Rybka 4 is final. (select games for fun)

[michiguel]

[d]r1b1k2r/5pp1/p3pn1p/q2p4/1b3B2/2NRQ3/PPP1BPPP/2K4R b kq - 3 14

O-O is a humongous blunder by FB, allowing Bxh6. What happened?

Well, bear in mind these are all blitz games only, and the above was played single-CPU, with ponder on. There was no time to test anything slower. That said, here is FB's output with both cores enabled:

I know... but still... it is a move that even I would reject immediately, and my engine Gaviota never considers O-O and sees Bxh6 in 4 seconds. With ponder on may be even faster. FB here takes ~40 seconds to abandon O-O. It just caught my attention why a top engine would have problems with a move that apparently is no so difficult.

Miguel

Analysis by FireBird 1.2 x64:

14...Bd7 15.a3 Bxc3 16.Rxc3 0-0 17.Qd4 Ne4 18.Rc7 Bb5 19.Bxb5 axb5 20.f3 Nf6 21.Qc5 Rad8 22.Bd2 Qa4 23.Qb4 d4 24.Qxa4 bxa4
+/= (0.44) Depth: 15 00:00:05 15666kN
14...Bd7 15.Kb1 Be7 16.Qg3 Kf8 17.Be5 h5 18.Qg5 Rh6 19.Rg3 Rg6 20.Qf4 Kg8 21.Rxg6 fxg6 22.Rd1 Rc8
+/= (0.41) Depth: 15 00:00:05 16888kN
14...Bd7 15.Kb1 Be7 16.Qg3 Kf8 17.Be5 h5 18.Qg5 Rh6 19.Rg3 Rg6 20.Qf4 Kg8 21.Rxg6 fxg6 22.Rd1 Rc8
+/= (0.41) Depth: 15 00:00:05 16900kN
14...0-0 15.Bxh6 Ne8 16.Qg3 Bd6 17.f4 Rb8 18.Kb1 Bd7 19.b3 Kh7 20.Bg5 Kg8 21.Rhd1 Bc5 22.Ne4 Nd6 23.Nxc5 Qxc5
+/= (0.33) Depth: 16 00:00:07 22952kN
14...0-0 15.Bxh6 Ne8 16.Qg3 Bd6 17.f4 Rb8 18.Kb1 Bd7 19.b3 Kh7 20.Bg5 Kg8 21.Rhd1 Bc5 22.Ne4 Nd6 23.Nxc5 Qxc5
+/= (0.33) Depth: 16 00:00:07 22952kN
14...0-0 15.Bxh6 Ne8 16.Qg3 Bd6 17.Qh4 Qc7 18.Be3 Rb8 19.Bd4 Qb7 20.Na4 Be7 21.Qf4 Bd6 22.Qg5 Qc6 23.Nc3 f6
+/= (0.36) Depth: 16 00:00:07 25224kN
14...0-0 15.Bxh6 Ne8 16.Qg3 Bd6 17.Qh4 Qc7 18.Be3 Rb8 19.Bd4 Qb7 20.Na4 Be7 21.Qf4 Bd6 22.Qg5 Qc6 23.Nc3 f6
+/= (0.36) Depth: 16 00:00:08 26413kN
14...0-0 15.Bxh6 Ne8 16.Qg3 Bd6 17.Qh4 Qc7 18.Bg5 f6 19.Bd2 f5 20.Kb1 Nf6 21.Re1 Bb7 22.Rh3
+/= (0.44) Depth: 17 00:00:14 47229kN
14...0-0 15.Bxh6 Ne8 16.Qg3 Ra7 17.Be3 Rb7 18.a3 Bd6 19.f4 f5 20.Bd4 Nf6 21.Rdd1 Bd7 22.Rhf1 Bc6 23.h4 Ne4 24.Nxe4 fxe4
+/= (0.43) Depth: 17 00:00:15 50468kN
14...0-0 15.Bxh6 Ne8 16.Qg3 Ra7 17.Be3 Rb7 18.a3 Bd6 19.f4 f5 20.Bd4 Nf6 21.Rdd1 Bd7 22.Rhf1 Bc6 23.h4 Ne4 24.Nxe4 fxe4
+/= (0.43) Depth: 17 00:00:15 50483kN
14...Be7 15.Kb1 Bd7 16.Qg3 Kf8 17.Rhd1 h5 18.Bd6 Rh6 19.Bxe7+ Kxe7 20.Qg5 Rg6 21.Qe5 Kf8 22.Rg3 Rxg3 23.fxg3 Qb6 24.Bxh5 Rb8 25.b3
+/= (0.41) Depth: 18 00:00:42 143mN
14...Be7 15.Kb1 Bd7 16.Qg3 Kf8 17.Rhd1 h5 18.Bd6 Rh6 19.Bxe7+ Kxe7 20.Qg5 Rg6 21.Qe5 Kf8 22.Rg3 Rxg3 23.fxg3 Qb6 24.Bxh5 Rb8 25.b3
+/= (0.41) Depth: 18 00:00:42 143mN
14...Be7 15.Kb1 Bd7 16.Qg3 Kf8 17.Be5 h5 18.Qg5 Rh7 19.Rf3 Be8 20.Qf4 Rh6 21.Rg3 h4 22.Rh3 Nd7 23.Bd4 Rb8
+/= (0.42) Depth: 19 00:00:45 152mN
14...Be7 15.Kb1 Bd7 16.Qg3 Kf8 17.Be5 h5 18.Qg5 Rh7 19.Rf3 Be8 20.Qf4 Rh6 21.Rg3 h4 22.Rh3 Nd7 23.Bd4 Rb8
+/= (0.42) Depth: 19 00:00:47 161mN

