Author: Dieter Buerssner
Date: 09:12:12 09/25/01
Go up one level in this thread
On September 25, 2001 at 09:46:05, Robert Hyatt wrote: >Remember what I said earlier. The collision probability is not the only >thing you have to deal with. I agree, and this was the whole point of the experiment. You may have seen, that even with a ridicously small hash value of 12 bit, there was essentially no change in the results (despite the detected collisions). >You have to have _enough_ collisions so that >the root score changes. This is a harder thing to cause. Your very small >table is probably helping you avoid this rather than hurting. No. The table got filled in about 5% of the search nodes. Thus through almost the whole search time, almost every probe can be a collision. >Because you >are probably depth-preferred and are not adding much to the table after the >first few seconds, at a guess... Accumulating large depths will make the collisions more critical instead of less critical. When probing, and you get a wrong result with small depth, this will influence move ordering. When larger depths accumulate in the table, you will get more and more wrong cutoffs in the search. After all, not the stroring relpacing is the time, that is critical, but the probing. Regards, Dieter
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.