Author: Heiner Marxen
Date: 17:04:39 04/29/01
Go up one level in this thread
On April 29, 2001 at 18:45:05, Paul wrote: >On April 29, 2001 at 14:07:27, leonid wrote: > >>It could be so. Real number of moves I don't know. For sure, I can say that this >>is mate between 10 and 13 moves. If 12 moves, indicated by Paul, is not the >>guess but what really its program found, then it is mate in 10, 11 or 12 moves. > >No ... my program also found a mate in 13 ... just tried to have some fun :) > >WM13 10 Qbxd8+ Qxd8 Qhxd8+ Nxd8 Qxd8+ Qxd8 gxf8=Q+ Rxf8 Qxf7+ Rxf7 h8=Q+ Nf8 >Qxf8+ Rxf8 Rxd8+ Kxd8 Nxe6+ Bxe6 Bg5+ Kd7 Qe7+ Kc8 Qexb8+ Qxb8 Qxb8# > >>I found mate for sure by selective in 13 moves. My program also searched >>position as far as 9 moves and indicated that mate in 9 do not existe. This is >>from where I became certain that this position is not easy. Will try, if I will >>have the chance, with my 600Mhz to look even 10 moves deep. Expected time is >>around 18 hours. It is difficult when you don't have free laptop for this work! > >So ... until someone finds something better, we're both champions my friend! ;) > >Paul Hi mate freaks! Within 1.5 hours Chest (K7/600, 350MB) found that there is no mate in 11. Since the effective branching factor here appears to be moderate for me (from #9 to #10: 3.83, and from #10 to #11: 3.545) I have just started the mate in 13. Expected time is around a full day, so do not hold your breath ;-) A quick glance through the statistics of the #11 computations shows, that the "anti" heuristic is very effective: within every mate in 3 the defender not only tries to find a good checking or capturing move (to reduce the size of the search tree), but he first tries to deliver a "mate in 1" to the attacker. For this #11 computation over 41% of all such mate attempts came up with a solution! That in turn does not only refute the last attacker move as a #3 candidate, but does refute it for any depth. As a consequence, from the 88 legal first moves of the attacker were left mere 17 on the top level as candidates for the mate in 11, since the #10 computed before that (iterative deepening) found all others more or less directly to be refuted for _all_ depths. Also, when the attacker has to move, he finds himself in check in over 93% of the cases. That helps, too :-) Chest for Champion! :-)) Heiner
This page took 0 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.