Simple fortress detection, engines fail

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

kgburcham
Posts: 2016
Joined: Sun Feb 17, 2008 4:19 pm

Re: Simple fortress detection, engines fail

Post by kgburcham »

bob wrote:
Houdini wrote:If you say "fortress" I reply "FiftyMoveDistance=12" and get the following result with Houdini 2.0c:

Code: Select all

8/8/8/5Bp1/7k/7P/4pPK1/8 b - -
Engine: Houdini 2.0c x64 (2048 MB)
17/38  0:01   -4.52    1...g4 2.Bxg4 e1Q 3.Bf3 Kg5 4.Bg4 Kf4 
                       5.Bf3 Qe7 6.Bg4 Ke5 7.Kg3 Kd4 8.Kf3 Kc3 
                       9.Bf5 Kd2 10.Kg4 Ke1 11.f4 Kf2 
                       12.Bg6 Qe2+ 13.Kh4 Kg2 14.f5 Qe4+ (8.330.990) 8175
The obvious question, why not run with that set to 12 all the time? Tuning a parameter to a specific position is not exactly solving the position as one would generally do. Setting the counter to 12 would be quite harmful overall, which makes it useless here as well.
Robert Hyatt
If you decide to go commercial with Crafty and I have to pay $125 for the mp version, would you please include this same option that Robert Houdart has included? It offers some variety and fun to the program. With the fifty move option in Houdini, I can use it to solve test positions and then switch right back to the #1 world default settings.
thanks in advance
kgburcham
Arpad Rusz
Posts: 273
Joined: Sat Apr 17, 2010 2:34 pm
Location: Budapest

Re: Simple fortress detection, engines fail

Post by Arpad Rusz »

J. Hasek (1929)

[d]k7/1p1p1p2/1PpKpP2/2P1P3/p7/8/6N1/3b4 w - - 0 1

Draw!
Arpad Rusz
Posts: 273
Joined: Sat Apr 17, 2010 2:34 pm
Location: Budapest

Re: Simple fortress detection, engines fail

Post by Arpad Rusz »

A.P. Kuznetsov (1965)

[d]7r/p1p1p1pk/3pPp1p/2pP1PP1/P1P4P/1K6/P6B/8 w - - 0 1

Draw

After 1.g6+ Kg8 we have the following position:

[d]6kr/p1p1p1p1/3pPpPp/2pP1P2/P1P4P/1K6/P6B/8 w - - 0 2

The author's solution is: 2.Bxd6!! :shock:
If that is the only move which draws, this could be one of the hardest test positions.
Vinvin
Posts: 5333
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Re: Simple fortress detection, engines fail

Post by Vinvin »

Arpad Rusz wrote:A.P. Kuznetsov (1965)


[d]6kr/p1p1p1p1/3pPpPp/2pP1P2/P1P4P/1K6/P6B/8 w - - 0 2

The author's solution is: 2.Bxd6!! :shock:
If that is the only move which draws, this could be one of the hardest test positions.
Amazing draw :twisted:

[Event "-"]
[Site "-"]
[Date "2012.07.06"]
[Round "?"]
[White "-"]
[Black "-"]
[Result "*"]
[SetUp "1"]
[FEN "6kr/p1p1p1p1/3pPpPp/2pP1P2/P1P4P/1K6/P6B/8 w - - 0 2"]

2. Bxd6 cxd6 3. h5 a5 4. Ka3 Kf8 5. Kb2 Ke8 6. Kb3 Kd8 7. Kb2 Kc7 8. Kb3 *

[d]7r/2k1p1p1/3pPpPp/p1pP1P1P/P1P5/1K6/P7/8 b - - 9 8
or

[Event "-"]
[Site "-"]
[Date "2012.07.06"]
[Round "?"]
[White "-"]
[Black "-"]
[Result "*"]
[SetUp "1"]

2. Bxd6 cxd6 3. h5 Kf8 4. a5 Ke8 5. Ka4 Kd8 6. a6 Kc7 7. Ka5 Rb8 8. a4 *

[d]1r6/p1k1p1p1/P2pPpPp/K1pP1P1P/P1P5/8/8/8 b - a3 0 8
carldaman
Posts: 2287
Joined: Sat Jun 02, 2012 2:13 am

Re: Simple fortress detection, engines fail

Post by carldaman »

I've been using FiftyMoveDistance=12 (or 13) a lot in my endgame & late middlegame analysis especially when I see difficulty in making progress.

