New chess Engines clones and News 2021

[Sylwy]

Well, I mean you are running Stockfish with a speed disadvantage (AVX2 vs BMI2) vs very similar engines...

I've already answered this thing. No.1 in this tournament is also AVX2!

[Sylwy]

1.-The current development version may be 10 ELOs better.
2.-Stockfish 14.1 of course has a bmi version, but you would have to click one link.

https://abrok.eu/stockfish/

1.-Do you really believe all these increases of Elo presented on abrok site .....? The M31 engine (not really new) still beats the latest SF version!
And then you make the longer tournaments with the final versions, not with the development ones (intermediate) ........ so at any moment a point, a comma changes ...... in the code (in which case only hyperbulett tournaments-which I do not agree with- can take place).

2.-Of course I have SF 14.1 bmi2:

https://stockfishchess.org/download/windows/

The problem was that at the start of the tournament I only had an AVX2 compilation for the M31 engine (no.1 in my tournament). I wanted to put them (SF 14.1-M31) on equal terms. Hard to understand?

Bonus:
======

Here is the last (and most economical) diagram of my tournament:

[Guenther]

1.-The current development version may be 10 ELOs better.
2.-Stockfish 14.1 of course has a bmi version, but you would have to click one link.

:lol:

https://abrok.eu/stockfish/

1.-Do you really believe all these increases of Elo presented on abrok site .....? The M31 engine (not really new) still beats the latest SF version!
And then you make the longer tournaments with the final versions, not with the development ones (intermediate) ........ so at any moment a point, a comma changes ...... in the code (in which case only hyperbulett tournaments-which I do not agree with- can take place).

It's like a kind of talking to flat earth believers...

So now you finally say all of those games (which are not avilable too) are run on a smartphone and most of them via a WIN emulator?
Why do you wonder then that a native compilation of an older SF is better? All of this is apples and coconuts and also much too less games.
Actually I wonder why I even reply here lol. I couldn't care less for random smartphone games with random compilations on random
start positions with zero documentation.

[Rubinus]

So I'm just saying that the games are few and the differences are minimal.1.5 points out of 127 games is just over 1%. < 10 ELO. If I put the exact same Stockfish in eight different directories and rename them in the uci or Arena settings, like SF dir A, B, etc, then that result will be about the same.

Plus, those engines are already so powerful that they rarely lose a straight library position. The difference is in the asymmetric ones. I believe, for example, Stockfish and Dragon are different programs, yet they play almost the same. The biggest difference is that if I let them play deliberately with a wide library, so that some variants are asymmetric, then the better ones Stockfish wins almost all of them, while Dragon only wins some. I tried Ethereal this way too, and it only won the completely won ones against Stockfish.

I was thinking of creating some sort of "only first" testsets with only one opening move, then we'd have 20 variants and twice as many with rematches. Or "first two", that would be 400 variants, 800 games - too many, we'd have to play about 1+1 to get real-time results.
The most powerful engines don't need libraries much anymore, on the contrary the long ones would rather hamper them, if they were created from some of those old "2600+" type instructions, they serve nowadays only to avoid playing one variant over and over again in games, which for example with Sedat's Perfect

[Sylwy]

Actually I wonder why I even reply here lol.

That's right! If you were silent you were a philosopher ....... say a word somewhere in the galaxy M31 ......

[Sylwy]

So I'm just saying that the games are few and the differences are minimal.1.5 points out of 127 games is just over 1%. < 10 ELO. If I put the exact same Stockfish in eight different directories and rename them in the uci or Arena settings, like SF dir A, B, etc, then that result will be about the same.

Plus, those engines are already so powerful that they rarely lose a straight library position. The difference is in the asymmetric ones. I believe, for example, Stockfish and Dragon are different programs, yet they play almost the same. The biggest difference is that if I let them play deliberately with a wide library, so that some variants are asymmetric, then the better ones Stockfish wins almost all of them, while Dragon only wins some. I tried Ethereal this way too, and it only won the completely won ones against Stockfish.