[d]2rq1r1k/6bp/bp4p1/p2Pp3/P2pP3/1QP2N2/1P3NPP/2R1R1K1 b - - 0 24 bm Rxf3

Rxf3 is a nice move.

The first game was the typical game in which an engine plays so passive and waits to be killed by the other. Pretty amazing that some of the tactical monsters could still play the style (or lack thereof) of the early 90's engines.

Miguel

Yes, but it isn't really all that surprising. If anything, Pablo has show that well with his endless series of time wins.

Albert

[yanquis1972]

FB sees Bxh6 instantly, it just doesn't see the threat.

[S.Taylor]

Beta testing for Rybka 4 has ended, and it is looking very strong.

Didn't he just release betas, starting weakly, and adding little by little, until it was just about strong enough to get good reviews, and the rest of the strength to be saved for the online edition?

If so, I'm not all that impressed if it just about pulls through. (Even if it is "looking very strong", it might just be enough for his own good).

I have no idea what you are talking about. A new engine comes out, and rather than see how it plays, your main concern is how the beta testing went or how it might compare to some other project of the author? Good luck with that....

You said "it is looking very strong". "very strong", is relative, and can be defined according to what peoples expectations are.

And you replied with a bunch of comments on the beta testing, something about saving Elo for another project of his, and whatnot.

Personally, if I were debating buying it, I couldn't care less about any of that. I would care about how it played, and whether or not it was a sufficient leap over its predecessor to warrant my money.

If he has 300 Elo lying around to use at will, all the power to him, but it has no affect on my decision.

For me, if it plays well, and it is a clear 150 elo leap over R3, then I would be happy enough about buying it, even if he has a +300 elo one lying around for other purposses.

I don't feel so good about it if it is only +80 elo, because I might be able to get better in another couple of months.

If the beta testing was to see how low he could get the elo to, then I would willingly grant him down to +150 over R3.

[michiguel]

FB sees Bxh6 instantly, it just doesn't see the threat.

Bxh6 is in the PV, but the score is not that high. It means it sees Bxh6, but it does not think is that good. I wonder whether this type of behavior is caused by some aggressive pruning technique.

Miguel

[Albert Silver]

Beta testing for Rybka 4 has ended, and it is looking very strong.

Didn't he just release betas, starting weakly, and adding little by little, until it was just about strong enough to get good reviews, and the rest of the strength to be saved for the online edition?

If so, I'm not all that impressed if it just about pulls through. (Even if it is "looking very strong", it might just be enough for his own good).

I have no idea what you are talking about. A new engine comes out, and rather than see how it plays, your main concern is how the beta testing went or how it might compare to some other project of the author? Good luck with that....

You said "it is looking very strong". "very strong", is relative, and can be defined according to what peoples expectations are.

And you replied with a bunch of comments on the beta testing, something about saving Elo for another project of his, and whatnot.

Personally, if I were debating buying it, I couldn't care less about any of that. I would care about how it played, and whether or not it was a sufficient leap over its predecessor to warrant my money.

