Author: Ricardo Gibert
Date: 12:33:10 08/03/99
Go up one level in this thread
On August 03, 1999 at 14:25:21, Robert Hyatt wrote:
>On August 03, 1999 at 10:39:37, Ricardo Gibert wrote:
>
>>On August 03, 1999 at 08:58:35, Robert Hyatt wrote:
>>
>>>On August 02, 1999 at 22:47:14, Ricardo Gibert wrote:
>>>
>>>>On August 02, 1999 at 20:49:56, Dann Corbit wrote:
>>>>
>>>>>On August 02, 1999 at 20:06:07, Ricardo Gibert wrote:
>>>>>
>>>>>>On August 02, 1999 at 18:15:27, Jeff Anderson wrote:
>>>>>>
>>>>>>>Could someone with access to the KQ v kr endgame tablebase tell me how many
>>>>>>>moves before Black mates in this position?
>>>>>>
>>>>>>I believe the tablebases are not optimized for finding the shortest mate. For
>>>>>>instance, in this case, it finds the shortest route to mate OR win the rook.
>>>>>>After winning the rook, it then finds the shortest mate from there. So it is
>>>>>>possible the "best" play generated by the table base is not the shortest mate.
>>>>>>It is also possible for a tablebase, organized differently, can give a solution
>>>>>>of a different length than another tablebase, though I would not expect that
>>>>>>here.
>>>>>>
>>>>>>So "perfection" is not guaranteed, but it does have the virtue of being optimal
>>>>>>in the light of the 50 move rule. In other words, it is possible that a position
>>>>>>is winnable without exceeding the 50 move rule, but the "shortest mate" would
>>>>>>exceed the 50 move rule. Of course, in that case it would not really be the
>>>>>>"shortest mate." I would be interested in seeing an example position of this if
>>>>>>someone has it.
>>>>>The Nalimov tablebase files have distance to mate. But as a confirmation, here
>>>>>is the output of Chest:
>>>>
>>>>With Nalimov, it is possible that is always the case for KQKR as memory
>>>>requirements are not a factor to generate a shortest mate EGTB for this ending,
>>>>but I've already given a reason why this is undesirable.
>>>>
>>>>The reason can manifest itself in the case where you have adjourned in an KQKR
>>>>ending (human vs human) and have only x numbers of moves to avoid the 50 move
>>>>rule. Using a tablebase for your adjournment analysis that gives "shortest mate"
>>>>instead of "shortest win of rook or mate" could be a problem.
>>>
>>>
>>>this can't possibly happen unless someone has a bug. KQ vs KR is _never_
>>>won in more than 50 moves. So at the point where the program reached this
>>>ending, it plays it optimally. Your case could not possibly happen. In
>>>other endings, this is a possible problem, but only if it is mate in > 50
>>>from the starting position where the ending is reached.
>>>
>>It appears you overlooked the phrase "human vs human". We both overlooked it can
>>also happen in Human vs computer where the human is the superior side.
>
>
>sure... but I then "don't get it". IE if the human can use the tablebases, he
>will mate optimally. If he can't, he has to either do it by himself or not at
>all. I don't quite see how the human could play for a while, then turn it over
>to a tablebase (which could certainly screw up there as optimal might be a
>mate in 20, while 40 moves have already been played and 20 more would take the
>game past move 50.
You "don't get it", because you have overlooked the word "adjourned".
I will give a more specific example: HUMAN plays an ending where he ADJOURNS. He
must mate OR make pawn move that preserves win OR capture that preserves win
within say 6 moves to avoid the 50 move rule. In his ADJOURNMENT ANALYSIS, if he
uses a SHORTEST MATE database, he may wrongly conclude he can't do it.
1. Instead of
distance to: mate
2. Better is single number representing
distance to: mate OR win preseving capture OR win preserving pawn move
(whichever comes soonest)
The 2nd way you ALWAYS win a winnable position. I find it hard to believe
Nalimov did his EGTB the way you assert (The 1st way). There is nothing I can
think of that would make the 2nd way listed above significant more difficult to
do. There is no good reason, I can think of, for using the 1st way in preference
to the 2nd one.
I hope this is more clear, otherwise, I give up.
>
>And this problem already exists. I believe the deepest mate in the 5 man
>files is 127, although Eugene can correct me here. And perhaps the deepest
>mates still have a capture or pawn push in them to defeat the 50 move rule.
>
>six piece files will be something else altogether, of course...
>
>
>>>
>>>>
>>>>Objectively speaking, it is never wrong to win the rook instead. Subjectively
>>>>speaking, your opponent is bound to resign after losing the rook anyway.
>>>>Shortest win can be faster than the shortest mate in that case.
>>>>
>>>>Your post is a little ambiguous. Are you saying Nalimov EGTB is a shortest mate
>>>>EGTB for all the 5 man endings? How would the tables be generated?
>>>
>>>this is correct. You start by enumerating all possible 5 man positions, and
>>>marking the ones that are "mate" and then working backward. Takes a lot of
>>>time to build, not time to probe.
>>>
>>>
>>>
>>>
>>>>
>>>>I would be surprised if all the endings covered by the Nalimov EGTB are of the
>>>>shortest mate variety. I would also be disappointed for the reason indicated.
>>>>Some endings (other than KQKR which a computer program can win in about 34
>>>>moves) would be "impossible" to win using such a TB due to the 50 move rule.
>>>>
>>>
>>>Eugene does use distance-to-mate in all files...
>>>
>>>
>>>>By the way, back in the days when a 20mhz 386 was a high end machine, I helped a
>>>>friend of mine learn how to play KQKR perfectly using the Thompson EGTB. I
>>>>myself learned how to do it within the 50 move rule. Not so easy against a
>>>>computer even though I am a master. Miles & Browne are 2 well known players that
>>>>have been embarassed. I don't think I can do it today.
