Problem 38408: a narrow distinction

[revenant]

Just curious if this one will be thrown out by the next generator run. What does Toga say the actual material difference is between 1... exf4 (the currently required solution) and 1... Qd4 (the move I tried first as it seems simplest; thank goodness it's an accepted alternative).  And, what happens if you throw some extra CPU at the position?  Does 1... exf4 stay solidly "best", or does 1... Qd4 perhaps pull up neck-and-neck with it?

Crafty 22.0 confirms that 1... exf4 certainly does win (see the list of cases in my comments at the problem itself), but then as an exercise I gave it 1... Qd4 and with 20 minutes of 2.4GHz CPU it produced:

2. Kf1 Qxf4 3. Qd3 Qc1+ 4. Ke2 Rc2+ 5. Qxc2 Qxc2+ 6. Ke3 Qxg2 7. Rb1 Qf2+ 8. Kd3 Qxf3+ 9. Kc2 g2 10. Kb2 e4 11. Rc1 e3 12. Rc7+ Kf6 13. Rc8 Qf2+ 14. Kb3 g1=Q 15. Rf8+ Ke5 16. Re8+ Kd4 17. Re7 -18.53

Which tells us that the difference is academic.  Whether we are judging by how much material is won how fast, or by how fast we reach checkmate, the computer cannot demonstrate an appreciable difference between the key first moves.  The truth is beyond the practical horizon of an exhaustive search, and even if it weren't it would be meaningless.

The problem now feels somewhat bogus, which is a shame because both lines have valuable lessons to teach.

For a composition which I wonder whether *any* of today's computers could solve, see the "Dance of the Elephants".  The mate is forced but it's 28 ply away from the initial position.  Is that in range of dedicated processors?

http://www.chessbase.com/puzzle/puzz16b.htm

http://www.chessgames.com/perl/chessplayer?pid=10156&kpage=1

[richard]

Hi revenant,

Initially toga sees the difference as just over +3 (about +8 versus +11). This then expands to +11 after about 2 mins (+10 versus +21). After 30 mins toga hasn't seen Qd4 catch up yet (it is at +11.37).  I haven't looked at the lines but based on the evals it seems toga sees a promotion to queen in one line but not the other. Qd4 may eventually catch up, one of the issues with positions like this is that the second best move still takes LOTS of material and once you are that much material up, gaining more material is inevitable, so if you leave the engine running for a LONG time it gets a chance to see the next material gain occur.

Based on the evals I'm seeing out of toga right now, this problem is unlikely to get rejected by the new generator.

Regarding the dance of the elephants there are some computer solvers that are dedicated to finding mates in N, they may have some chance with that position.

Regards,
Richard.

[slacker00]

Problem Blitz Rating: 1455.9
Blitz Average seconds: 13.3
Blitz attempts: 13
Blitz success rate: 61.54%

I think this is the interesting part.  61%?!  These people obviously have no clue as to what's really going on in this problem.  More user moves might have helped cull out the clueless, but I'm not sure how forcing the main line is.  There's simply some wierd chemistry with this problem where the answer seems to miss the point of the tactics.

The problem itself is extremely difficult, needing to see at least a half dozen moves or more in at least a half dozen variations.  Eventually, I just said "screw it", and made the correct move, despite only ruling out the major alternatives and not seeing lines like 5.Kg1 Rc1+ 6.QxR Qb6+ 7.Kf1 Qf2#.  My mental "horizon" really breaks down right around 8 or 9 plies.

 

[drahacikfm]

Yes, there are a lot of problems with low ratings that are actually very difficult.  It's usually because the problems stops after you make the first move and you are marked correct.  But the tough part comes when the opponent replies with a tricky move, and you have to find some very difficult moves to even avoid losing.  But you don't have to find those moves because the problem stopped after the easy first move.  So the rating is low.

On many of these problems, I just cannot figure out what I am going to do on move 2, after I take that hanging piece on move 1 and the computer makes the tricky move I am thinking about... often it's pushing a pawn threatening to queen, and I can't figure out how to stop the queen.  After several minutes of not seeing how to stop the queen, I give up and just take the hanging piece on move 1, and boom, the problem is over.   I don't have to stop that pawn from queening.  The people who solve the problem fast are either geniuses who quickly see how to stop that pawn, or people who didn't even consider the pawn push.  The second is most likely.