Shirov's Bh3, Houdini solves it

[lech]

In fact, it's not "playing chess", it's "solve a position where you know there's unusual things to play" and you search unusual moves ... computers solve in "normal play mode"

Sting does it in "normal play mode". It lets Sting play on a very high level too.

[geots]

I remember this problem very wel, because a chess friend putted it to me in a board, as a challenge, after I told him that even being, myself, just an average player, I was pretty good at solving problems. He did it and in half a minute or so I gave him the bh3 solution.
It was not a matter of calculation, but just intuition. Is the strongest part of my mind, as it fits to a writer.

My best
Fern

Befitting a writer- and you are one of the very best- first necessary ingredient is "creativity"- which you can't cultivate. You are born with it- as you were- or you weren't. As for "intuition"- I'm not sure if that can be "learned" or not. I wouldn't think so.


My 1 & 1/2 cents,

george

[yanquis1972]

i remember when this thread came about, thinking this was about to be a dead test position...sadly, not the case. the new engines prune out Bh3 without tricks, even houdini 5 with 50 move distance set to 15 doesnt see it after 15 minutes and 20BN searched. SF, komodo suffer similarly.

[tpoppins]

Topalov - Shirov, Linares 1998 (r10)
[d]8/8/4kpp1/3p1b2/p6P/2B5/6P1/6K1 b - - 0 47
Current Let's Check status:

Going by the depth-to-points ratio the first entry appears to be Tactical Mode, the third NMP OFF.
Today's SF stubbornly sticks to 47...a3:

Code: Select all

Analysis by Stockfish 031216 x64 POPCNT:

47...a3 48.Kf2 Bc2 49.Ke1 Be4 50.g3 Kf5 51.Kf2 a2 52.Ke3 Kg4 53.Bxf6 Kxg3 54.Ke2 Bg2 55.Ke3 Kg4 56.Kf2 Bf3 57.Ke3 Be4 58.Kd2 Bb1 59.Ke3 Kg3 60.Kd2 Kf4 61.Bd4 Kf5 62.Bg7 Be4 63.Bc3 Kf4 64.Bb2 Kg3 65.Bf6 Kg2 66.Ke3 Kh3 67.Ke2 Bb1 68.Ke3 Kg4 69.Kd2 Be4 70.Ke2 Kf4 71.Kd2 Kf3 72.Bb2 Kf2 73.Be5 Bb1 74.Bf6 Kg2 75.Ke3 Bf5 76.Kd2 Kf3 77.Be5 Ke4 78.Bb2 Bg4 79.Bc3 Bf3 80.Bb2 Bg2 81.Ke2 Kf5 
  -+ (-2.71)  Depth: 67/69   00:03:05  4192MN, tb=29611011

Perhaps it will change its mind later ...

[tpoppins]

DS13 does find 47...Bh3, after trying a variety of other bisho moves and at 10x node count of the Houdini from 2012 (see the first page of this thread).

Code: Select all

Analysis by Deep Shredder 13 x64:

...
47...Bd3 48.Kf2 a3 49.Ke3 Be4 50.g3 Kf5 51.Bd4 Kg4 52.Bxf6 Kxg3 53.Kd2 Kf4 54.Ke2 Bf5 55.Bc3 Ke4 56.Bb4 a2 57.Bc3 Bc8 58.Kd2 Bh3 59.Bb2 Bf1 60.Ba1 Kf4 61.Bg7 Bg2 62.Bf6 Ke4 63.Ba1 Bf1 64.Bc3 Bc4 65.Bb2 Ba6 66.Ba1 Kf4 67.Bf6 Kg3 68.Ke3 Kg4 69.Kd2 Bb5 70.Ke3 Bc4 71.Kd2 Ba6 72.Ke3 Kg3 73.Kd2 Kf4 74.Bb2 Bc8 75.Bf6 Kg3 76.Kd3 Bh3 77.Kd2 Bg4 78.Kc2 Be2 79.Kc1 
  -+ (-1.98)  Depth: 64/78   00:07:39  18888MN, tb=48634365
47...Bh3 
  -+ (-1.98 ++)  Depth: 64/78   00:08:40  21149MN, tb=53646768
47...Bh3 48.Kf2 Kf5 49.gxh3 Ke4 50.Bb2 f5 51.Ke1 d4 52.Ke2 
  -+ (-2.04 ++)  Depth: 64/78   00:09:11  22359MN, tb=60080877
...
47...Bh3 48.Kf2 Kf5 49.gxh3 Ke4 50.Bb2 f5 51.Bh8 d4 52.Kg2 a3 53.Bf6 a2 54.Kf2 f4 55.Kg1 a1Q+ 56.Kf2 Qb2+ 57.Kg1 Qc3 58.Kf2 Qxh3 59.Bd8 Qe3+ 60.Kf1 Kf5 61.Bf6 Kxf6 62.Kg2 d3 63.Kf1 Qe8 64.Kg1 Qh8 65.h5 
  -+ (-4.27 ++)  Depth: 64/83   00:13:57  37229MN, tb=184642066

[tpoppins]

That was all with 3+4+5 Syzygy; perhaps the 6-men TBs will speed it up?

