Author: martin fierz
Date: 23:05:45 09/26/04
Go up one level in this thread
On September 27, 2004 at 01:18:43, Sergei S. Markoff wrote: >>wow sergei, you have even more questions than i do :-) > >:) > >>now, i have invented (and patented LOL) a new method for root move ordering. >> it is a combination of the two above. first, i shift new best moves to the >> top of the list. then i order the list based on node count but i only swap >> two moves at i,i+1 if the condition > >>nodecount[i+1] > X*nodecount[i] > >I have another idea. May be it's better to sort (by nodecount) independently two >groups of moves: the moves that was best moves in previous iterations and moves >that was not. The first group of moves must be searched first. And also may be >it's good idea to add some "expire" function that will decide that move from >group 1 now can be transfered to group 2. The easy way is to expire moves which >was best for early iterations (root_depth-4, for example). hi sergei, i also thought of something similar to your scheme - keep previous best moves at the top of the list always. but i think my version is better, because if one move generates a significantly larger search tree than the previous best moves, it can jump to #2 on the move list, which it can't in your version. the expiration is a good idea, mine also 'expire' by moving slowly down the list if they don't generate larger subtrees. the main appeal of my suggestion is simplicity. it is identical to what people are doing today, but with an additional constant. and no need to keep in mind what moves previously belonged etc. but of course if your version works even better i will implement it immediately for 1 more rating point :-) cheers martin > >>i need your results to compare :-) > >I'm experimenting on it :) I will publish my results.
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.