Patricia 4

[pohl4711]

IMHO, right now the balance of Celo and EAS is really, really good in Patricia 4: Patricia 4 is -90 Celo weaker than Stockfish final HCE, means a strength somewhere in between Stockfish 9 and Stockfish 10. And that is already way above superhuman-level, and the aggressiveness of Patricia 4 is just awesome!

And whats also great is: Patricia 4 has a very small nnue-net, which leads to very high nps. It is as fast as the good, old HCE nodecrunchers (Komodo 14.1 HCE has -10% less nps than Patricia 4 (!))
So, Patricia 4 will perform very well, even though when running on slow hardware (cheap Smartphones or Mephisto Phoenix (if Millenium does not put Patricia 4 on their machine, they are just idiots)). Very good!

[tapio]

Thank you very much. I'm interested only in an interesting opponent to play against me. I'm super weak (and always blunder), so I'm hoping strength 1 plays well. It will probably still beat me

[jefk]

bigger gambit book (posted and explained in this forum some time ago);
polyglot format

https://sourceforge.net/projects/chess- ... n/download

decent and sharp book for P4 (crazy/unsound gambits were removed
and less sound gambits have lower playing frequencies); includes many
relatively unknown (or new) gambits for White and Black

anyone knows where this trap.ctg book can be found?
meanwhile i found this,
https://fauziakram.gumroad.com/l/traps

but i don't think i'm going to spend 19 bucks on it

[tapio]

We've talked about it in Patricia 3 thread. The engine reports 5 multipv moves to the GUI at the same time, which is some display overload. Always, if strength <20, I think. Didn't you want to limit the reporting?

[Whiskers]

We've talked about it in Patricia 3 thread. The engine reports 5 multipv moves to the GUI at the same time, which is some display overload. Always, if strength <20, I think. Didn't you want to limit the reporting?

I completely forgot, sorry!

[Peter Berger]

When performing final testing with Pohl, I came to the uncomfortable realization that she is a good deal less aggressive in unbalanced books than more balanced books. At her strength, the reason is obvious - the only chance of winning a balanced opening against another strong engine is to go ham with the sacrifices, so nearly all her wins were sacrifices leading to an artificially high EAS score.

Pohl's testing still supports it being the world's most aggressive engine, however; she's also much stronger and it's probably going to take me at least a couple months to get her EAS back to 400,000+ in conditions where it matters. So, release time it is. I hope you all enjoy this new Patricia.

First: thank you very much for the release!
But your/Stefan's experiences also add to some other recent discussions. We see a clear difference between "modern" (sth like UHO openings) and "classical" conditions ( balanced positions) when it is about engine behaviour and style.
And this is not only happening to Patricia of course for obvious reasons.

[pohl4711]

First: thank you very much for the release!
But your/Stefan's experiences also add to some other recent discussions. We see a clear difference between "modern" (sth like UHO openings) and "classical" conditions ( balanced positions) when it is about engine behaviour and style.
And this is not only happening to Patricia of course for obvious reasons.

It is clear, that using UHO openings lower the EAS-scoring, because in 50% of the games (when playing black), the engine has a worse position, when the engines start thinking. It is clear, that this makes it harder to win with a sac or win very quick (short wins), compared to balanced openings (or playing UHO with white).

In general, EAS-calculation and EAS-points depend heavily on the playing-conditions: UHO lowers EAS-scoring. The higher the aggressiveness of the opponents, the lower the EAS-scoring. etc. So, my advice is always not to look at the absolute EAS-points, but on the ranking of an EAS-ratinglist of all playing engines in the tournament/test/ratinglist. The most aggressive playing engine will always be on rank 1. But the EAS-scoring can differ a lot, depending on openings, opponents, timecontrol, overall strength-level of the engines...

[nmcrazyim5]

Thanks for the v4 release. I'm using it to analyze the world chess championship games. Interesting divergence in opening suggestions from SF and Torch (browser version) at several moments in the games I analyzed

[Hai]

Does the engine run on macOS?

[JManion]

bigger gambit book (posted and explained in this forum some time ago);
polyglot format

https://sourceforge.net/projects/chess- ... n/download

decent and sharp book for P4 (crazy/unsound gambits were removed
and less sound gambits have lower playing frequencies); includes many
relatively unknown (or new) gambits for White and Black

anyone knows where this trap.ctg book can be found?
meanwhile i found this,
https://fauziakram.gumroad.com/l/traps

but i don't think i'm going to spend 19 bucks on it

I saw that site, but the Fauzi traps book dl appears very small, so I do not think it's what we are looking for.