk Classical Vs LCZero. I remember when Stockfish NNue was playing versus LCZero after 35 games the score was 21 to 14 in favor of SF NNUE which is exactly the same score that SF NNUE Vs SF Classical at the moment, so SF classical vs LCZero will be a very close match :shock: https://tcec-chess.com/live.html
SF NNUE vs LC0 was much closer after 35 games. SF NNUE made the difference only in the second half of the match where openings were less balanced. Of course time control was much longer as well, which likely increased the draw rate.

MMarco
Posts: 132
Joined: Sat Apr 11, 2020 11:09 pm
Full name: Marc-O Moisan-Plante

Re: Are we sure that Stockfish NNUE is better than the Normal Stockfish ?

Post by MMarco » Tue Oct 20, 2020 6:07 pm

Chessqueen wrote:
Tue Oct 20, 2020 3:24 am
syzygy wrote:
Mon Oct 19, 2020 12:42 pm
Gary Internet wrote:
Sat Oct 17, 2020 7:24 am
Chessqueen wrote:
Sat Oct 17, 2020 3:23 am
They should be testing the normal Stockfish Vs Stockfish NNUE instead of versus Komodo https://tcec-chess.com/live.html
They will be. The winner is SF Classic vs Komodo then goes on to play SFNNUE. You will get the match you want. Be patient.
Next match I'd like to see is SF NNUE Hybrid vs SF NNUE Pure.
The next match will be Stockfisk Classical Vs LCZero. I remember when Stockfish NNue was playing versus LCZero after 35 games the score was 21 to 14 in favor of SF NNUE which is exactly the same score that SF NNUE Vs SF Classical at the moment, so SF classical vs LCZero will be a very close match :shock: https://tcec-chess.com/live.html
Last SuFi was a tie after 35 games: 4-4-27 (
https://tcec-chess.com/#div=sf&game=1&season=19 )
I predict Lc0 will clobber SF Classical. +15 at least! 🙂

Chessqueen
Posts: 1078
Joined: Wed Sep 05, 2018 12:16 am
Full name: Nancy M Pichardo

Re: Are we sure that Stockfish NNUE is better than the Normal Stockfish ?

Post by Chessqueen » Wed Oct 21, 2020 2:19 am

MMarco wrote:
Tue Oct 20, 2020 6:07 pm
Chessqueen wrote:
Tue Oct 20, 2020 3:24 am
syzygy wrote:
Mon Oct 19, 2020 12:42 pm
Gary Internet wrote:
Sat Oct 17, 2020 7:24 am
Chessqueen wrote:
Sat Oct 17, 2020 3:23 am
They should be testing the normal Stockfish Vs Stockfish NNUE instead of versus Komodo https://tcec-chess.com/live.html
They will be. The winner is SF Classic vs Komodo then goes on to play SFNNUE. You will get the match you want. Be patient.
Next match I'd like to see is SF NNUE Hybrid vs SF NNUE Pure.
The next match will be Stockfisk Classical Vs LCZero. I remember when Stockfish NNue was playing versus LCZero after 35 games the score was 21 to 14 in favor of SF NNUE which is exactly the same score that SF NNUE Vs SF Classical at the moment, so SF classical vs LCZero will be a very close match :shock: https://tcec-chess.com/live.html
Last SuFi was a tie after 35 games: 4-4-27 (
https://tcec-chess.com/#div=sf&game=1&season=19 )
I predict Lc0 will clobber SF Classical. +15 at least! 🙂

I do NOT think so, they will be very close probably 52.5 47.5 in favor of LCZero


PS: I believe that in order to make the 100 games fair they should have 30 games from the start position with no Opening, another 30 games with balanced positions and the rest of 40 games with White having a small advantage, but NOT a HUGE Advantage.

Jouni
Posts: 2223
Joined: Wed Mar 08, 2006 7:15 pm

Re: Are we sure that Stockfish NNUE is better than the Normal Stockfish ?

Post by Jouni » Wed Oct 21, 2020 6:31 am

Now NNUE is +80 to classical with 172 cores. Big difference to CCC tournament were they have almost same score with 90 cores. The truth is somewere between I quess :) .
Jouni

Post Reply