not considering trading to winning pawn endgame as correct solution

[uri blass]

I failed in the following position(did not save number of the position but saved the position)

8/7p/p3k3/1p4p1/P2B1p2/1PPr1P1P/6PK/8 b - - 0 1

I found Rxd4 with winning pawn endgame but the computer decided that my solution is wrong.

b4 is also winning but the first thing that I think about is trading to pawn endgame and I know that time is important in this test so I did not check other alternatives.

[richard]

Hi Uri,

The problem generator currently runs without endgame table bases so it can sometimes miss winning moves with endgames (where the win is further out than the search time allows it to see).  For the next verification run of the problem set I'm hoping to run with tablebases enabled but if you can tell me the problem number for this one I can manually remove it from the set.

Regards,
Richard.

[uri blass]

Hi Richard,

I do not expect tablebases to help here because there are too many pawns.
I checked the problem with rybka3 and even after Rxd4 cxd4 rybka
needs a long time to see that black is completely winning.

I do not know how can I see problems that I tried to solve in the past and to see the problem numbers.

I posted a fen of the position that chess programs understand

Here is a direct link:

http://www.jinchess.com/chessboard/?p=---------------pp---k----p----p-P--B-p---PPr-P-P------PK--------&tm=b

Uri

[uri blass]

I looked at the FAQ and found that the problem id was 46393

[uri blass]

 There is also a win in pawn endgame in problem 44061

Rxe6 is a faster way to win but I checked that the pawn endgame after
1.Qxh6+ Ke8 (Ke7 2.Qf6+ Ke8 3.Qh8+ is the same) 2.Qh8+ Ke7 3.Qxd8+ wins for white.

[richard]

Thanks Uri,

I've disabled both of those, let me know when you find others.

Table bases can still help with some of these, as even though there are too many pieces to directly look up the solution, the engine can avoid looking down lots of branches by using the tablebases once it gets to more simplified positions, and this allows the engine to see more for a given period of time, thus increasing the chances of seeing the endgame correctly. 

Sometimes tablebases are not enough and only spending longer analysing each position will help. This can also be true of non-endgame positions, there was a thread recently discussing situations where evaluations can change after 18+ ply which is roughly how far the problem generator sees with current time settings (it can sometimes see further if the position is relatively simple).

Regards,
Richard.