The question is -- can code be implemented that measures this lack of progress (in a game situation, not analysis) for several moves (based on eval), and then proceeds to reset FMD to 12, on the fly? This way the feature will only be turned on if there is a reason for it (as a last resort).

Regards,
CL
User avatar
Houdini
Posts: 1471
Joined: Tue Mar 16, 2010 12:00 am

Re: Simple fortress detection, engines fail

Post by Houdini »

carldaman wrote:I've been using FiftyMoveDistance=12 (or 13) a lot in my endgame & late middlegame analysis especially when I see difficulty in making progress.

The question is -- can code be implemented that measures this lack of progress (in a game situation, not analysis) for several moves (based on eval), and then proceeds to reset FMD to 12, on the fly? This way the feature will only be turned on if there is a reason for it (as a last resort).

Regards,
CL
I'd rather look at it the other way round: enable FiftyMoveDistance=12 by default, and try to recognize situations where it doesn't work.
Obviously for mating against king alone, e.g. KBN v K , but that is rather trivial.

Robert
Vinvin
Posts: 5333
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Re: Simple fortress detection, engines fail

Post by Vinvin »

Houdini wrote:
carldaman wrote:I've been using FiftyMoveDistance=12 (or 13) a lot in my endgame & late middlegame analysis especially when I see difficulty in making progress.

The question is -- can code be implemented that measures this lack of progress (in a game situation, not analysis) for several moves (based on eval), and then proceeds to reset FMD to 12, on the fly? This way the feature will only be turned on if there is a reason for it (as a last resort).

Regards,
CL
I'd rather look at it the other way round: enable FiftyMoveDistance=12 by default, and try to recognize situations where it doesn't work.
Obviously for mating against king alone, e.g. KBN v K , but that is rather trivial.

Robert
... and where there's long manoeuvres with pieces (no pawn move and no take). Other engines simply decrease score after many moves with no pawn move and no take ...
User avatar
Houdini
Posts: 1471
Joined: Tue Mar 16, 2010 12:00 am

Re: Simple fortress detection, engines fail

Post by Houdini »

Vinvin wrote:... and where there's long manoeuvres with pieces (no pawn move and no take). Other engines simply decrease score after many moves with no pawn move and no take ...
Can you provide a position?
Vinvin
Posts: 5333
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Re: Simple fortress detection, engines fail

Post by Vinvin »

Houdini wrote:
Vinvin wrote:... and where there's long manoeuvres with pieces (no pawn move and no take). Other engines simply decrease score after many moves with no pawn move and no take ...
Can you provide a position?
Salut Robert, quel genre de positions ?
Last edited by Vinvin on Fri Jul 06, 2012 12:36 pm, edited 1 time in total.
User avatar
Houdini
Posts: 1471
Joined: Tue Mar 16, 2010 12:00 am

Re: Simple fortress detection, engines fail

Post by Houdini »

Arpad Rusz wrote:A.P. Kuznetsov (1965)

[d]7r/p1p1p1pk/3pPp1p/2pP1PP1/P1P4P/1K6/P6B/8 w - - 0 1

Draw

After 1.g6+ Kg8 we have the following position:

[d]6kr/p1p1p1p1/3pPpPp/2pP1P2/P1P4P/1K6/P6B/8 w - - 0 2

The author's solution is: 2.Bxd6!! :shock:
If that is the only move which draws, this could be one of the hardest test positions.
There we go :).

