Houdini 1.5 is released

[M ANSARI]

I agree, please keep posts on subject of initial post. If you would like to discuss other matters then create another post.

[FlavusSnow]

Actually I just consider Houdini one of the many Rybka 3 derivatives and regularly test Rybka 4 against it ...

But aren't they all just Fruit derivatives and forks?

[FlavusSnow]

First: Nice Work.

Any chance we can get a 64-bit Linux compile?

How about Chess960 / Fischer Random support?

[Dann Corbit]

Feature request for Houdini 1.6:
When there is a tablebase mate, list the full pv to the end.

E.g.
+M8 Nxb3

does not show me the full idea.

You can easily collect it on the last iteration by just making the suggested moves and calling the EGTB again until we have exhausted the pv.

Dann, Houdini 1.5 should produce the full mate line when the root position is an EGTB mate, it already does exactly what you suggest.
Can you give me an example position in which it doesn't?

It may have been the GUI that did it, or perphaps it was a fail high when time ran out. I will have to look into it to find an example. I was just watching analysis and saw it happen once.

[JManion]

In CB 10, when I open a new game, and turn on Houdini 1.5 x64 for infinite analysis, the engine and the program will freeze every few moves.

I will be unable to move any pieces, and the depth and node counts will freeze.

I have tried this on a few different computers XP, Vista, and 7. All of them running 64 bit OS.

In Chessbase 9 and 10 I am having the freezing issue. However in DF12 I do not seem to have the issue, so I assume it has something to do with the GUI.

[yanquis1972]

will houdini crack the 3000 barrier on ipon?

http://www.inwoba.de/

still almost 700 games to go, so doubful & it will probably regress to robert's estimate. but pretty interesting.

[yanquis1972]

quickly running through your engine & while it seems to be an improvement, it also (as far as i can tell) still lacks the 'creativity' that it shares with its lineage (i don't believe any of ippolit, crafty or stockfish are particularly good speculative engines). do you have any plans to address this and/or have you attempted to do so with 1.5?

Houdini 1.5 is more aggressive and more tactical than version 1.03, but it's still based on a very solid positional game.
Most of the human World Champions would fit that description, so it's not a disgrace .

i have noticed the tactical improvements, but i was curious if perhaps you're considering any emphasis on king safety or other dynamic factors. i have begun using rybka 3 dynamic again lately because it will often suggest moves & evaluations which are more humanlike (& ultimately accurate when tested against other engines). the cost of this was very few elo in comparison to rybka 3 & i think vas hardly worked on it at all. it would seem to me that if done right & with emphasis it could not only add elo, but make for a -better- all around engine for analysis purposes.

obviously not a criticism & congrats on releasing what ostensibly appears to be the strongest engine in the world. i just notice that seems to be one area for improvement, partic. now that you have that crown.

[yanquis1972]

here is an example of what i am talking. i always use this position to check an engines understanding of dynamic compensation.
[d]r1b2rk1/pp1ppp1p/5bp1/q7/3nP2Q/1BN1B3/PPP2PPP/R4RK1 w - -

supernehz of course played Qxf6!!?. i believe this move draws.


houdini performs about in line with other top programs, with an evaluation that slowly increases for the move from a pawn plus down to about half a pawn in this case after 30 minutes of thinking on 8CPU in 6mpv mode.

[+0.17] d=28 12.Qh6 Bg7 13.Qg5 Qxg5 14.Bxg5 e6 15.Rfd1 f5 16.Be3 Nc6 17.exf5 Rxf5 18.Rab1 Be5 19.Ne2 Rf7 20.Nd4 Nxd4 21.Bxd4 Bc7 22.Bc5 Be5 23.Bd6 Bf6 24.c3 (0:21:31) 21490463kN
[+0.09] d=27 12.Qg4 d6 13.Qd1 Nc6 14.Rb1 Bg7 15.Qe1 b5 16.Nd5 Qxe1 17.Rfxe1 e6 18.Nc7 Rb8 19.a4 a6 20.Red1 Na5 21.axb5 axb5 22.Ba2 b4 23.Bf4 b3 24.cxb3 (0:17:42) 17540593kN
[-0.13] d=27 12.Qh3 d5 13.Qh6 dxe4 14.Bxd4 Bxd4 15.Nxe4 Bg7 16.Qh4 Qc7 17.Rfe1 Bxb2 18.Rad1 Bf5 19.Ng5 h5 20.Nxf7 Bf6 21.Ne5 Kg7 22.Qf4 Qc3 23.Rd2 Rad8 24.Rde2 e6 25.h3 b5 26.Qc1 (0:18:15) 18101854kN
[-0.53] d=27 12.Qxf6 Ne2 13.Nxe2 exf6 14.Nc3 Qd8 15.Nd5 Re8 16.Bd4 Re6 17.f4 Rc6 18.e5 f5 19.Rfe1 h6 20.c3 Re6 21.Nf6 Kh8 22.Bd5 Qa5 23.b4 Qa3 24.Re3 Kg7 25.Bc5 Qb2 (0:20:57) 20896789kN
[-0.60] d=27 12.Qf4 Qxc3 13.bxc3 Ne2 14.Kh1 Nxf4 15.Bxf4 d6 16.Be3 Bxc3 17.Rab1 b6 18.Bg5 Re8 19.Ba4 Be6 20.Bxe8 Rxe8 21.Rfd1 Bxa2 22.Ra1 Bxa1 23.Rxa1 f6 24.Be3 Be6 25.Rxa7 Kf7 26.Bf4 Rd8 27.h4 d5 28.e5 (0:20:57) 20896789kN
[-0.73] d=27 12.Qg3 Qxc3 13.bxc3 Ne2 14.Kh1 Nxg3 15.fxg3 Bxc3 16.Rad1 Kg7 17.Rd3 Be5 18.Bf4 Bxf4 19.gxf4 Rb8 20.e5 b5 21.Rc3 f6 22.exf6 Kxf6 23.Rd1 Kg7 24.g3 Bb7 25.Bd5 Bxd5 26.Rxd5 (0:20:57) 20896789kN

