Who is who ?

beram · Post by **beram** » Sun Jun 22, 2014 10:45 pm

For those who believe in engine playing styles I want to challenge them to tell "who is who" in this recent LTC game between Komodo and Houdini
And of course with arguments about style of play

[pgn]
[Event "T"]
[Site "C"]
[Date "2014.06.07"]
[White "X"]
[Black "Y"]
[Result "0-1"]
[ECO "D30"]
[TimeControl "CCRL40/40"]

1.d4 d5 2.c4 e6 3.Nf3 Nf6 4.Bg5 Be7 5.e3 h6 6.Bh4 O-O 7.Nc3 b6 8.Be2 Bb7 9.O-O Nbd7 10.Bg3 Ne4 11.cxd5 Nxg3 12.hxg3 exd5 13.Rc1 Nf6 14.Qb3 c6 15.Rfd1 Bd6 16.Bd3 Re8 17.Re1 Re7 18.Qc2 Rb8 19.Nh4 Re8 20.e4 dxe4 21.Nxe4 Nxe4 22.Bxe4 c5 23.Bxb7 Rxe1+ 24.Rxe1 Rxb7 25.Nf5 Bf8 26.dxc5 Bxc5 27.Qe2 Kh7 28.Qe4 Rd7 29.b4 Bf8 30.Ne7+ g6 31.Nc6 Qf6 32.a3 Rc7 33.Ne5 a5 34.Rd1 Qe6 35.Qf4 Bg7 36.Re1 a4 37.Nf3 Qc4 38.Qxc4 Rxc4 39.Re7 Rc3 40.Rxf7 Rxa3 41.Rb7 Rb3 42.Rxb6 g5 43.Kf1 Rb1+ 44.Ke2 Rb2+ 45.Ke3 a3 46.Ra6 a2 47.Nd2 Rxd2 48.Kxd2 a1=Q 49.Rxa1 Bxa1 50.Kd3 Kg6 51.b5 Kf5 52.b6 Be5 53.Ke3 Bc3 54.b7 Be5 55.Kd3 Bb8 56.Ke3 Ba7+ 57.Kd2 Ke6 58.Kc3 h5 59.Kd3 Kd5 60.Ke2 Kc6 61.Kd3 Kxb7 62.f4 g4 63.Ke4 0-1
[/pgn]

Graham Banks · Post by **Graham Banks** » Sun Jun 22, 2014 11:40 pm

LOL - I have absolutely no idea, but I'll hazard a guess and say that Komodo was playing black.

Based on:
- early fianchetto of the bishop
- endgame play on the queenside

However, even though I've watched hundreds of games play through from both engines, I'm still not 100% certain on this particular game.

BeyondCritics · Post by **BeyondCritics** » Mon Jun 23, 2014 12:05 am

beram wrote:For those who believe in engine playing styles I want to challenge them to tell "who is who" in this recent LTC game between Komodo and Houdini
And of course with arguments about style of play

[pgn]
[Event "T"]
[Site "C"]
[Date "2014.06.07"]
[White "X"]
[Black "Y"]
[Result "0-1"]
[ECO "D30"]
[TimeControl "CCRL40/40"]

1.d4 d5 2.c4 e6 3.Nf3 Nf6 4.Bg5 Be7 5.e3 h6 6.Bh4 O-O 7.Nc3 b6 8.Be2 Bb7 9.O-O Nbd7 10.Bg3 Ne4 11.cxd5 Nxg3 12.hxg3 exd5 13.Rc1 Nf6 14.Qb3 c6 15.Rfd1 Bd6 16.Bd3 Re8 17.Re1 Re7 18.Qc2 Rb8 19.Nh4 Re8 20.e4 dxe4 21.Nxe4 Nxe4 22.Bxe4 c5 23.Bxb7 Rxe1+ 24.Rxe1 Rxb7 25.Nf5 Bf8 26.dxc5 Bxc5 27.Qe2 Kh7 28.Qe4 Rd7 29.b4 Bf8 30.Ne7+ g6 31.Nc6 Qf6 32.a3 Rc7 33.Ne5 a5 34.Rd1 Qe6 35.Qf4 Bg7 36.Re1 a4 37.Nf3 Qc4 38.Qxc4 Rxc4 39.Re7 Rc3 40.Rxf7 Rxa3 41.Rb7 Rb3 42.Rxb6 g5 43.Kf1 Rb1+ 44.Ke2 Rb2+ 45.Ke3 a3 46.Ra6 a2 47.Nd2 Rxd2 48.Kxd2 a1=Q 49.Rxa1 Bxa1 50.Kd3 Kg6 51.b5 Kf5 52.b6 Be5 53.Ke3 Bc3 54.b7 Be5 55.Kd3 Bb8 56.Ke3 Ba7+ 57.Kd2 Ke6 58.Kc3 h5 59.Kd3 Kd5 60.Ke2 Kc6 61.Kd3 Kxb7 62.f4 g4 63.Ke4 0-1
[/pgn]

