how does one learn to find the best move over a good move?

[jadeiguana]

I'm pretty bad at chess, but I enjoy puzzles and the occasional casual game.  Usually when solving these puzzles I'll sit down and really think about each possible move that looks good, but I've found myself moving as soon as I see a winning move even if it's not the best move (in an attempt to speed up my play as long calculation is boring for casual play).  The latest example I came across was problem 8375.  I immediately saw that taking the rook would leave me up a rook and knight vs non-threatening pawns while missing the very simple mate right in front of me.

I am curious how important seeing a much better move is in terms of improving my play.  Perhaps more importantly, how should one go about training to avoid making a solid move over a game winning move?  (I'd argue in the case I gave that my solution is game winning, but obviously this happens to me in less drastic cases where the difference matters more.)

[richard]

I'd consider #8375 a borderline case. In the new generator I do try to avoid situations similar to this, however when the mate is so close I think requiring them is probably justifiable.  Having said that there are still a large number of solution that end in mate that have alternative material winning moves. I'm hoping many of these problems can be eliminated (especially the ones where the mates are more difficult and going for the material advantage over the tricky mate is therefore much more justifiable).

At the end of the day, I don't really think this issue impacts the usefulness of the site much.  In my opinion finding a tactic that was not the "correct" one still helps tactical vision, as does being shown the mating sequence if you missed it. I do however want to reduce the frequency of these problems as it can be frustrating to choose a move that is the right one to play in over-the-board play, only to be marked incorrect and have points knocked off your rating.

I think the "when you find a good move, look for a better one" advice should be ignored in cases where the "good move" is clearly winning and the better move can't be found quickly.  I think it could be argued in #8375 that the mate was relatively easy to find. While the other move would have won, you could have made a mistake in the endgame which finding the mate protects you from. ideally positions like this should probably reward both moves but the ability to provide that feature is still a long way off.  For now I'd try to be happy you found a reasonable move and try not to get too angry about the lost rating points :-)  The next problem set update will greatly reduce this type of problem.

Regards,
Richard.

[jadeiguana]

Heh, I think you misinterpreted my post.  I understand you are working on a solution to the problem of an easy win via material advantage versus a long complex mate, but I was not complaining about the fact that my solution was winning.  I was merely using that as an example to illustrate a problem that I have where I find a good move and miss a great one.  I was curious what methods people apply (or how they learn) to avoid missing a better move.  Assuming a good move has been found, how does one justify potentially wasting time looking for an even better move?

I hope I articulated my question better this time.

[richard]

Hi Jadelguana,

Sorry I probably came across as a bit defensive, I did understand there was a more general question in there, I probably didn't do a very good job of addressing it :-)  The reason I brought up the problem generator is that I think it is important to keep in mind that sometimes on chess tempo your "not the best move" may actually BE the best move from a pragmatic point of view.  I'll leave the question of how to know when it is appropriate to look for better moves and how best to look for those moves to members who are better qualified than myself (In my own game I'm happy with any move that isn't an outright blunder :-)   ).

Regards,
Richard.

[slacker00]

The way to find the best move rather than just a good move is practice, patience, experience, time, etc. 

Lately, I've been getting tripped up by finding a mate in 3, when there's a mate in 2 out there.  It seems like there's more  of these on Chess Tempo than anywhere else I've ever  solved problems.  It sucks to get a problem wrong when in  a real game, it would only take an extra move to deliver checkmate.  Now I'm just trying harder to see if there's a better  mate even after I find a mate in X.  I'm thinking about  writing down a list of these problem  types and  post it when I get a half dozen or so.  Some  partial  credit should be available for finding a lesser mate in X.

