Author: Gian-Carlo Pascutto
Date: 13:01:56 08/27/01
Go up one level in this thread
On August 27, 2001 at 14:37:34, Pat King wrote: >I wasn't planning on doing straight pvs exactly. Any re-searches would come from >mtd, all min-window. Rather, i was thinking of using the move pv[][] structure >that many pvs implementations use for my move ordering in place of the hash >table, or to supplement it. You will crash and burn when trying to use MTD without a perfectly working hashtable, and there is no sense in trying to hack around that. If you need the pv[] structure to keep the pv's intact or to improve the move ordering, then that is a signal that your hashing is broken and it _will_ cost you dearly in MTD. Also keep in mind that during the MTD search you will not have the PV's you'd expect. So if you are worried about having weird-looking PV's with the MTD it's probably ok. You should end up with a decent one at the end of the search though. If you know your hash table is broken, look no further. Fix it first. Improve it. MTD cannot beat PVS without a good working hash. If I am sounding repetitive, it's because I am utterly shocked at your suggestion. Please _read_ Aske Plaat's full thesis. It's available on the web, well written, and it will learn you everything you need to know. I will sleep better if I know that there is at least one more chessprogrammer out there that actually understands what he's doing. And it _will_ pay off for you in the future. -- GCP
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.