I took your challenge and clicked through the game.

I think 32.a3?! is a serious positionell concession, not to say a blunder. No wonder that white later got into difficulties.
[d]5b2/p2r1p1k/1pN2qpp/8/1P2Q3/6P1/P4PP1/4R1K1 w - - 2 32

In the last position white has serious drawing chances, since black has the wrong coloured bishop.
[d]8/bk6/8/7p/4KPp1/6P1/6P1/8 b - - 1 63

Taking a closer look: 63..Bf2 64.Kf5 Bxg3 65.Kg5 h4 66.Kxg4 ... Now white wants to move the king to h1 with a draw. Black cannot spoil this plan. E.g. 66...Be1 67.Kh3. Or 66...Kc6 67.Kf3 Kd5 68.Ke2! Bxf4 69.Kf2 Bh2 70.g4! h3 71.g5 with a draw. Or 70..hxg3 Kg2 with a "known" forttress:
[d]8/8/8/3k4/8/6p1/6Kb/8 b - - 1 71
Other ideas do not work either for black.
Therefore the last position is a draw and the game was wrongly adjudicated, seemingly both engines "believed" that black is winning.

Now to your question:
I know nothing about the playing styles of houdini or komodo. But i know that komode has a gm as team member and houdini has not. Therefore i guess wildly, that komodo hat black and houdini had white. Since a move like 32.a3 is like a slap in the face, either very good or very bad. Most likely very bad. A gm does feel that errors.

jhellis3 · Post by **jhellis3** » Mon Jun 23, 2014 12:14 am

I'll guess:

W - Komodo

B - Houdini

Lyudmil Tsvetkov · Post by **Lyudmil Tsvetkov** » Mon Jun 23, 2014 1:02 am

beram wrote:For those who believe in engine playing styles I want to challenge them to tell "who is who" in this recent LTC game between Komodo and Houdini
And of course with arguments about style of play

[pgn]
[Event "T"]
[Site "C"]
[Date "2014.06.07"]
[White "X"]
[Black "Y"]
[Result "0-1"]
[ECO "D30"]
[TimeControl "CCRL40/40"]

1.d4 d5 2.c4 e6 3.Nf3 Nf6 4.Bg5 Be7 5.e3 h6 6.Bh4 O-O 7.Nc3 b6 8.Be2 Bb7 9.O-O Nbd7 10.Bg3 Ne4 11.cxd5 Nxg3 12.hxg3 exd5 13.Rc1 Nf6 14.Qb3 c6 15.Rfd1 Bd6 16.Bd3 Re8 17.Re1 Re7 18.Qc2 Rb8 19.Nh4 Re8 20.e4 dxe4 21.Nxe4 Nxe4 22.Bxe4 c5 23.Bxb7 Rxe1+ 24.Rxe1 Rxb7 25.Nf5 Bf8 26.dxc5 Bxc5 27.Qe2 Kh7 28.Qe4 Rd7 29.b4 Bf8 30.Ne7+ g6 31.Nc6 Qf6 32.a3 Rc7 33.Ne5 a5 34.Rd1 Qe6 35.Qf4 Bg7 36.Re1 a4 37.Nf3 Qc4 38.Qxc4 Rxc4 39.Re7 Rc3 40.Rxf7 Rxa3 41.Rb7 Rb3 42.Rxb6 g5 43.Kf1 Rb1+ 44.Ke2 Rb2+ 45.Ke3 a3 46.Ra6 a2 47.Nd2 Rxd2 48.Kxd2 a1=Q 49.Rxa1 Bxa1 50.Kd3 Kg6 51.b5 Kf5 52.b6 Be5 53.Ke3 Bc3 54.b7 Be5 55.Kd3 Bb8 56.Ke3 Ba7+ 57.Kd2 Ke6 58.Kc3 h5 59.Kd3 Kd5 60.Ke2 Kc6 61.Kd3 Kxb7 62.f4 g4 63.Ke4 0-1
[/pgn]