[richard]

Hi,

I should only be a few days away from updating the problem set. The new set has a large reduction in mate in N versus mate in N+1 situations.  I still tolerate some ambiguity, for example if there is a mate in 1 AND a mate in 2 I allow that position as I think it's reasonable for users to be able to spot an immediate mate.  For your example of finding a mate in 3 when there was a mate in 2, that problem should be removed in the soon-to-be-updated problem set.   For mates of 3 or more moves, any position which has a mate within 2 moves of the quickest mate will also be rejected.  This means that if there is say a mate in 3 as quickest mate, and alternative mates of 4 or 5 moves then the position will be rejected during problem generation.  Positions with a mate in 3 as quickest mate, but alternative mates of 6 moves or greater will be allowed, and you will therefore be expected to find the quickest mate in these situations.  I think this is a reasonable tradeoff between precision and realism.   Whilst finding a mate in 6 when there was a mate in 3 will still win you the game, it still means you missed something on the board and have increased the risk of making a miscalculation due to the length of your move sequence.

I'd still be very interested in seeing your list, as it will be a good test of how much of an improvement the new problem set has made.  I have my fingers crossed that the majority of the positions on your list will be knocked out of the new set, and any that remain will be a good indication of where I can make further improvements.

Regards,
Richard.

[slacker00]

Hi Richard.  I'm excited about your updates.  I absolutely agree that mate 3 should be found even when a mate in 6 is available.  But, I've gotten  tripped up so many times  with the mate in X versus mate in X+1.  I'm  actually getting more  persistent at looking for better mates,  now, which actually might be good. 

Anyway, I started compiling a list last night.  Here's 4 so far.  I'll keep compiling my list.

15088-not real bad, the mate in 3 is pretty good.  But I had a  mate  in  4

22893-they've got a mate in 4, I march the pawn for mate in 5.

14733-they had mate in  2, but I had mate in 3

3700-they had a mate  in 4, I had a mate  in 5.  1.Qe5+ Kg4 2.h3+ Kh4 3.g3+ Kxh3 4.Rh1+ Kg4 5.Rh4#

[richard]

Hi Richard.  I'm excited about your updates.  I absolutely agree that mate 3 should be found even when a mate in 6 is available.  But, I've gotten  tripped up so many times  with the mate in X versus mate in X+1.  I'm  actually getting more  persistent at looking for better mates,  now, which actually might be good. 

Anyway, I started compiling a list last night.  Here's 4 so far.  I'll keep compiling my list.

15088-not real bad, the mate in 3 is pretty good.  But I had a  mate  in  4

22893-they've got a mate in 4, I march the pawn for mate in 5.

14733-they had mate in  2, but I had mate in 3

3700-they had a mate  in 4, I had a mate  in 5.  1.Qe5+ Kg4 2.h3+ Kh4 3.g3+ Kxh3 4.Rh1+ Kg4 5.Rh4#

Thanks slacker00,

The good news so far is that I checked those 4 with the new set and all of them have been rejected so far.  Apart from Mate in 1 I should have removed all the Mate N+1 problems and any that remain after the upcoming update would be a bug that I could fix fairly easily. The trickier ones are the "mate on offer, but an easier material winning non-mate move exists" problems.  I should have removed a considerable number of these but not quite as agressively as the longer mates. I'll be interested to see how many of these still exist  (I do check a fair number myself, but with so many problems I can really only take a [hopefully!] representative sample on my own).

Unfortunately these improvements do come at a cost. A number of decent problems have probably been culled off and the new problem set is definitely slightly easier (as measured by average ratings). Hopefully I've left enough hard problems in for the quality improvement to be an overall gain.

Regards,
Richard.

[btherl]

Hi Iguana,

I think a lot of it is to do with the order you examine moves.  The first piece you should look at is the queen.  Then look to see if any other pieces can help the queen.  That will find you nearly all mates.  If there's no queen then you're probably looking for a material win, unless the king is trapped somehow and vulnerable.

Only after I've verified that the queen doesn't pose much danger will I start looking for material wins rather than mates.  In rarer situations you can mate without using the queen, but you can usually find such positions by noticing the enemy king is in an unusual or exposed position.

Another rule of thumb is to look at checks first, which are much faster to calculate.  Only after you've verified that no checks win should you look at non-check moves (which allow your opponent many possible responses).