Some "bright" positions

[Uri Blass]

After playing:

Code: Select all

d2d1 c1d1 h3d3 d1e1 g5h4 e1f1 d3f3 f1g1 h4f2 g1f1 f2c5 f1e1 c5b4 e1d1 f3d3 d1c1 b4a3 c1b1 d3b3 b1a1 a3b2 a1b1 b2e5 b1c1 e5f4 c1d1 b3d3 d1e1 f4g3 e1f1 d3f3 f1g1 g3f2 g1f1 f2c5 f1e1 c5b4 e1d1 f3d3 d1c1 b4a3 c1b1 d3b3 b1a1

we reach the position where the PV diverges:

[D]8/K6P/5p2/8/8/BR6/p1p1p1p1/k6b w - - 21 1

Here Chest continues with Bb2+ which Hamsters sees as +M12 but then Hamsters thinks h8=Q is +M8 with the following PV:

Code: Select all

h8=Q g1=Q+ Ka6 Bb7+ Rxb7 Qf1 Bb2+ Kb1 Bxf6+ Kc1 Bb2+ Kd1 Qd4+ Ke1 Bc3+

I have only Shredder 11 to check now and it falls for +M8 too, so I can't see the refutation... what does your engine say?

Chest says it is mate in 8

Uri

[Uri Blass]

Here's another position where bright gives wrong mate (this seldom happens):

...

This is mate in 34.

Is this mate proven? Hamsters finds mate in 34 but in some configurations it improves to mate in 30. Someone with Chest?!?

Chest says mate in 34:


WinCHEST Ver.3.19f+, 01-Feb-2006
Options = -M256 -Z34 -eL -R3 -K3 -P5 -X16 -0 -5 -SUr
Input file: STDIN
Reading job:
% created by ChestUCI Ver.4.4
W: Ka7 Rd2 Rh3 Bg5 Ph7 (5)
B: Kc1 Bh1 Pa2 Pc2 Pe5 Pe2 Pf6 Pg2 (8)
FEN: 8/K6P/5p2/4p1B1/8/7R/p1pRp1p1/2k4b w - -
analysing (special-mate in 34 moves):
Solution (in 34 moves):
d2d1
Time (virt) = 0.359 sec

PV= d2d1 c1d1 h3d3 d1e1 g5h4 e1f1 d3f3 f1g1 h4f2 g1f1 f2c5 f1e1 c5b4 e1d1 f3d3 d1c1 b4a3 c1b1 d3b3 b1a1 a3b2 a1b1 b2e5 b1c1 e5f4 c1d1 b3d3 d1e1 f4g3 e1f1 d3f3 f1g1 g3f2 g1f1 f2c5 f1e1 c5b4 e1d1 f3d3 d1c1 b4a3 c1b1 d3b3 b1a1 a3b2 a1b1 b2f6 b1c1 f6g5 c1d1 b3d3 d1e1 g5h4 e1f1 d3f3 f1g1 h4f2 g1f1 f2g3 f1g1 h7h8q a2a1q h8a1 c2c1q a1c1 e2e1q c1e1
d2d1 c1d1 h3d3 d1e1 g5h4 e1f1 d3f3 f1g1 h4f2 g1f1 f2c5 f1e1 c5b4 e1d1 f3d3 d1c1 b4a3 c1b1 d3b3 b1a1 a3b2 a1b1 b2e5 b1c1 e5f4 c1d1 b3d3 d1e1 f4g3 e1f1 d3f3 f1g1 g3f2 g1f1 f2c5 f1e1 c5b4 e1d1 f3d3 d1c1 b4a3 c1b1 d3b3 b1a1 a3b2 a1b1 b2f6 b1c1 f6g5 c1d1 b3d3 d1e1 g5h4 e1f1 d3f3 f1g1 h4f2 g1f1 f2g3 f1g1 h7h8q a2a1q h8a1 c2c1b a1c1 e2e1b c1e1

You used chest wrongly and did not give it to search for the shortest mate.
Chest cannot find the shortest mate in a reasonable time here.

Uri

[F.Huber]

You used chest wrongly and did not give it to search for the shortest mate.
Chest cannot find the shortest mate in a reasonable time here.

Well, I won´t say that he´s used Chest wrongly - in fact the problem is that this position is contained in the ChestUCI database as #34 and that´s just not correct (as I´ve checked now) - here´s the better solution:

ChestUCI Ver.4.6:
CPU: Celeron 400MHz
FEN: 8/K6P/5p2/4p1B1/8/7R/p1pRp1p1/2k4b w - - (5+8)
Position-Analysis: C0/R0/K3/P6/X17 W:5/35
Parameters found in Clipboard: #30; 00:05 @ C0/R3/K3/P4/X34;
Search for Special-Mate [C0/R3/K3/P4/X34] in 30 ... (Hash=64MB)
30 00:05 181.261 40.280 +M30 1.Rd1+
Search completed ... (Time=5.22s)
Mate in 30 found ! (1 Solution in 00:05)
30/30 00:05 232.023 44.448 +M30 1.Rd1+ Kxd1 2.Rd3+ Ke1 3.Bh4+ Kf1 4.Rf3+ Kg1 5.Bf2+ Kf1 6.Bc5+ Ke1 7.Bb4+ Kd1 8.Rd3+ Kc1 9.Ba3+ Kb1 10.Rb3+ Ka1 11.Bb2+ Kb1 12.Bxe5+ Kc1 13.Bf4+ Kd1 14.Rd3+ Ke1 15.Bg3+ Kf1 16.Rf3+ Kg1 17.Bf2+ Kf1 18.Bc5+ Ke1 19.Bb4+ Kd1 20.Rd3+ Kc1 21.Ba3+ Kb1 22.Rb3+ Ka1 23.h8Q g1Q+ 24.Ka6 Bb7+ 25.Rxb7 Qf1 26.Bb2+ Kb1 27.Bc3+ Kc1 28.Qh6+ Qf4 29.Qxf4+ Kd1 30.Qd2+
Best move: Rd1, Value: +Mate in 30, Depth: 30/30, Time: 00:05,809, 232.023 Nodes, 44.448 N/sec