If he has 300 Elo lying around to use at will, all the power to him, but it has no affect on my decision.

For me, if it plays well, and it is a clear 150 elo leap over R3, then I would be happy enough about buying it, even if he has a +300 elo one lying around for other purposses.

I don't feel so good about it if it is only +80 elo, because I might be able to get better in another couple of months.

If the beta testing was to see how low he could get the elo to, then I would willingly grant him down to +150 over R3.

Unfortunately, that is always true, and of SO many things. Ex: I can buy this 3.2GHz CPU now for $200 or wait a few months and get a faster one for that much. I could buy this new video card that plays Crysis at 40 FPS (frames per second) for $150, or I could wait 3-4 months and get one that plays it at 46 FPS, etc.

I'll break it to you right now: in this sort of thing, forget about tomorrow, because this quandary will NEVER end. If the level of play is good for you TODAY, and you can afford it of course, go for it.

In any case, if you can afford it, I'd go for one of the packages, either by CB or by ChessOk, depending on your interest. Both have their perks, and bring far more to the table than just the engine.

[S.Taylor]

In any case, if you can afford it, I'd go for one of the packages, either by CB or by ChessOk, depending on your interest. Both have their perks, and bring far more to the table than just the engine.

Thanks for pointing out the 2 good options. I will try to weigh up the pros and cons.
And, of course, in the end, everyone should do what is best for him. Money seems to mean quite alot, too. I think that there are people who feel the business world is corrupt. Like when prices just go up and up and up, because in the mean time they can get it.

[bob]

[d]r1b1k2r/5pp1/p3pn1p/q2p4/1b3B2/2NRQ3/PPP1BPPP/2K4R b kq - 3 14

O-O is a humongous blunder by FB, allowing Bxh6. What happened?

Well, bear in mind these are all blitz games only, and the above was played single-CPU, with ponder on. There was no time to test anything slower. That said, here is FB's output with both cores enabled:

I know... but still... it is a move that even I would reject immediately, and my engine Gaviota never considers O-O and sees Bxh6 in 4 seconds. With ponder on may be even faster. FB here takes ~40 seconds to abandon O-O. It just caught my attention why a top engine would have problems with a move that apparently is no so difficult.

Miguel

mainly depends on king safety scores as to how deep one has to search to reject o-o...

Analysis by FireBird 1.2 x64:

