Author: Poschmann
Date: 12:25:43 10/28/01
Go up one level in this thread
<snip> >I'm pretty sure it is always possible to determine a true mate-in-n value >from such a tablebase. > >- Suppose the current position has a value of +5 (so the true value is somewhere > between 4*N+1 and 5*N). >- First do a 1 ply search -- that is, test each of our moves. If you find > one that is -4, then you pick that move. In this case you know the original > position must have an actual value of 4*N+1. (Look at it this way, if N=4, > then you know the value is from mate-in-17 and 20. If it is possible to make > a move and hit a position that is between loss-in-13 and 16, then the original > position MUST have been mate-in-17. >- If you don't see a -4, then take each of the moves that produces a -5 position > and search deeper (anything -6 or more can be discarded). >- Keep increasing the depth until you can produce a -4. >- The actual mate-in-n value can be determined by how deep you have to look; > the maximum depth you might have to search is determined by N. > >So I'm pretty sure it is always possible to play optimally once you actually >get a tablebase position. I'm not sure how many tablebase lookups will be >required to do it though. > >- paul You are right. You can determine the true distance to mate by a deeper search (brute force! as Uri Blass mentioned in the other answer). But I suppose that there are a lot of tablebase accesses to do so (several thousands with N=4!?). Its the question, what is faster increasing in future: size or access time of bulk memory? Ralf
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.