Leela Chess Zero Beats Stockfish 106-94...

[Sedat Canbaz]

Are these games before four and half years!?

Indeed...Deeds's shared database belongs to 2020 year
But we are in 2024...something strange here for sure!
Actually then whole thread is pointless here since
Too old news to discuss all over...but this is also true that
- My stated points are not dead.. it does not matter 2020
Or 2024...they can be used as guide...even in future as well!

[Sedat Canbaz]

They are old games because zamchess said :

Leela Chess Zero Beats Stockfish 106-94 In 13th Chess.com Computer Chess Championship !

And they are posted here because sedat said :

- the choice of openings favors lc0
Since games are not available...no comments!

I was thinking that 13th CCCC CS is new...
If played in 2020... Let's forget about it...)

Actually we can still discuss all about...
For this I strongly suggest SCCT results!
At least all played data belongs to Oct 2024 !
Sure I refer about why latest SF dev is weaker...

[ZamChess]

games

Thanks DeeDs for the games !

[ZamChess]

Sorry ! It's my fault...
When I saw this title, I was under the impression that it was the last tourney.
After posting DeeDs with the parties, I realized the mistake I made. Again, sorry.

[Sedat Canbaz]

Sorry ! It's my fault...
When I saw this title, I was under the impression that it was the last tourney.
After posting DeeDs with the parties, I realized the mistake I made. Again, sorry.

Do not worry my friend.. it's ok,
No one's work is perfect here )

[Homayoun]

Sorry ! It's my fault...
When I saw this title, I was under the impression that it was the last tourney.
After posting DeeDs with the parties, I realized the mistake I made. Again, sorry.

No matter Zamchess. Your post caused our dream became true at least for a moment. We want to see leela or at least another engine would beat Stockfish again. This Stockfish has made chess fans tired by its great victories and results.
All the bests

[deeds]

Stockfish is beaten by its own playing style.
Example with an experience file which contains only Stockfish's moves :
https://www.open-chess.org/viewtopic.php?f=4&t=4145

Other engines have beaten Stockfish.
Example with a system which uses any learning engine and its own experience file :
https://www.open-chess.org/viewtopic.ph ... =10#p27324

[Sedat Canbaz]

Wow.. what a thread ?) we discuss even for 2020 )

Anyhow nice to discuss... since 1st happens..
At least never seen before for sure ))

And just my 2 cents more over these issues:

Well, all these CCCC 2020 matches are not fair at all,
Sure if CCCC duel was played in 2024..I mean with
Fair cond. etc then would be ideal..but unfortunately not !
Sure here I refer if CPU vs CPU eng matches..not vs GPU
And only then their results would be great job for Lc0..
But it seems is not so at all, right ?)

If not so clear, (about 13th CCCC),
The mentioned Duel belongs to Apr 2020
That means Stokfish is played as NON-NN
In short, SF's size: 1-2 MB size, where
Lc0's size is unknown, but I assume that
Much much larger size than Stockfish!

In other words,
Lc0 is played as NN, via huge NN assistance...
Where SF is played under Classic conditions..
Moreover, no one can predict here CPU vs GPU
Speed difference...

If still not so clear much, give me so much assistance..
I will be able to beat any Strong player on this planet!

Btw, it's something like in 2018: Stockfish 8 vs AlphaZero,
What happened in 2018 ? AlphaZero won, but how much
Their match was fair ? this is other question of course )

Hope helps..

[mwyoung]

Btw, it's something like in 2018: Stockfish 8 vs AlphaZero,
What happened in 2018 ? AlphaZero won, but how much
Their match was fair ? this is other question of course )

Let us answer this question. And take CPU vs GPU out of the equation.

Here I played Stockfish dev 18 vs Stockfish 8

But unlike in 2017, I played Stockfish 8 at full power. With 64 CPU, 64 GB hash, opening book, and tablebases.

I played the NN engine Stockfish dev 18 with 1 CPU, and 1 GB hash.

Here like AlphaZero, the NN engine destroyed Stockfish 8 even more then Alphazero. Even with a handicap. This leaves little doubt that Alphazero would beat Stockfish 8. With fair conditions....

Code: Select all

 esult:
--------------------------------------------------------------------------------------
  #  name                      games            wins           draws          losses   score    elo    +    -
  1. Stockfish dev 18 (1 CPU)    100       38 100.0%       58 100.0%        4 100.0%      67     49   27   26
  2. Stockfish 8 (64 CPU)        100        4 100.0%       58 100.0%       38 100.0%      33    -49   26   27

Cross table:
--------------------------------------------------------------------------------------
  #  name                         score   games         1         2
  1. Stockfish dev 18 (1 CPU)        67     100         x      67.0
  2. Stockfish 8 (64 CPU)            33     100      33.0         x

Tech:
--------------------------------------------------------------------------------------

Tech (average nodes, depths, time/m per move, others per game), counted for computing moves only, ignored moves with zero nodes:
  #  name                        nodes/m         NPS  depth/m   time/m    moves     time
  1. Stockfish dev 18 (1 CPU)    135421K     1899481     67.3     71.3     75.3   5365.6
  2. Stockfish 8 (64 CPU)      12078233K   160715876     53.0     75.2     75.0   5633.4
     all ---                    5952052K    83241753     60.2     73.2     75.1   5499.5
 

[Sedat Canbaz]

Btw, it's something like in 2018: Stockfish 8 vs AlphaZero,
What happened in 2018 ? AlphaZero won, but how much
Their match was fair ? this is other question of course )

Let us answer this question. And take CPU vs GPU out of the equation.

Here I played Stockfish dev 18 vs Stockfish 8

But unlike in 2017, I played Stockfish 8 at full power. With 64 CPU, 64 GB hash, opening book, and tablebases.

I played the NN engine Stockfish dev 18 with 1 CPU, and 1 GB hash.

Here like AlphaZero, the NN engine destroyed Stockfish 8 even more then Alphazero. Even with a handicap. This leaves little doubt that Alphazero would beat Stockfish 8. With fair conditions....

Code: Select all

 esult:
--------------------------------------------------------------------------------------
  #  name                      games            wins           draws          losses   score    elo    +    -
  1. Stockfish dev 18 (1 CPU)    100       38 100.0%       58 100.0%        4 100.0%      67     49   27   26
  2. Stockfish 8 (64 CPU)        100        4 100.0%       58 100.0%       38 100.0%      33    -49   26   27

Cross table:
--------------------------------------------------------------------------------------
  #  name                         score   games         1         2
  1. Stockfish dev 18 (1 CPU)        67     100         x      67.0
  2. Stockfish 8 (64 CPU)            33     100      33.0         x

Tech:
--------------------------------------------------------------------------------------

Tech (average nodes, depths, time/m per move, others per game), counted for computing moves only, ignored moves with zero nodes:
  #  name                        nodes/m         NPS  depth/m   time/m    moves     time
  1. Stockfish dev 18 (1 CPU)    135421K     1899481     67.3     71.3     75.3   5365.6
  2. Stockfish 8 (64 CPU)      12078233K   160715876     53.0     75.2     75.0   5633.4
     all ---                    5952052K    83241753     60.2     73.2     75.1   5499.5
 

Hello Mwyoung,

Very very interesting test I like it!
And thanks a lot..

So, NNUE 1 CPU of 2025 seems to be much
Stronger than NON-NNUE 64 CPU of 2017...

Btw, can you share the played games please..

Greetings