pawnocchio 2.0 Pre, very strong, stronger as strong ... shortly ... strong!

[Frank Quisinsky]

Hi there,

I wrote this to Ipman in Reckless thread!
pawnocchio 2.0 Pre is since days available!

I am looking in the last days on my 3.0 i7 ultrabook a bit and must say ...
Really potential for place 1.

Pure feeling, with special openings, and my own test-positions.
Newer I saw a better result for the complicated Dutch openings. Also King Indian ... fantastic results.

Now running for ETOC-G (for my engine-overview).
https://www.amateurschach.de/main/_engines.htm

I am very sure ...
it can be go very very high ...

Code: Select all

                                 points   1:0 1/2   0:1     %   m-avg     1/0<45       rk
-----------------------------------------------------------------------------------------
01. Stockfish 18 JA             1.084,0   870 428     2   83,38   81   19/  0  + 19   03#  *2
02. Reckless 0.9.0              1.079,5   859 441     0   83,04   75   34/  0  + 34   03#  *2
03. PlentyChess 7.0.37 Pre      1.067,5   838 459     3   82,12   74   17/  0  + 17   03#
04. Obsidian 16.15 CE dev       1.063,5   827 473     0   81,81   77   23/  0  + 23   04#

So, you can see 83,38% is to beat.

LIVE:
https://www.amateurschach.de/fling/etoc-g_t049.html

pawnocchio 2.0 Pre is fighting ...
I await really a lot ... pure feeling!

Best
Frank

PS: For ETOC-G all TOP-250 engines have to play vs. the same group of 26 opponents x 50 games = 1.300 games.
This need a while, I think around one year. Every day you can find a live test-run on my engine-overview.

[Frank Quisinsky]

A bit to ETOC-G:

Code: Select all

Reasons for ETOG-G:
-------------------
01. To have own results for my engine-overview of "TOP-250" engines!
    A very simple idea for testing engines without focusing on Elo.
02. The aim is to avoid high draw rates.
03. The issue is no inflation of playing strength.
04. I would finally like to see a list, where all engines had the same opponents,
    the same hardware and the same conditions.
05. Achieved results should not become “old” and “uninteresting” in the list.
    Long-term improved comparability of results achieved.
06. Easy parameters regarding speculation about playing styles in the first,
    in my opinion the most important phase of the game.
    The number of quick wins / losses, average number of moves as indicators.
    The many extremely high ratings are mostly due to the transition to the
    endgame and, above all, the very important pawn structures.
07. ETOC-G is applicable to engines that reach min. approximately humans
    super grandmaster level.
    Sorry, I can't test over 1.500 available engines alone in homework.
08. Fun for me and I hope for my visitors in the foreground!
09. Easy to copy for anyone.
    Perhaps someone will start an ETOC-T (ETOC-T = Tokyo) at some point.
    On such an easy test-system, many people can work to the same time.
10. Avoid headaches and sleepless nights when your thirst for knowledge about
    engines is not quenched?! :-)

You know, point 10 is most important? The insidious "Computer-Chess-Syndrom"!

The question is:
Do the best engines always have to play against each other? No, they don't.
With a balance opening book, the goal is to limit the number of draws.
With lesser draws the result will be better and better.
With the goal of achieving the greatest possible success with the least amount of effort.
Important are only many opponents, should not be weaker as 900 Elo (Elo-range).

After around 50 test-runs ETOC-G works fine and my test work before wasn't for nothing.

Enough ...
Back to pawnocchio ... again, it can go very, very high.
Test-run need around 20 hours, so around in 18 hours ready.

Best
Frank

[Frank Quisinsky]

389.0 - 80.0
= 82.94%

Very close to SF18, not a surprise after my test-results I created before.
83,38% is the result from SF 18.

So, now after 469 games it's more and more interesting!
83,38% to beat!!

[Frank Quisinsky]

Hi there,

very strong results by pawnocchio 2.0.0 Pre.
Most interesting is the phase before the endgame started?!

Here pawnocchio shows incredible playing strength.
Strongest engines are here not better, perhaps a bit weaker ... I have not enough games for it to say more.
But it seems, the pawn structures are excellent.

I know now why the program have this name.
It tricks the other programs a bit with its pawn structures, and they fall for it.
Pinocchio - Pawnocchio



Fact is:
An interesting development.
An very hard fighter for the half point more in endgames

You have to be into that.
I tend to prefer the attackers but again ... this program is definitely interesting, too.
Given its strength, the program apparently achieves good analysis results in the Dutch Opening as well.
Can't add all that in TalkChess, sorry!