Code: Select all

Analysis by Deep Shredder 13 x64:

....
47...Bc2 
  -+ (-2.05 ++)  Depth: 67/80   00:08:40  18987MN, tb=93209512
47...a3 48.Kf2 Be4 49.g3 a2 50.Ke2 Kf5 51.Ke3 Kg4 52.Bxf6 Kxg3 53.Kd2 Bb1 54.Ke2 Kg4 55.Kd2 Be4 56.Ke2 Bf3+ 57.Kd2 Bg2 58.Ke3 Bh3 59.Kd2 Kf4 60.Bd4 Ke4 61.Bb2 Bc8 62.Ba1 Be6 63.Bb2 Bh3 64.Ba1 Bc8 65.Bb2 Ba6 66.Bc3 Kf3 67.Bf6 Kf4 68.Bd4 Kf5 69.Ba1 Bc8 70.Ke3 Kg4 71.Bf6 Ba6 72.Kd2 Bb7 73.Kc2 Kf5 74.Bb2 Ba6 75.Kb3 Bc4+ 76.Kc2 Ke4 77.Kd2 Bb5 78.Bc3 Ba6 79.Bb2 Bf1 80.Bc3 Bb5 81.Bb2 Kf3 82.Bc3 Bc6 
  -+ (-2.04)  Depth: 68/89   00:12:06  26379MN, tb=133030211
47...Bc2 
  -+ (-2.04 ++)  Depth: 68/89   00:12:43  27761MN, tb=139462365
47...a3 48.Kf2 Be4 49.g3 Kf5 50.Ke3 a2 51.Kf2 Bc2 52.Ke3 Kg4 53.Bxf6 Kxg3 54.Kd2 Bf5 55.Ke3 Bc8 56.Ke2 Ba6+ 57.Ke3 Bf1 58.Kd2 Kf4 59.Bc3 Ke4 60.Ba1 d4 61.Bb2 Bh3 62.Ba1 Bc8 63.Bb2 Bf5 64.Ba1 Bd7 65.Bb2 Be8 66.Ke2 Bb5+ 67.Kd2 Bd7 68.Ba1 Bf5 69.Ke2 Bg4+ 70.Kd2 Be6 71.Bb2 Bc4 72.Kc2 Bb3+ 73.Kd2 Bf7 74.Ba1 
  -+ (-2.11 ++)  Depth: 69/89   00:15:41  34573MN, tb=202899047
47...a3 48.Ba1 Bh3 49.gxh3 Kf5 50.Bd4 Ke4 51.Ba1 d4 52.h5 gxh5 53.Kf1 h4 54.Kg2 f5 55.Kf2 a2 56.Ke2 
  -+ (-2.18 ++)  Depth: 69/89   00:28:22  66242MN, tb=717624117
47...a3 48.Kf2 Be4 49.g3 a2 50.Kg1 Bc2 51.Kg2 Bd1 52.Kf2 Kf5 53.Ke3 Kg4 54.Bxf6 Kxg3 55.Kd2 Bb3 56.Kc1 Kf4 57.Kd2 Kf3 58.Kd3 Kf4 59.Ba1 Bc4+ 60.Kd2 Ke4 61.Bc3 d4 62.Ba1 Be6 63.Bb2 Ke5 
  -+ (-2.11 --)  Depth: 69/89   00:32:02  76034MN, tb=816208268
47...a3 48.Kf2 Be4 49.g3 Kf5 50.Bb4 a2 51.Bc3 Bc2 52.Kf3 Bd1+ 53.Ke3 Kg4 54.Bxf6 Kxg3 55.Kd2 Bb3 56.Ke3 Bc4 57.Kd2 Kg4 58.Ke3 Ba6 59.Kd2 Kf5 60.Bd4 Ke4 61.Ba1 d4 62.Bb2 Bc4 63.Ba1 d3 64.Bf6 Kf5 65.Ba1 Kg4 66.Bf6 Kg3 67.Kc1 g5 68.hxg5 Bb3 
  -+ (-1.96 --)  Depth: 69/89   01:20:45  180336MN, tb=1289388761

Good thing SSDs are cheap these days, huh?