I was thinking of creating some sort of "only first" testsets with only one opening move, then we'd have 20 variants and twice as many with rematches. Or "first two", that would be 400 variants, 800 games - too many, we'd have to play about 1+1 to get real-time results.
The most powerful engines don't need libraries much anymore, on the contrary the long ones would rather hamper them, if they were created from some of those old "2600+" type instructions, they serve nowadays only to avoid playing one variant over and over again in games, which for example with Sedat's Perfect

Wait ! There will be several hundred games on each engine (and if I want I can increase them at any time). With the 1st place the difference is bigger (in fact the performances of the 1st place face to SF 14.1 interest me). I already said: SF 14.1-M31 ...... the rest .... salt and pepper (smartphones, emulators, random compilations, zero documentation.....blah-blah-blah.... ) .....

[AlexChess]

@Sylwy which time controls do you use and how many positions calculates your hardware (on average) ?
I'm also very disappointed by Stockfish 14.1 results (out from top 10) ! Luckily there are also the Chessman / Skynet compiles of the latest dev versions that are much stronger and include Polyglot book openings support: http://outskirts.altervista.org/forum/v ... 367#p38367

Our tournaments are perfecly complementary: All top 3100-3700 ELO engines and All 3600+ ELO Stars

Best regards, Alex

Test conditions:
-TC=4'+2" (12-15 minutes per game)
-Hash=256 MB
-GUI: Arena 3.5.1
-Books: IM_4mvs.abk (8 plies book) for all engines
-default settings for both engines
-1 thread-CPU=Intel i5-7400-3GHz (Kaby Lake)
-6-men Syzygy bases for all engines
-OS: Windows 10 Home.

OK, then my hardware speed is 2x on average than your. But I have just cutted the time controls by half and doubled the number of games from 72 to 144 to have a lower error margin, so our results are comparables.



PS: If trolls are annoying you, simply use add foe from the control panel clicking their name, as I finally discovered (I read their boring comments only because you quote them ) It would be better to have an option for blocking them from commenting in our topics, but only moderators can do that...

Best regards, Alex

[AlexChess]

Actually I wonder why I even reply here lol.

That's right! If you were silent you were a philosopher ....... say a word somewhere in the galaxy M31 ......

+1

[Rubinus]

Wait ! There will be several hundred games on each engine (and if I want I can increase them at any time). With the 1st place the difference is bigger (in fact the performances of the 1st place face to SF 14.1 interest me). I already said: SF 14.1-M31 ...... the rest .... salt and pepper (smartphones, emulators, random compilations, zero documentation.....blah-blah-blah.... ) .....

So your competition, do what you want, I'm not particularly convinced so far. I tried Raubfish a few years ago, the results were inconclusive. And I need different sounding engines rather than several almost identical ones. I'm not fundamentally opposed to clones in principle, as long as I can see the added value. The Rybka and Houdini, for example, had that, in my opinion, if I had been the one to decide and not the old ICGA guys, Vasik would have kept the medals. But here I don't see any big contribution between Stockfish, Sugar and those other pieces.

[Sylwy]

Wait ! There will be several hundred games on each engine (and if I want I can increase them at any time). With the 1st place the difference is bigger (in fact the performances of the 1st place face to SF 14.1 interest me). I already said: SF 14.1-M31 ...... the rest .... salt and pepper (smartphones, emulators, random compilations, zero documentation.....blah-blah-blah.... ) .....

So your competition, do what you want, I'm not particularly convinced so far. I tried Raubfish a few years ago, the results were inconclusive. And I need different sounding engines rather than several almost identical ones. I'm not fundamentally opposed to clones in principle, as long as I can see the added value. The Rybka and Houdini, for example, had that, in my opinion, if I had been the one to decide and not the old ICGA guys, Vasik would have kept the medals. But here I don't see any big contribution between Stockfish, Sugar and those other pieces.

It happens that-sometimes-students surpass their teachers ......... translated: sometimes clones / derivatives exceed the original. That's life. An interesting article (if you read it all):

https://glarean-magazin.ch/2020/06/01/c ... ournament/