>>>>
>>>>My friend showed up for a big tournament in Canada (St. Johns?) where a guy was
>>>>showing off his program that could play KQKR. I guess he got a kick out of
>>>>embarassing all the titled players there. My friend consistently won the ending
>>>>in the "optimal" number of moves. The guy couldn't believe it.
>>>>
>>>
>>>
>>>this ending is remarkably easy to win, even without databases. I ran several
>>>tests last year and was surprised that even at a couple of seconds per move,
>>>Crafty could win using no tablebases, against a version of itself that did.
>>>
>>>
>>>
>>>>>Reading job:
>>>>>W: Kg2 Rf2 (2)
>>>>>B: Kg4 Qa2 (2)
>>>>>FEN: 8/8/8/8/6k1/8/q4RK1/8 b - -
>>>>>analysing (mate in 14 moves):
>>>>>No solution in 14 moves.
>>>>>refu 1: Qa8+ Kf1 [ 13-]
>>>>>solu 1: Rf3 [ 4+]
>>>>>solu 2: Kh2 [ 12+]
>>>>>solu 3: Kg1 [ 10+]
>>>>>refu 2: Qd5+ Kf1 [ 13-]
>>>>>solu 4: Rf3 [ 4+]
>>>>>solu 5: Kg1 [ 4+]
>>>>>solu 6: Kh2 [ 13+]
>>>>>refu 3: Qa3 Kg1 [ 13-]
>>>>>solu 7: Rd2 [ 5+]
>>>>>solu 8: Rc2 [ 11+]
>>>>>solu 9: Rf6 [ 12+]
>>>>>solu 10: Rf4+ [ 4+]
>>>>>solu 11: Rf7 [ 13+]
>>>>>solu 12: Rf1 [ 8+]
>>>>>solu 13: Kh2 [ 5+]
>>>>>solu 14: Rf5 [ 6+]
>>>>>solu 15: Rf3 [ 4+]
>>>>>solu 16: Rf8 [ 4+]
>>>>>solu 17: Re2 [ 4+]
>>>>>refu 4: Qa4 Rd2 [ 13-]
>>>>>solu 18: Rf6 [ 11+]
>>>>>solu 19: Rf7 [ 12+]
>>>>>solu 20: Rf8 [ 13+]
>>>>>refu 5: Qa5 Re2 [ 13-]
>>>>>solu 21: Rf4+ [ 4+]
>>>>>solu 22: Rc2 [ 12+]
>>>>>solu 23: Rf6 [ 11+]
>>>>>solu 24: Rf7 [ 7+]
>>>>>solu 25: Rf1 [ 6+]
>>>>>solu 26: Rf8 [ 13+]
>>>>>solu 27: Rb2 [ 11+]
>>>>>solu 28: Rf3 [ 4+]
>>>>>solu 29: Kh2 [ 12+]
>>>>>solu 30: Rf5 [ 4+]
>>>>>refu 6: Qa6 Rd2 [ 13-]
>>>>>solu 31: Rf4+ [ 4+]
>>>>>solu 32: Re2 [ 3+]
>>>>>refu 7: Qa7 Rc2 [ 13-]
>>>>>solu 33: Rf4+ [ 4+]
>>>>>solu 34: Kg1 [ 3+]
>>>>>refu 8: Qa1 Rf8 [ 13-]
>>>>>solu 35: Rf4+ [ 4+]
>>>>>solu 36: Rf6 [ 4+]
>>>>>solu 37: Rb2 [ 3+]
>>>>>refu 9: Qb3 Kg1 [ 13-]
>>>>>solu 38: Rf4+ [ 4+]
>>>>>solu 39: Rd2 [ 5+]
>>>>>solu 40: Rf6 [ 11+]
>>>>>solu 41: Rf5 [ 5+]
>>>>>solu 42: Ra2 [ 3+]
>>>>>solu 43: Rf8 [ 13+]
>>>>>solu 44: Rf1 [ 6+]
>>>>>solu 45: Kh2 [ 5+]
>>>>>solu 46: Rf3 [ 4+]
>>>>>solu 47: Rf7 [ 4+]
>>>>>solu 48: Re2 [ 4+]
>>>>>refu 10: Qc4 Rd2 [ 13-]
>>>>>solu 49: Rf6 [ 11+]
>>>>>solu 50: Rf4+ [ 3+]
>>>>>solu 51: Rf5 [ 6+]
>>>>>solu 52: Rf3 [ 4+]
>>>>>solu 53: Rf7 [ 4+]
>>>>>solu 54: Rf1 [ 6+]
>>>>>solu 55: Re2 [ 3+]
>>>>>solu 56: Rc2 [ 3+]
>>>>>solu 57: Ra2 [ 3+]
>>>>>solu 58: Rf8 [ 13+]
>>>>>refu 11: Qe6 Rd2 [ 13-]
>>>>>solu 59: Rf4+ [ 4+]
>>>>>refu 12: Qg8 Rd2 [ 13-]
>>>>>solu 60: Rf4+ [ 4+]
>>>>>solu 61: Rf3 [ 4+]
>>>>>refu 13: Qb1 Rf8 [ 13-]
>>>>>solu 62: Rf4+ [ 4+]
>>>>>solu 63: Rf7 [ 12+]
>>>>>solu 64: Rf6 [ 11+]
>>>>>solu 65: Rf5 [ 4+]
>>>>>solu 66: Rc2 [ 3+]
>>>>>solu 67: Ra2 [ 3+]
>>>>>Time (user) = 856.00 sec (ca. 14.3 min)
This page took 0.01 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.