Fire 8 = Ethereal 13.00 NNUE

[Chessqueen]

So far the score is equal, but I will NOT run 1000 games to find out, 20 games should give me an idea
Fire 8.NN.MC.3 x64

Engine Score Fi Et S-B
1: Fire_8_x64_pext 4.0/7 ······· ====101 12.00
2: Ethereal-pext-avx2 3.0/7 ====010 ······· 12.00

7 of 20 games played
Name of the tournament: Fire8
Site/ Country: MININT-UB2PIMJ, United States
Level: Blitz 5/5
Hardware: Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz with 15.9 GB Memory
Operating system: Windows 10 Enterprise Professional (Build 9200) 64 bit
PGN-File: C:\Program Files (x86)\Arena\Tournaments\IGel.pgn
Website:
E-Mail Address:

[Ras]

20 games should give me an idea

You have no idea. 20 games is nothing.

[amanjpro]

20 games should give me an idea

You have no idea. 20 games is nothing.

He gives him an idea that one engine is not infinitly weaker than the other , that is the only conclusion one can draw form 20 games

[Chessqueen]

20 games should give me an idea

You have no idea. 20 games is nothing.

And you believe that if X engine is holding Y engine into an even score after 20 games that either X engine or y engines will eventually come up ahead by 20 games at the end of 1000 games, I do NOT think so, chess engines are not like human and if we use the same opening book (Perfect2021.abk) for both engines after 1000 games, unless we use stupid opening book like TCEC than we can see a big difference after 1000 games.

My conclusion is that Fire 8.NN.MC.3 x64 is equal in strength to Ethereal 13.00 NNUE using same opening book Perfect2021.abk

Engine Score Fi Et S-B
1: Fire_8_x64_pext 10.5/20 ···················· ====101===1=====0=== 99.75
2: Ethereal-pext-avx2 9.5/20 ====010===0=====1=== ···················· 99.75

20 games played / Tournament is finished
Name of the tournament: Fire8
Site/ Country: MININT-UB2PIMJ, United States
Level: Blitz 5/5
Hardware: Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz with 15.9 GB Memory
Operating system: Windows 10 Enterprise Professional (Build 9200) 64 bit
PGN-File: C:\Program Files (x86)\Arena\Tournaments\IGel.pgn
Website:
E-Mail Address:

[Ras]

And you believe

I don't have to believe because I know how statistics work.

If both engines are about equal, you can expect +/-SQRT(N) as error margin. N=20 is very small for even that formula to apply, so it will give you an optimistic lower bound for your expected error. The expected result with even opponents will be between 5.5 and 14.5 points. Your error bar is 170 Elo at least.

So you're obviously talking nonsense. 20 games are ridiculous to conclude equal strength.

[Chessqueen]

And you believe

I don't have to believe because I know how statistics work.

If both engines are about equal, you can expect +/-SQRT(N) as error margin. N=20 is very small for even that formula to apply, so it will give you an optimistic lower bound for your expected error. The expected result with even opponents will be between 5.5 and 14.5 points. Your error bar is 170 Elo at least.

So you're obviously talking nonsense. 20 games are ridiculous to conclude equal strength.

Simply put it into a test, as I mentioned before the error bar is not 170 Elo unless you are using different opening books at the most it would be probably 70 Elo. Go ahead start your test of 1000 games and we will talk again next week.

[AndrewGrant]

Still waiting for you to confirm this is actually NNUE and not the github version.