High-rated ambiguous problems.

[drahacikfm]

Instead of making a new topic for each problem, I will just list here the high-rated ambiguous problems that I find, with short notes for each.  Then you can check if your next generator disables these.

30952:  5.Bxe6 or 6.Bxe6 are completely winning, piece up endgame.

[richard]

Thanks Drahacik,

If you have Rybka analysis for these that would be useful also, although I imagine for some of these it is obvious without needing to get the engine out.

Regards,
Richard.

[drahacikfm]

20922:  Rybka, depth 20, over 2 minutes thinking time:
1.Qf3  +5.12
1.cxd5 +5.12
Now that's perfect ambiguity 
Also, the move I played, 1.h4 is +3.11

[drahacikfm]

19008: I got this one correct with 1...Bf1, but after the alternative winning move 1...Rf2 2.Qxe5+ Qxe5+ 3.Rxe5 Bf1, Rybka thinks best defense is 4.Re1 Bxg2 5.Re2 Rxe2 6.Nxe2 Bxd5 which is a simple endgame of bishop and 5 pawns against knight and 3 pawns.  Most Grandmasters would just resign instead of trying to defend that.  But hard for the generator because it evaluates less than +3.0.

[richard]

Thanks,

Keep them coming, I will not respond to all of them but it will be great having this thread to use as a source of test cases.  Re: the +3.0 and GMs, I'm considering going as low as +2.5 for "good moves" in the next generator, previously I didn't want to throw away good tactics based on an evaluation that small but if I get support for allowing winning alternatives (in the way discussed a few days, i.e. not accepted as answer, but don't fail problem) then I don't have to be concerned as I can keep the position.

Regards,
Richard.

[drahacikfm]

24206:  Alternative move 2.Nxe6 and White is up a queen for a knight and it's White's move too!  Rybka +4.82 depth 16.

Also, on move 1, alternative winning move:  1.Ng5 +3.44 and others over +2.5

[drahacikfm]

14423:  a low-rated ambiguous problem.  1...Bd4+ 2.Kg2 (or anywhere else) e6 and Black is up a rook for one pawn.  No need for Rybka on this one.

[drahacikfm]

33502:  After 1...Nf5 2.Qf4, the alternative move 2...Re3 is +5.60 Rybka, depth 19.  The move I played 2...Rf3, is +3.61.

Yes, the tolerance for "winning moves" should be lowered to at least +2.5, maybe even lower.  There are databases with millions of games out there, so I don't think you have to worry about not finding enough problems.  Better to have completely clean problems.

[drahacikfm]

14335: Alternative move 3.Bxb7 is a full bishop up and Black's king is exposed.  Completely winning.  Rybka +2.99 depth 20.

Also, the weird alternative move 4.Rd8 leaves White up a rook, and Rybka depth 21 says +4.91

[drahacikfm]

8927:  The alternative I played, 1...Qc6, is +3.81 Rybka depth 20.

[drahacikfm]

12821: I played 2.Nxf5, which is +3.11 Rybka depth 22.

[drahacikfm]

23624: I played the alternative winning move 1...Bf4 and after 2.Nxf4 Rxf4 Black has a crushing attack and Rybka says +4.62 depth 24.  Also 2.Ne3 is mate in 8 for Black.  So White just loses the exchange for nothing and Black still has a big attack.

[drahacikfm]

31645:  1.Qxe5+ is a knight-up endgame, easy win, more than +3.0 Rybka.

[drahacikfm]

10536:  Alternate winning move 2.Nxg7 is +3.09 Rybka.

[tmr]

I think these ambiguous winning line problems are a lot more frustrating than the mate in n/n+1 problems.  With the mate problems at least you know you didn't find the quickest mate.  With the ambiguous lines the difference a lot of times isn't so clear cut.  I think you mentioned that the proportion of mate problems is much less in the current problems set.  That means these problems will be causing a lot more frustration.