Which was the latest version of Rybka?

[towforce]

...Shredder 12, the champion before [Rybka] also had problems with the king's safety and, after the opening phase, played in a manner that was anything but aggressive.

Sounds as though playing positionally well in the middle game counts for more than a strong end game: by the time you get to the end game, it's already over.

This would also explain Uri's complaint that Stockfish isn't good at quickly killing the opponent in won positions.

[Eelco de Groot]

...Shredder 12, the champion before [Rybka] also had problems with the king's safety and, after the opening phase, played in a manner that was anything but aggressive.

Sounds as though playing positionally well in the middle game counts for more than a strong end game: by the time you get to the end game, it's already over.

This would also explain Uri's complaint that Stockfish isn't good at quickly killing the opponent in won positions.

You can get in hot water here, but for alternative explanation you just have to read what Fabien Letouzey wrote, only very recently here but it was known of course, about the attacking power of his own engine. Yes, endgame technique was more of a strength I think he would say himself. A bit surprising though that this still was not 'attacked' by Vas himself in Rybka 4, or even version three when a lot was rewritten. Apparently King safety was stil under development then.

[Frank Quisinsky]

Code: Select all

You wrote:
This would also explain Uri's complaint that Stockfish isn't good at quickly killing the opponent in won positions.

From that perspective, I have never looked into it in any detail.
Thats interesting!

I wonder that Reckless hold +- the Elo strength from Stockfish, but is able to win clearly more fast games.
Move-average is lesser!

Really interesting are PlentyChess, Integral, Tarnished.
Move average is super, nice to see that in the late mid-games ... clearly draw games ended faster.
In my opinion absolutely correct to wait of the late-midgame. Explain very nice the move-average from this three interesting programs.

But the topic is what Uri find out.
Hm, room for experiments, thanks Uri!

[Frank Quisinsky]

Code: Select all

You can get in hot water here, but for alternative explanation you just have to read what Fabien Letouzey wrote, only very recently here but it was known of course, about the attacking power of his own engine. Yes, endgame technique was more of a strength I think he would say himself. A bit surprising though that this still was not 'attacked' by Vas himself in Rybka 4, or even version three when a lot was rewritten. Apparently King safety was stil under development then.

Also, not nice the king-safety from Senpai 2.
Fabien wrote me that he forced more the pawns for more open positions, made a bit on king safety.
Clearly better as in Senpai 2 now.

Also Fruit, was never a great attacker.

Whats funny is:
The great attacker: Patricia, Wasp or the CSTal, Revenge, Pedone 3 etc. aren't able to win vs. stronger NN engines games very fast.
But more fast wins vs. HCE engines or "weaker" NN engines as most of the TOP-Engines today. How can this be explained?

I can't seem to figure out a good explanation.

[syzygy]

Thats right but ...

Shredder 12, the champion before also had problems with the king's safety and, after the opening phase, played in a manner that was anything but aggressive. Back then, it was harder to figure out which programs played aggressively and didn’t have those problems. Correspondence chess players, in particular, were constantly asking because they were totally dissatisfied with Shredder and Rybka (first playing phase). I worked for a chess magazine and was constantly bombarded with these questions. Correspondence chess players aren't stupid; they spot fundamental problems right away. Of course, most people in the computer chess community unfortunately only care about Elo. They want the fastest car, no matter how terrible it looks ... that's really bad, in my opinion.

But the problems correspondence players found will have been related to Elo, too. Their games were just at a high enough level for Rybka's weakness to become noticeable. But if you mean "engine to engine"/rating list Elo, then yes, I see what you mean.

[Frank Quisinsky]

I agree!

Corr. player asked ... what can I use for the early mid-games.

I provided the information, but the problem was ...
If engines like Spark (the time around 2010), Junior, Hiarcs, Texel different others are able to play more fast wins ...

Only when attacks fail do they often leave behind very poor pawn structures, and this tends to lead to defeat rather than victory. Aggressive pawn moves against the king's position leave their mark if the attack fails ... I mean this constellation is more often.

Computer chess is anything but simple. Writing about playing styles has always been a double-edged sword.

Back to Rybka:
Mid-games in three phases.
Phase 1 = 2/5
Phase 2 = 1/5 ... that is the point from Rybka ... all the loses between move number 45-55 ... terrible!
Phase 3 = 4/5 ... before endgame started

Shredder 12:
Phase 1 = 1/5 ... more very fast loses
Phase 2 = 2/5
Phase 3 = 4/5

Small differences between Shredder and Rybka but fact is ...
Both engines are not good after opening book moves, so the correspondence player tell me often ... in the mid-games a strong GM can make the points vs. Rybka, Shredder ... not in endgames.

I often wonder that Chessbase offer Fritz 15 / Fritz 16 with Rybka.
Because they have so many experts ... but offer such an engine.
And that years later ...