rybka 4 virtually instantly understands there is massive dynamic compensation with an invaluation in the seconds mark of ~-.28. i'll stop the analysis here after just a few minutes --

[+0.18] d=21 12.Qh6 Bg7 13.Qg5 Qxg5 14.Bxg5 e6 15.f4 Nxb3 16.cxb3 Bd4 17.Kh1 d5 18.exd5 f6 19.Bh4 Bxc3 20.bxc3 exd5 21.Rfe1 Kf7 22.Rad1 Be6 23.Bf2 Rfe8 24.h3 Rac8 25.Bd4 b6 (0:04:39) 216226kN
[+0.13] d=21 12.Qg4 d6 13.Qd1 Nc6 14.Qd2 Bg7 15.Rad1 Bg4 16.f3 Bxc3 17.bxc3 Be6 18.f4 Rfc8 19.f5 gxf5 20.exf5 Bxf5 21.Qf2 (0:05:29) 249439kN
[-0.18] d=21 12.Qxf6 Ne2 13.Nxe2 exf6 14.Nc3 Re8 15.Bd4 Kg7 16.Nd5 Re6 17.Rfe1 d6 18.Rad1 Re5 19.Bc3 Qd8 20.f4 Rxd5 21.Bxd5 Be6 22.Bxe6 fxe6 23.e5 Qb6 24.Bd4 Qa5 25.exf6 Kf7 26.a3 Qf5 27.g3 (0:05:30) 250132kN
[-0.26] d=20 12.Qh3 d5 13.Qh6 dxe4 14.Bxd4 Bxd4 15.Nxe4 Bg7 16.Qh4 Bxb2 17.Rab1 Ba3 18.Bc4 a6 19.Rfd1 Kg7 20.Ng5 h5 21.Qd4 e5 22.Qe4 Qc7 23.Bd5 Be7 24.Bxb7 Bf5 25.Qd5 Bxc2 (0:04:24) 205797kN\

(this was also mpv 6, but as we all know R4 is buggy to the point of worthless with mpv)

won't clog the post up w/ more output, but rybka 3 dynamic instantly sees full compensation and even gives the move a slight plus. this no doubt comes at some sacrifice to elo, but i was absolutely astounded at the intelligence of the program when i first saw it. i think vas was right to focus on king safety in R4 & i believe it's still a step a lot of the other top free engines haven't focused on, for whatever reason.

[Houdini]

i have noticed the tactical improvements, but i was curious if perhaps you're considering any emphasis on king safety or other dynamic factors. i have begun using rybka 3 dynamic again lately because it will often suggest moves & evaluations which are more humanlike (& ultimately accurate when tested against other engines). the cost of this was very few elo in comparison to rybka 3 & i think vas hardly worked on it at all. it would seem to me that if done right & with emphasis it could not only add elo, but make for a -better- all around engine for analysis purposes.

obviously not a criticism & congrats on releasing what ostensibly appears to be the strongest engine in the world. i just notice that seems to be one area for improvement, partic. now that you have that crown.

I actually agree with you that king safety is certainly not Houdini's strongest point. And there are many other areas where improvements are possible, it's a never-ending work in progress...

[yanquis1972]

i have noticed the tactical improvements, but i was curious if perhaps you're considering any emphasis on king safety or other dynamic factors. i have begun using rybka 3 dynamic again lately because it will often suggest moves & evaluations which are more humanlike (& ultimately accurate when tested against other engines). the cost of this was very few elo in comparison to rybka 3 & i think vas hardly worked on it at all. it would seem to me that if done right & with emphasis it could not only add elo, but make for a -better- all around engine for analysis purposes.

obviously not a criticism & congrats on releasing what ostensibly appears to be the strongest engine in the world. i just notice that seems to be one area for improvement, partic. now that you have that crown.

I actually agree with you that king safety is certainly not Houdini's strongest point. And there are many other areas where improvements are possible, it's a never-ending work in progress...

down to 500+ games on IPON, only one point drop in elo...seems you may have underestimated the strength improvement. w/ a 50 elo lead on the (commercial) competition...might be a nice time to experiment.

ok i'm done on the subject. promise. for now.