[d]5b2/p2r1p1k/1pN2qpp/8/1P2Q3/6P1/P4PP1/4R1K1 w - - 2 32

The doubled g3 pawn part of the king shelter and the preference for knights over bishops tells the whole story: white is Komodo and black Houdini. (nedded just one second for the answer after I saw the above position

, any rewards here?)

But, also very telling is the fact that you always like to see Houdini winning.

gerold · Post by **gerold** » Mon Jun 23, 2014 1:31 am

jhellis3 wrote:I'll guess:

W - Komodo

B - Houdini

I agree. Plus one.

Graham Banks · Post by **Graham Banks** » Mon Jun 23, 2014 10:52 am

Lyudmil Tsvetkov wrote:........the preference for knights over bishops tells the whole story: white is Komodo and black Houdini........

I considered this too.
Looking forward to finding out the answer.

Lyudmil Tsvetkov · Post by **Lyudmil Tsvetkov** » Mon Jun 23, 2014 11:25 am

BeyondCritics wrote:In the last position white has serious drawing chances, since black has the wrong coloured bishop.
[d]8/bk6/8/7p/4KPp1/6P1/6P1/8 b - - 1 63

Taking a closer look: 63..Bf2 64.Kf5 Bxg3 65.Kg5 h4 66.Kxg4 ... Now white wants to move the king to h1 with a draw. Black cannot spoil this plan. E.g. 66...Be1 67.Kh3. Or 66...Kc6 67.Kf3 Kd5 68.Ke2! Bxf4 69.Kf2 Bh2 70.g4! h3 71.g5 with a draw. Or 70..hxg3 Kg2 with a "known" forttress:
[d]8/8/8/3k4/8/6p1/6Kb/8 b - - 1 71
Other ideas do not work either for black.
Therefore the last position is a draw and the game was wrongly adjudicated, seemingly both engines "believed" that black is winning.

Black should win easily.
Somehow you do not consider the simple 63...Kc6 (instead of Bf2) 64.Kf5 Kd5 65. Kg5 Ke4 66. Kh5 Kf5 and black wins.

[d]8/b7/8/5k1K/5Pp1/6P1/6P1/8 w - - 0 58

Ajedrecista · Post by **Ajedrecista** » Mon Jun 23, 2014 1:16 pm

Hello:

[d]8/bk6/8/7p/4KPp1/6P1/6P1/8 b - - 1 63

Houdini started to give a probably winning eval for 63.- ..., Bf2:

Code: Select all

 15/32	00&#58;01	     314.794	765.000	+1,99	Ba7f2 Ke4f5 Bf2xg3 Kf5g5 h5h4 Kg5xg4 Kb7c6 f4f5 Kc6d7 Kg4g5 Kd7e7 Kg5g6 Bg3f2 Kg6g5 Ke7f7 Kg5f4 Kf7f6 Kf4f3 Bf2g3 Kf3g4 Bg3e1 Kg4f4 Be1f2 Kf4f3

Then it thinks that this move draws:

Code: Select all

 23/45	00&#58;21	  34.301.814	1.661.000	+0,02	Ba7f2 Ke4f5 Bf2xg3 Kf5g5 h5h4 Kg5xg4 Kb7c6 f4f5 Kc6d7 Kg4g5 Kd7e7 Kg5g6 Bg3e1 f5f6+ Ke7f8 f6f7 Be1g3 Kg6f6 Bg3f2 Kf6g6 Bf2e1 Kg6f6 Be1g3 Kf6g6 Bg3f2 Kg6f6 Bf2e3 Kf6g6 Be3b6 Kg6g5 Bb6d8+ Kg5g4 Kf8xf7 g2g3 Kf7g7 g3xh4