14...Bd7 15.a3 Bxc3 16.Rxc3 0-0 17.Qd4 Ne4 18.Rc7 Bb5 19.Bxb5 axb5 20.f3 Nf6 21.Qc5 Rad8 22.Bd2 Qa4 23.Qb4 d4 24.Qxa4 bxa4
+/= (0.44) Depth: 15 00:00:05 15666kN
14...Bd7 15.Kb1 Be7 16.Qg3 Kf8 17.Be5 h5 18.Qg5 Rh6 19.Rg3 Rg6 20.Qf4 Kg8 21.Rxg6 fxg6 22.Rd1 Rc8
+/= (0.41) Depth: 15 00:00:05 16888kN
14...Bd7 15.Kb1 Be7 16.Qg3 Kf8 17.Be5 h5 18.Qg5 Rh6 19.Rg3 Rg6 20.Qf4 Kg8 21.Rxg6 fxg6 22.Rd1 Rc8
+/= (0.41) Depth: 15 00:00:05 16900kN
14...0-0 15.Bxh6 Ne8 16.Qg3 Bd6 17.f4 Rb8 18.Kb1 Bd7 19.b3 Kh7 20.Bg5 Kg8 21.Rhd1 Bc5 22.Ne4 Nd6 23.Nxc5 Qxc5
+/= (0.33) Depth: 16 00:00:07 22952kN
14...0-0 15.Bxh6 Ne8 16.Qg3 Bd6 17.f4 Rb8 18.Kb1 Bd7 19.b3 Kh7 20.Bg5 Kg8 21.Rhd1 Bc5 22.Ne4 Nd6 23.Nxc5 Qxc5
+/= (0.33) Depth: 16 00:00:07 22952kN
14...0-0 15.Bxh6 Ne8 16.Qg3 Bd6 17.Qh4 Qc7 18.Be3 Rb8 19.Bd4 Qb7 20.Na4 Be7 21.Qf4 Bd6 22.Qg5 Qc6 23.Nc3 f6
+/= (0.36) Depth: 16 00:00:07 25224kN
14...0-0 15.Bxh6 Ne8 16.Qg3 Bd6 17.Qh4 Qc7 18.Be3 Rb8 19.Bd4 Qb7 20.Na4 Be7 21.Qf4 Bd6 22.Qg5 Qc6 23.Nc3 f6
+/= (0.36) Depth: 16 00:00:08 26413kN
14...0-0 15.Bxh6 Ne8 16.Qg3 Bd6 17.Qh4 Qc7 18.Bg5 f6 19.Bd2 f5 20.Kb1 Nf6 21.Re1 Bb7 22.Rh3
+/= (0.44) Depth: 17 00:00:14 47229kN
14...0-0 15.Bxh6 Ne8 16.Qg3 Ra7 17.Be3 Rb7 18.a3 Bd6 19.f4 f5 20.Bd4 Nf6 21.Rdd1 Bd7 22.Rhf1 Bc6 23.h4 Ne4 24.Nxe4 fxe4
+/= (0.43) Depth: 17 00:00:15 50468kN
14...0-0 15.Bxh6 Ne8 16.Qg3 Ra7 17.Be3 Rb7 18.a3 Bd6 19.f4 f5 20.Bd4 Nf6 21.Rdd1 Bd7 22.Rhf1 Bc6 23.h4 Ne4 24.Nxe4 fxe4
+/= (0.43) Depth: 17 00:00:15 50483kN
14...Be7 15.Kb1 Bd7 16.Qg3 Kf8 17.Rhd1 h5 18.Bd6 Rh6 19.Bxe7+ Kxe7 20.Qg5 Rg6 21.Qe5 Kf8 22.Rg3 Rxg3 23.fxg3 Qb6 24.Bxh5 Rb8 25.b3
+/= (0.41) Depth: 18 00:00:42 143mN
14...Be7 15.Kb1 Bd7 16.Qg3 Kf8 17.Rhd1 h5 18.Bd6 Rh6 19.Bxe7+ Kxe7 20.Qg5 Rg6 21.Qe5 Kf8 22.Rg3 Rxg3 23.fxg3 Qb6 24.Bxh5 Rb8 25.b3
+/= (0.41) Depth: 18 00:00:42 143mN
14...Be7 15.Kb1 Bd7 16.Qg3 Kf8 17.Be5 h5 18.Qg5 Rh7 19.Rf3 Be8 20.Qf4 Rh6 21.Rg3 h4 22.Rh3 Nd7 23.Bd4 Rb8
+/= (0.42) Depth: 19 00:00:45 152mN
14...Be7 15.Kb1 Bd7 16.Qg3 Kf8 17.Be5 h5 18.Qg5 Rh7 19.Rf3 Be8 20.Qf4 Rh6 21.Rg3 h4 22.Rh3 Nd7 23.Bd4 Rb8
+/= (0.42) Depth: 19 00:00:47 161mN

[d]2rq1r1k/6bp/bp4p1/p2Pp3/P2pP3/1QP2N2/1P3NPP/2R1R1K1 b - - 0 24 bm Rxf3

Rxf3 is a nice move.

The first game was the typical game in which an engine plays so passive and waits to be killed by the other. Pretty amazing that some of the tactical monsters could still play the style (or lack thereof) of the early 90's engines.

Miguel

Yes, but it isn't really all that surprising. If anything, Pablo has show that well with his endless series of time wins.

Albert

[djbl]

seems very odd to me why they would want to beta test r4 against 'so-called clones'??

[Nimzovik]

Of course they must test against the other clones. Why test against R3? Is not R 4 a clone of R3?

[djbl]

well, i think i'll wait for the next stockfish and firebird before i go blowing 60-70 quid. maybe now though vas can finally admit that all this clone talk was merely a ruse to gain a bit of time. i think r4 was all ready to go late last year but the strength of the new engines put paid to that, and since then he has been trying to claw back ground on engines that he should have been able to lick, being that they were merely (allegedly) poor imitations of his own masterpiece n'all. and it does seem strange how chessbase lackies who have so discredited these other engines and refused all others to use them have chosen to test their own engines mostly against them. rank hypocrisy if you ask me. they accuse, forebid others to use them, and yet they make use of them themselves for testing. talk about double standards.