Best
Frank

[peter]

Fine engine indeed. Prerelease 2.0 shows even some progress to 1.9 in my EloStatTS- screening list, but in CuteChess GUI it doesn't run, tried to restart with different hardware- TCs several times, each one ended at once with illegal moves from pawnocchio, here's one of the many examples of games that ended like this:
[pgn][Event "Dra-paw"]
[Site "?"]
[Date "2026.04.08"]
[Round "1"]
[White "Pawnocchio2.0"]
[Black "Dragon3.3"]
[Result "0-1"]
[ECO "B01"]
[PlyCount "12"]
[GameDuration "00:00:00"]
[GameEndTime "2026-04-08T12:54:22.119 Mitteleuropäische Sommerzeit"]
[GameStartTime "2026-04-08T12:54:22.103 Mitteleuropäische Sommerzeit"]
[Opening "Scandinavian (centre counter) defense"]
[Termination "illegal move"]
[TimeControl "60/180+1"]

1.e4 {book} 1...d5 {book} 2.exd5 {book} 2...Qxd5 {book} 3.Nc3 {book} 3...Qa5 {book} 4.Nf3 {book} 4...Nf6 {book} 5.d4 {book} 5...c6 {book} 6.Bc4 {book} 6...Bg4 {book, White makes an illegal move: 0000} 0-1
[/pgn]
Hardware was single thread of 16x4.3GHz CPU, 30 concurrencies.

Edit: tried 5 concurrencies with 6 threads per engine and 10 concurrencies single thread too, each time the same, always illegal moves from pawnocchio at once after book moves' ends.

BTW in Fritz20 no problems at 50 games with 15 threads each and 10'+5", no concurrencies, rather even result to Dragon3.3, which was the opponent, but of course much too small sample size, regards

[cc2150dx]

Hey Frank. Is this the same version https://github.com/JonathanHallstrom/pa ... ase-builds if so, I can't find the pawnocchio-2.0-dev-a8f09e7-windows-x86-64-avx2.exe version. Am I missing something ?

Thanks

[peter]

Not sure about the meaning of the v2 to v4 and the znver- versions by myself neither, in such cases I always start downloading the highest one v- number I think to be possible for my system, here I was lucky already with znver5.exe for my Ryzen 9950x, so I guess this one will be something like vnni512 maybe.

Works well in all GUIs I tried so far except (see above) CuteChess GUI. Have another one little match between Dragon 3.3 and the pawnocchio prerelease 2.0 in Shredder GUI in meantime, so far it runs like this with 6 threads of the 16x4.3GHz CPU, 3'+1" and UHO 2024 85-94cp 6mvs:

Dragon3.3 - Pawnocchio2.0
34.5 - 46.5 (81: +14, =41, -26) 42.59% Perf = -51 Elo
( 42.5% TP = -51 Elo 68%->[-86,-28] 95%->[-122,-5] 99.7%->[-161,+17] )
81 games played

[cc2150dx]

Thanks for the tip Peter. Pretty impressive results so far. Glad to know it works with Fritz. I'll check with Aquarium.

[cpeters]

Hardware was single thread of 16x4.3GHz CPU, 30 concurrencies.

Edit: tried 5 concurrencies with 6 threads per engine and 10 concurrencies single thread too, each time the same, always illegal moves from pawnocchio at once after book moves' ends.

Could you elaborate please? Especially this:

5 concurrencies with 6 threads per engine

10 concurrencies single thread

Shouldn't be a problem. Are there log-files?

greetings

[cc2150dx]

Not sure about the meaning of the v2 to v4 and the znver- versions by myself neither, in such cases I always start downloading the highest one v- number I think to be possible for my system, here I was lucky already with znver5.exe for my Ryzen 9950x, so I guess this one will be something like vnni512 maybe.

Works well in all GUIs I tried so far except (see above) CuteChess GUI. Have another one little match between Dragon 3.3 and the pawnocchio prerelease 2.0 in Shredder GUI in meantime, so far it runs like this with 6 threads of the 16x4.3GHz CPU, 3'+1" and UHO 2024 85-94cp 6mvs:

Dragon3.3 - Pawnocchio2.0
34.5 - 46.5 (81: +14, =41, -26) 42.59% Perf = -51 Elo
( 42.5% TP = -51 Elo 68%->[-86,-28] 95%->[-122,-5] 99.7%->[-161,+17] )
81 games played

znver3 works on my Ryzen 5 3600.