SF after one hour and a half made no progress (I have fail-highs/lows suppressed by default). There is this in the log, though:

Code: Select all

...
<< info depth 68 currmove a4a3 currmovenumber 1
<< info nodes 4202841955
<< info nodes 4315225720
<< info nodes 4431037411
<< info nodes 4546407738
<< info depth 68 currmove f5e4 currmovenumber 2
<< info depth 68 currmove d5d4 currmovenumber 3
<< info depth 68 currmove f5g4 currmovenumber 4
<< info depth 68 currmove e6f7 currmovenumber 5
<< info depth 68 currmove f5b1 currmovenumber 6
<< info depth 68 currmove f5d3 currmovenumber 7
<< info depth 68 currmove f5h3 currmovenumber 8
<< info nodes 4657346996
...
<< info nodes 12641541918
<< info depth 68 currmove f5h3 currmovenumber 1
<< info nodes 12754562537
...
<< info nodes 134093202888
>> stop
<< info depth 68 seldepth 78 multipv 1 score cp 819 nodes 134184006694 nps 25156257 hashfull 999 tbhits 477463595 time 5334021 pv f5h3
<< bestmove f5h3 ponder g2g3

12642 MN, so about 9 minutes. How long till a full PV is anybody's guess; an hour and a half wasn't enough in this case.

[Lyudmil Tsvetkov]

i remember when this thread came about, thinking this was about to be a dead test position...sadly, not the case. the new engines prune out Bh3 without tricks, even houdini 5 with 50 move distance set to 15 doesnt see it after 15 minutes and 20BN searched. SF, komodo suffer similarly.

Bh3 is certainly the flashiest of moves, but even more certain not the best. so why play it?

I guess every black move wins: a3 wins, g5 should win, d4 and Kd6 with a sac. almost certainly also win, I would myself play Be4 and Kf5, which also wins quite easily.

I guess the problem is simply wrong and not the way engines handle this position.

[Eelco de Groot]

i remember when this thread came about, thinking this was about to be a dead test position...sadly, not the case. the new engines prune out Bh3 without tricks, even houdini 5 with 50 move distance set to 15 doesnt see it after 15 minutes and 20BN searched. SF, komodo suffer similarly.

Bh3 is certainly the flashiest of moves, but even more certain not the best. so why play it?

I guess every black move wins: a3 wins, g5 should win, d4 and Kd6 with a sac. almost certainly also win, I would myself play Be4 and Kf5, which also wins quite easily.

I guess the problem is simply wrong and not the way engines handle this position.

You have got to learn the classics Lyudmil I once had the pleasure of being able to participate in a training with GM Sipke Ernst. The training was more intended for KNSB players, but not enough people were attending and for a little sum, everyone could join, even though I was no longer a club member at the time. The Frisian Grandmaster brought up the Topalov - Shirov position in a Bishop endgame discourse and for once I knew something that players many hundreds of Elo points stronger had not yet seen. I recognized the position! Apart from the key move, I did not know much theory but it was nice to see that reading CCC had taught me some useful chess knowledge!

The maneuver is very deep as Dvoretsky also says, and most engines will just prune it now... If you search it with 13 moves Multi PV, even Stockfish will find the win though Stock could probably do better if he recognized that other moves make no progress maybe (different colour bishops and such).

Here is a lesson from the late great Mark Dvoretsky, in his ChessCafé column. I posted this .pdf link several times before but probably the posts are no longer in the Talkchess archives, the one in the earlier CCC archives before Talkchess should probably still exist, showing that Rebel Q5T tactical personality also was able to produce Bh3, on a Windows 98 SE Celeron I think, at some 500 Mhz!

http://billwall.phpwebhosting.com/dvore ... tsky06.pdf

[BBauer]

May be it was a lucky run:

Code: Select all

FEN: 8/8/4kpp1/3p1b2/p6P/2B5/6P1/6K1 b - - 0 1
St-mod-16-12-02:
Found 510 tablebases
 50/56	02:12	 173.020k	1.306k	+2,26	a4-a3 Kg1-f2 Bf5-c2 Kf2-e2 Bc2-e4 g2-g3 Ke6-f5 Bc3-d4 Kf5-g4 Bd4xf6 Kg4xg3 Ke2-d2 Kg3-f4 Bf6-d4 a3-a2 Bd4-f6 Kf4-g4 Kd2-e3 Kg4-f5 Bf6-c3 Be4-h1 Ke3-d2 Bh1-f3 Kd2-c2 Kf5-f4 Kc2-d3 Bf3-g2 Kd3-d2 Kf4-g3 Bc3-f6 Kg3-g4 Kd2-c1 Bg2-e4 Kc1-d2 Kg4-h3 Kd2-e2 Kh3-g3 Ke2-d2 Kg3-g2 Kd2-e3 Be4-c2 Ke3-d2 Bc2-b1 Bf6-a1 Kg2-h3 Ba1-f6 Kh3-g4 Kd2-c1 Kg4-f4 Kc1-d2 Bb1-f5 Bf6-d4 Kf4-g4 Bd4-f6 Kg4-f3 Bf6-a1
 51/56	02:17	 182.220k	1.323k	+2,26	a4-a3 Kg1-f2 Bf5-c2 Kf2-e2 Bc2-e4 g2-g3 Ke6-f5 Bc3-d4 Kf5-g4 Bd4xf6 Kg4xg3 Ke2-d2 Kg3-f4 Bf6-d4 a3-a2 Bd4-f6 Kf4-g4 Kd2-e3 Kg4-f5 Bf6-c3 Be4-h1 Ke3-d2 Bh1-f3 Kd2-c2 Kf5-f4 Kc2-d3 Bf3-g2 Kd3-d2 Kf4-g3 Bc3-f6 Kg3-g4 Kd2-c1 Bg2-e4 Kc1-d2 Kg4-h3 Kd2-e2 Kh3-g3 Ke2-d2 Be4-b1 Kd2-e3 Bb1-f5 Ke3-d2 Kg3-f3 Bf6-a1 Bf5-e4 Ba1-c3 Kf3-g2 Bc3-f6 Kg2-g3 Kd2-c1 Be4-b1 Kc1-d2 Kg3-g2 Bf6-a1
 52/57+	04:39	 436.623k	1.563k	+2,33	Bf5-h3
 52/57+	05:20	 538.195k	1.680k	+2,41	Bf5-h3
 52/57	10:53	 1.378.699k	2.109k	+3,53	Bf5-h3

but stockfish found it in 4:39 min on a amd laptop with 4 threads 1024 mb hash and 6-piece syzygy.

@Ludmil: As you can see it is a mate in 4 or 5 moves

Kind regards
Bernhard

[Lyudmil Tsvetkov]

i remember when this thread came about, thinking this was about to be a dead test position...sadly, not the case. the new engines prune out Bh3 without tricks, even houdini 5 with 50 move distance set to 15 doesnt see it after 15 minutes and 20BN searched. SF, komodo suffer similarly.

Bh3 is certainly the flashiest of moves, but even more certain not the best. so why play it?

I guess every black move wins: a3 wins, g5 should win, d4 and Kd6 with a sac. almost certainly also win, I would myself play Be4 and Kf5, which also wins quite easily.

I guess the problem is simply wrong and not the way engines handle this position.

You have got to learn the classics Lyudmil I once had the pleasure of being able to participate in a training with GM Sipke Ernst. The training was more intended for KNSB players, but not enough people were attending and for a little sum, everyone could join, even though I was no longer a club member at the time. The Frisian Grandmaster brought up the Topalov - Shirov position in a Bishop endgame discourse and for once I knew something that players many hundreds of Elo points stronger had not yet seen. I recognized the position! Apart from the key move, I did not know much theory but it was nice to see that reading CCC had taught me some useful chess knowledge!

The maneuver is very deep as Dvoretsky also says, and most engines will just prune it now... If you search it with 13 moves Multi PV, even Stockfish will find the win though Stock could probably do better if he recognized that other moves make no progress maybe (different colour bishops and such).

Here is a lesson from the late great Mark Dvoretsky, in his ChessCafé column. I posted this .pdf link several times before but probably the posts are no longer in the Talkchess archives, the one in the earlier CCC archives before Talkchess should probably still exist, showing that Rebel Q5T tactical personality also was able to produce Bh3, on a Windows 98 SE Celeron I think, at some 500 Mhz!

http://billwall.phpwebhosting.com/dvore ... tsky06.pdf

do not believe that for a moment.

Berger-Kotlermann is very simple, here you have 4 more pawns, black wins one way or another. the eval advantage is simply very huge, and I trust eval before I decide to trust anyone or anything else.

sorry, no time to suggest specific lines, but do not believe for a moment Bh3 is the only solution. Dvoretsky analysed even before latest top engines appeared, and some of his analysis is based on engine output.