Of course I don´t remember from where I have this mate puzzle, but I´m quite sure that it was given as #34 - otherwise I would have tried a shorter solution.
The problem in this case is the variant 1.Rd1+ Kb2 - although this black defense leads to a much shorter mate, the ´MaxMoves´restriction X16 from the ChestUCI database doesn´t hold anymore for this shorter variation, and so Chest fails to find a #30 with the parameter X16.

So the correct line in ChestUCI.epd for this positions should be:
8/K6P/5p2/4p1B1/8/7R/p1pRp1p1/2k4b w - - bm #30; 00:05 @ C0/R3/K3/P4/X34;

Thanks to all here for one more correction of my ChestUCI database,
Franz.

[George Tsavdaris]

You used chest wrongly and did not give it to search for the shortest mate.
Chest cannot find the shortest mate in a reasonable time here.

Well i thought Chest only searches for the shortest mate no matter what parameters you give it.
I think this indeed happens and perhaps Franz can confirm it.

I know that when Chest reports time solutions of 0.3 s or something most times it's because it has the position on its database. Just like this time. So i guessed everything was OK since until now there was no error reported in its database.
But this time an error has been found.....

[George Tsavdaris]

The problem in this case is the variant 1.Rd1+ Kb2 - although this black defense leads to a much shorter mate, the ´MaxMoves´restriction X16 from the ChestUCI database doesn´t hold anymore for this shorter variation, and so Chest fails to find a #30 with the parameter X16.

So the correct line in ChestUCI.epd for this positions should be:
8/K6P/5p2/4p1B1/8/7R/p1pRp1p1/2k4b w - - bm #30; 00:05 @ C0/R3/K3/P4/X34;

Does Chest only searches for the shortest mate no matter what parameters someone sets?
I mean if we set Special=C0/R0/K3/P0/X0 for example then Chest searches for the shortest mate no matter what. Right?
So if it finds a mate in 18 then we can be sure that there is no mate in less than 18. Is this correct?

[F.Huber]

Does Chest only searches for the shortest mate no matter what parameters someone sets?
I mean if we set Special=C0/R0/K3/P0/X0 for example then Chest searches for the shortest mate no matter what. Right?
So if it finds a mate in 18 then we can be sure that there is no mate in less than 18. Is this correct?

No George, that´s not correct - I thought this would be clear since ChestUCI exists already for some years.

Whenever you search for a special mate (i.e. with any of the special parameters set unequal to zero), then Chest only searches for the shortest mate under those special conditions!.

So if with your example K3 Chest finds a #18, then this is only the shortest mate (in 18), where at no place in the whole move tree the defender (usually black) has more than 3 ´KingSquares´ (IOW 2 possible king moves).
Thus there might exist a shorter mate, but anywhere in this (possibly) shorter solution there will definitely be a position with more than 3 KingSquares for the black king.

Short conclusion:
A guarantee for the absolutely shortest mate you´ll only have with Chest´s brute-force mode (i.e. with ´SearchMode=Normal´ or with all special parameters set to zero) - but with this brute-force method of course you won´t be able to solve most of the longer mates (like #15, #30 or even upto #126 ... )

[Jouni]

Wow bright beats Chest by far Please Franz explain how You got right
parameters here? I run Chest in Autoturbo mode (different search orders) and it isn't near the solution in 10 minutes (P2400). Brigt is about 1000000x faster here...

Jouni

[F.Huber]

Wow bright beats Chest by far

Yes, that´s indeed an impressive performance of Bright, but for one single position everything is possible! Make a statistics of a few 100 positions from the ChestUCI database, and I´m quite sure that Bright´s results won´t be nearly that good.

Please Franz explain how You got right
parameters here? I run Chest in Autoturbo mode (different search orders) and it isn't near the solution in 10 minutes (P2400).

Well, I started after 20 moves of the solution, because someone said that he got a different PV from here. The remaining #10 was quite easy to find for Chest, and then I got the complete parameter set with ChestUCIs ´ParameterAnalysis´ mode, going back from the final to the starting position - a matter of a few minutes.
Ok, arriving at the very first move (i.e. at the staring position) you´ll need some experience with the handling of ChestUCI ...

[George Tsavdaris]

No George, that´s not correct - I thought this would be clear since ChestUCI exists already for some years.

Whenever you search for a special mate (i.e. with any of the special parameters set unequal to zero), then Chest only searches for the shortest mate under those special conditions!.

Of course this is what i meant! I possibly didn't express myself correctly....
If you set some parameters, then what it finds it is obviously with these parameters! Since it didn't search with any other.

[George Tsavdaris]

Yes, that´s indeed an impressive performance of Bright, but for one single position everything is possible! Make a statistics of a few 100 positions from the ChestUCI database, and I´m quite sure that Bright´s results won´t be nearly that good.

Yes and also with Bright you can't be sure it's the shortest mate while you Chest you are(for the given parameters for not being misunderstood again )....
Unless Bright has been designed to find the shortest mate before it announces a mate, but i doubt.

A question:
I'm aware of some more specialized program for solving mates etc. Like WinChloe for example. Do you have any idea if WinChloe for example is faster from Chest-UCI in finding mates?
Because i have seen some amazing time results by WinChloe in the past in some positions....