And later finds 63.- ..., Kc6 as the winning move for black:

Code: Select all

 23/51	00&#58;31	  49.207.351	1.605.000	+6,04	Kb7c6 Ke4f5 Kc6d5 Kf5g6 Kd5e4 f4f5 Ba7f2 f5f6 Bf2xg3 Kg6xh5 Ke4f5 f6f7 Bg3d6 Kh5h4 Bd6e7+ Kh4g3 Be7c5 Kg3h4 Bc5d6 Kh4h5 Bd6e7 g2g3 Be7d6 Kh5h6 Bd6f8+ Kh6h5 Bf8e7 Kh5h6 Kf5e4 Kh6g6 Ke4f3 Kg6f5 Kf3xg3 Kf5e5 Be7f8 Ke5e6 Kg3h4 Ke6d7 g4g3 Kd7e8 g3g2 Ke8xf8 g2g1Q Kf8e7 Qg1c5+ Ke7e8 Qc5e5+ Ke8d7
&#91;...&#93;
 36/88	45&#58;34	4.681.365.835	1.713.000	+M40	Kb7c6 Ke4d3 Kc6c5 Kd3e2 Kc5d5 Ke2f1 Kd5e4 Kf1e1 Ke4f5 Ke1f1 Ba7c5 Kf1e1 Bc5g1 Ke1e2 Bg1h2 Ke2f2 Kf5e4 f4f5 Ke4xf5 Kf2e2 Bh2xg3 Ke2d3 h5h4 Kd3e3 Bg3f4+ Ke3f2 Kf5e4 Kf2e2 Bf4g3 Ke2d2 h4h3 g2xh3 g4xh3 Kd2c3 h3h2 Kc3b4 h2h1Q Kb4c5 Ke4f5 Kc5c4 Qh1c6+ Kc4d4 Bg3e1
 36/88	1&#58;25&#58;19	8.188.285.745	1.601.000	+M40	Kb7c6 Ke4d3 Kc6c5 Kd3e2 Kc5d5 Ke2f1 Kd5e4 Kf1e1 Ke4f5 Ke1f1 Ba7c5 Kf1e1 Bc5g1 Ke1e2 Bg1h2 Ke2f2 Kf5e4 f4f5 Ke4xf5 Kf2e2 Bh2xg3 Ke2d3 h5h4 Kd3e3 Bg3d6 Ke3f2 Kf5e4 g2g3 h4h3 Kf2g1 Bd6xg3 Kg1f1 Ke4f3 Kf1g1 h3h2+ Kg1h1 Bg3f2 Kh1xh2 g4g3+ Kh2h3 g3g2 Kh3h2 g2g1Q+ Kh2h3 Qg1h1+

I stopped the search after more than 85 minutes. There were 220,017 TB hits to the full set of Gaviota TBs.

Houdini was able to give a mate search after a little more than 45 minutes on my PC of year 2006. The true distance to mate is surely shorter than 40 moves (just my opinion).

Regarding the original question of the thread: I have no idea on who is who.

Regards from Spain.

Ajedrecista.

BeyondCritics · Post by **BeyondCritics** » Mon Jun 23, 2014 5:31 pm

Hi all,

thanks for your analysis. Finally we are coming down to the truth of this position.
And yes, Kc6 is really not so far fetched as idea. My arrogance blinded me

.
Normally i do check with stockfish, but this time i simply skipped that, because i was so proud applying the fortress idea, that no engine currently really "understand". Unfortunately this position is reduced down to a mate score nowadays...

Who is who ?

Who is who ?

Re: Who is who ?

Re: Who is who ?

Re: Who is who ?

Re: Who is who ?

Re: Who is who ?

Re: Who is who ?

Re: Who is who ?

Re: Who is who?

Re: Who is who?