Author: Heiko Mikala
Date: 15:35:56 10/17/99
Go up one level in this thread
On October 16, 1999 at 12:49:29, Frank Phillips wrote: >I have been experimenting with DarkThought type pre-pre-frontier and >pre-frontier pruning with mixed results and wondered what other’s experience is >of this type of forward pruning. Unless I missed it, I noticed Crafty does no >forward pruning (except null move) and no longer even razors the last ply. Hi Frank! I tested DarkThought's so called Extended Futility Pruning (I'll call it EFT from now on) with mixed results too. The first results were very bad, but my experience is, that you need speed and search depth to make it work. Like Andrew said, this can be a killer in blitz games, especially if your program is slow and/or you're using a slow computer. And my results in test suites like BS or LCTII have been a bit worse than withouht EFT. Nevertheless, my program is very slow (around 20-30k NPS on a Cyrix MII 300) but I have been using all three stages of DT's EFT for some months now, and it worked fine. It let's you miss some things, but it also makes you search a lot deeper which seems to help more, than the missing of a few things seems to hurt. I have also tried Crafties razoring technique (which mentions "Kent et. al." as the original authors), but my experience with DT's combined pruning/razoring scheme has been far better. Maybe my implementation of Crafties technique is wrong, or my EFT implementation is too conservative, because I think in general Crafties scheme should be much safer, but my program makes more tactical mistakes with Crafties scheme than with DT's EFT. But, to be honest, I'm not really 100% happy with it. It's a bit too selective for my taste, and I'll continue to search for better things. Greetings, Heiko.
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.