Author: Robert Hyatt
Date: 13:12:21 04/04/98
Go up one level in this thread
On April 04, 1998 at 08:48:03, Stanislav Goldovski wrote: >Dear CCC-members, > >I have a few questions concerning programming >transposition tables. Any help is highly appreciated! > >1. What possibilities are there to increase the efficiency of > hash tables for a given size? For example, I read that > one shouldn't let horizon positions overwrite the more > important positions near the root. Is that correct? > If yes, how is that implemented? when you store a position in the table, store the depth *below* that position as well, commonly called the "draft". then when you do a store into the table, you compare the depth remaining at this position in the search to the draft for the entry in the table you are wanting to replace, and keep the one with the largest value.. > >2. What app. speed increase should I expect in the middlegame, > due to the better move ordering from the hash table? here you have to run tests to see, because this can vary wildly depending on what you are doing for move ordering. If the hash move is your main ordering approach, it will help. If you have other ideas to use, it's not as important... > >3. I heard that hash tables are more effective at higher depths. > In other words, they will bring more with higher time controls. > Is it so, and if yes - why? the deeper you search, the more permutations in move order that lead to the same position there are... > >Thanks a lot in advance for your help, >Stan
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.