Author: leonid
Date: 12:50:42 05/24/01
Go up one level in this thread
On May 24, 2001 at 10:30:17, Heiner Marxen wrote: >On May 23, 2001 at 22:45:11, leonid wrote: > >>On May 23, 2001 at 19:35:00, Heiner Marxen wrote: >> >>>On May 23, 2001 at 15:37:07, leonid wrote: >>> >>>>[D]RrqkqrR1/qPNbNPq1/BnqBqnB1/bqQQQqb1/Q2K2Q1/3B4/8/8 w - - >>> >>>This one has an even smaller effective branching factor. It has the same >>>set of solution moves, but it is a mate in 12: >> >>Here I was not sure where is the mate. Your finding is useful. 10 moves took >>already 6 hours and 28 minutes to say that no mate exist. For 11 moves it could >>be expected as 50 hours. Branching factor started to grow for my program just >>after 7 moves. It was between 4 and 5 move - 5.4. Mate was found for sure in 13 >>moves by selective. Took 7 sec. Big number of possible solutions make this >>position quit useful for possible logic verification. >> >>Cheers, >>Leonid. > >Hmm, Chest does not suffer from a bad branching factor, quite the opposite: > >depth seconds speed (= hash table speed up) ># 1 0.00 0.96 1- 0 ># 2 0.00 1.00 1- 0 ># 3 0.02 0.93 81- 0 ># 4 0.09 1.02 507- 0 ># 5 0.32 1.24 1941- 0 ># 6 0.92 2.75 5583- 0 ># 7 2.73 4.82 16205- 0 ># 8 9.95 5.71 58319- 0 ># 9 19.95 11.80 115465- 0 ># 10 39.25 17.14 227722- 0 ># 11 86.12 18.24 506580- 0 ># 12 163.57 13.09 978074- 0 > >Looks like the hash table kicks in quite nicely, as seen from the speed factor. >It reduces the effective branching factor down to around 2! > >Have a nice day! >Heiner Hi, Heiner! One question but about something else. Do you remember in your "mate2" search what pieces you look at first for responding side? Side that respond on attack. When I regarded quickly your description, I had the impression that we did almost the same work on this level even if names are different. Only I am curious to know what pieces you search first for responding side. Had the impression that it is not for the kings moves that your program look frist. My program search for one legal king's moves and only later one legal move for the other pieces. When king is searched later, speed is almost identical but slightly lower. All difference could be around 1%. NPS will be much higher. Could go up 25%. I speak all the time about what is in your program 3 plys search. I just went today to this part and looked if something in the past was not done in its best. Tried to change the generation of moves for in mate in 1 (in your its is mate in 2) but reached the same response that I remember was found many years ago. I still wonder how you did this part. Cheers, Leonid.
This page took 0.02 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.