Code: Select all

6kr/p1p1p1p1/3pPpPp/2pP1P2/P1P4P/1K6/P6B/8 w - -

Engine: Houdini DEV 3T (4096 MB)
by Robert Houdart

20/34  0:01   -1.86    1.Kc3 Kf8 2.Bf4 Ke8 3.Kd3 Kd8 4.Ke4 h5 
                       5.Bc1 Kc8 6.Bd2 Kb7 7.Be1 c6 8.Ba5 Rg8 
                       9.Bd2 Kb6 10.Be1 Rh8 11.Bd2 Rb8 
                       12.a5+ Kc7 (9.911.938) 6990 
...
25/49  0:11   -1.98    1.Kc3 h5 2.Bf4 Kf8 3.Kd3 Ke8 4.Bc1 Kd8 
                       5.Ke4 c6 6.dxc6 (83.814.754) 7075 

25/49  0:12   -1.96    1.Bf4 Kf8 2.h5 Ke8 3.Bd2 Kd8 4.Kc2 Kc8 
                       5.Kd3 Kb7 6.Ke4 c6 7.Kd3 Rb8 8.Kc2 Ka6 
                       9.Be1 Kb6 10.Bd2 Ka6 11.Be1 cxd5 
                       12.cxd5 Kb7 13.Kd3 Ka6 14.Kc4 (87.965.737) 7076 
...
30/54  1:40   -2.55    1.Bf4 Kf8 2.h5 Ke8 3.Bd2 Kd8 4.Kc2 Kc8 
                       5.Kd3 Kb7 6.Ke4 c6 7.Kd3 Rb8 8.a3 Kc8 
                       9.Kc2 cxd5 10.cxd5 Kc7 11.Ba5+ Kb7 
                       12.Be1 Ka6 13.Bd2 c4 14.Bb4 (697.039.255) 6933 
...
36/80  42:08  -3.62    1.Bf4 Kf8 2.Bc1 Ke8 3.Kc3 Kd8 4.Bd2 c6 
                       5.Kc2 Kc7 6.Bc1 h5 7.Bd2 Kb7 8.Bc3 Ka6 
                       9.a3 cxd5 10.cxd5 Kb6 11.Kd3 a5 
                       12.Bd2 Rb8 13.Bc1 Ka6 14.Bf4 (16.335.719.358) 6459 

36/80  42:21  -3.45++  1.a5 (16.419.311.066) 6460 

36/80  42:21  -3.11++  1.a5 (16.420.928.555) 6460 

33/80  42:21  -2.22++  1.a5 (16.421.551.311) 6460 

33/80  42:24  -0.10++  1.a5 (16.436.736.975) 6459 

33/80  42:27  0.00     1.a5 a6 2.Bxd6 cxd6 3.h5 Kf8 4.a3 Ke8 
                       5.Kc2 Kd8 6.Kb3 Kc7 7.Ka4 Kc8 8.Kb3 Kb8 
                       9.Ka4 Rc8 10.Kb3 Rc6 11.Ka4 Rc7 
                       12.Kb3 Kc8 13.Ka4 Kd8 14.Kb3 (16.452.658.003) 6459 

34/80  42:28  0.00     1.a5 a6 2.Bxd6 cxd6 3.h5 Kf8 4.a3 Ke8 
                       5.Kc2 Kd8 6.Kb3 Kc7 7.Ka4 Kc8 8.Kb3 Kb8 
                       9.Ka4 Rc8 10.Kb3 Rc6 11.Ka4 Rc7 
                       12.Kb3 Kc8 13.Ka4 Kd8 14.Kb3 (16.461.793.003) 6458 

35/80  42:29  0.00     1.a5 a6 2.Bxd6 cxd6 3.h5 Kf8 4.a3 Ke8 
                       5.Kc2 Kd8 6.Kb3 Kc7 7.Ka4 Kc8 8.Kb3 Kb8 
                       9.Ka4 Rc8 10.Kb3 Rc6 11.Ka4 Rc7 
                       12.Kb3 Kc8 13.Ka4 Kd8 14.Kb3 (16.466.473.981) 6458 

36/80  42:33  0.00     1.a5 a6 2.Bxd6 cxd6 3.h5 Kf8 4.a3 Ke8 
                       5.Kc2 Kd8 6.Kb3 Kc7 7.Ka4 Kc8 8.Kb3 Kb8 
                       9.Ka4 Rc8 10.Kb3 Rc6 11.Ka4 Rc7 
                       12.Kb3 Kc8 13.Ka4 Kd8 14.Kb3 (16.491.363.715) 6457 
Houdini suggests that a5 followed by Bxd6 is a draw too.