Author: Simon Finn
Date: 15:47:31 11/13/99
Go up one level in this thread
On November 13, 1999 at 17:08:27, Eugene Nalimov wrote: >On November 13, 1999 at 16:17:17, Simon Finn wrote: > >>On November 13, 1999 at 01:07:29, Eugene Nalimov wrote: >> >>>Of course much more interesting task will be to produce DTM table that takes >>>into account 50 moves rule, but unfortunately I don't see good algorithm for >>>that. The simplest approach would be to add additional counter to the each >>>position, but that would mean 2 times more memory when compared to the first >>>version of my generator, and 16 times when compared to the current (not >>>published yet) version. If that is acceptable, modification can be done in a day >>>or so. >> >>What about? >> >>(1) Split the tablebases that contain pawns into multiple tablebases >> in such a way that any pawn move causes a conversion. >> For example, krpkr could split 6 ways: kr7kr+kr6kr+kr5kr+kr4kr+kr3kr+kr2kr. >> >> (This might be a good idea anyway, as it would reduce the memory >> required to compute these tablebases, and should also reduce the >> number of iterations required. The tablebases can always be >> re-aggregated for backawards compatibility with existing code.) >> >>(2) Run the current DTM algorithm, but terminate after 50 iterations. >> >>This works because the Nth iteration of the algorithm computes >>DTM under the restriction that any line that doesn't leave the current >>tablebase in N moves or fewer is a draw. The 50 move rule is just a >>special case of this. > >That's not true. From the second you are starting to talk 'doesn't leave the >current tablebase' you are talking about DTC, not DTM. Right. I want to compute DTM subject to the restriction that DTC <= 50. > Currently when program >passes "lost in N" to the preceeding positions, it doesn't know was >conversion/pawn move at move N-1 or 1. I don't think it needs to know this. If it can pass "lost in N" on the Mth iteration, then it knows that DTC <= M. That's enough (for M = 50) to implement the 50 move rule. Simon > >Eugene > >>Simon
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.