High-rated ambiguous problems.

[tama]

I agree very nice problem and congratulations for hitting number one!

Even if u only solved 2 problems in standard since the new algorithim came out.

But dont think you r in the clear ill be Number one soon!

[drahacikfm]

23458:  Very bad problem.  The problem gives 1.fxe3, but Rybka depth 17 says 1.Bxe6+ is better.  Hard to gain rating points when you find the best move and get marked wrong!

[drahacikfm]

12861:  Another very bad problem:  1.Rff2 and 1.Rf6 are both +5.42 Rybka, depth 21.  And you get marked wrong for one of them, and not the other.

[drahacikfm]

32403: Another horrible problem.  1.Rxb4+ followed by QxQ is an endgame a knight and a pawn up, +5.26 on Rybka.  Problem was attempted 24 times, 22 people played 1.Rxb4.  The problem wants you to take the queen first, then take the bishop.

[tama]

I inconterd this problem in the past, there is no clear tactic, anymove wins endgame almost 100% problem 32403

[drahacikfm]

15640: Ambiguous problem.  Alternate 2.Rxc4 Rxc4 3.Qxc4 Qxf2 is a full bishop up. Alternate 2.Qxf6+ Kxf6 3.Bd5 Ne3 4.Rxc1 Nxd5 is exchange and a pawn up.  Both must be included as alternate winning moves in new problem set.

(Edit:  I see I covered this problem before, but because of ambiguous moves on move 1, not on move 2!  And Richard said this was thrown out by the new generator, so ok).

[xjamesbx]

Hope you don't mind if I post one of the vague ones I encountered, drahacikfm.

7665:  In the starting position, black is up a knight and a bishop for a pawn.  I figured the move b4 would hold on to all the material (and it does), but the actual solution is Ne5.  Skeptical, I plugged the position into Fritz and Crafty and they both thought b4 and Ne5 were very much winning moves.  After a few minutes, Ne5 is found to be a tad better, but both are completely winning and around +7.00 in favor of black.

The problem is rated close to 1900.

[drahacikfm]

Yes, 7665 should be thrown out not only because there is more than one winning move, but because Black is up 5 points of material at the start.

[drahacikfm]

19176:  A nice problem, but it goes on one move too long.  1.Rc7 Be7 2.Nxc8 Rbc8 3.Rxc8 is all fine.   But after 3...Bxg5, the problem's 4.fxg5 is best, but the alternate 4.Rxe8+ Qxe8 5.fxg5 is just a clear rook up, +6.59 Rybka.  And 4.Rxb7 is +3.96 Rybka.

So this problem could be saved if you just prune it after 3.Rxc8.

[richard]

I've just checked all of these against the (hopefully!) soon to be released updated problem set.

The new generator removed 90% of the problem's raised as having ambiguity issues (18/20).  Of the two remaining , one was left as is due to the winning alternative having a toga envaluation too small for the current generator to consider winning.

The second problem left in the set was replaced with a new problem with the same position , but an alternative move for one user's suggestion but unfortunately not Drahacik's (again toga didn't like the move enough for it to be considered winning by the current generator).

Overall this seems a fairly good result, as long as there is not a lot of problems with alternative lines just below the current winning line threshold there should be a noticeable improvement in problem set quality. I'll consider lowering the threshold for winning alternatives on the next verification run.

The results for each problem:
Killed
30952
20922
19008
24206
14423
33502
14335
8927
12821
31645
10536
11867
24694
14201
23458
12861
32403
15640
7665
19176

Kept
19329 Toga wouldn't give Drahacik's alternative of Rxe4 more than 2.21 after a 3 minutes, I suspect the passed pawn on d6 worried toga, but in the long run it doesn't seem to be much of a threat.

Replaced
23624 replaced by 37578 (which provides the alternative mrcm suggested in the problem comment, but still marked drahacik's Bf4 as wrong. Toga gives a much lower evaluation
to Bf4 than rybka (about +2 after a couple of mins, moving up to +2.42 after 6 mins)..after running for a few mins toga also liked h5 enough to have made it an alternative move if I'd waited that long).


Regards,
Richard.

[drahacikfm]

Too bad so many had to be thrown out, instead of keeping them with alternate moves.  I would have thought more would stay around now that you are allowing two alternate moves in every position.  Would allowing 3 alternate moves (and calculating 5 moves in every position, a 20% increase in generator time) allow you to keep a lot more?

Or maybe a lot of these were thrown out because the best move was not enough better than the second-best move?

[richard]

Too bad so many had to be thrown out, instead of keeping them with alternate moves.  I would have thought more would stay around now that you are allowing two alternate moves in every position.  Would allowing 3 alternate moves (and calculating 5 moves in every position, a 20% increase in generator time) allow you to keep a lot more?

Or maybe a lot of these were thrown out because the best move was not enough better than the second-best move?

I'm considering going up to 5 moves for the next generator run. The most common reason problems were thrown out was too many alternatives, however I also tightened up the "position already won at start of tactic" threshold so that change threw quite a few problems out also.  I also made a few more situations ambiguous that were previously being marked wrong (e.g. mate in 2 versus mate in 1 was previously marked wrong but is now an alternative). This increases the number of problems that are thrown out due to too many alternatives and going to 5 moves might bring some of those back. I think there is also at least one case in the material winning mates versus the best mate where I could probably be allowing an alternative instead of throwing the problem out.

There should have been no more problems throw out due to the best move being not enough better than the second-best move as this particular threshold should be the same as previous runs.

Regards,
Richard.

[angrypirate]

thanks Dr. H for posting these......


youre doing us all a service