Author: Bob Durrett
Date: 17:04:33 12/21/03
Go up one level in this thread
On December 20, 2003 at 10:53:56, Jay Hysenbeg wrote: >hi, >i was wondering what happens when u use low hash for a long time control. to the >point where the hash is normaly filled before the eng has made its move. >thank you >jay h. Your use of the words "long time control" suggests that you may be thinking about situations in which it is practical and desirable to use long time controls. Most humans, when playing in real time against the chess program, would not wish to wait a "long time" for the program to select it's next move. The use of very long time controls in such games would be neither practical nor desirable. On the other hand, automated analysis [I call this "overnight analysis"] would not require the presence of the human and no human would have to "sit and wait." Very long analysis times could be set for automated analysis. The engine could do it's work while it's human user was doing something else, such as sleeping or at work. An interesting question is what happens to the contents of the hash table during automated overnight post-mortem analysis of a game. If the game contains many errors on both sides [as is typical of an amateur blitz game], and if the analysis is set to examine moves where the evaluation of the previous position in the game differs from that of the position being examined by only a very small amount, then the product of the "overnight analysis" can be considerable. In this case, it seems to me that size of the hash table, and the way it is used, are especially important. Testing the effects of hash on fast games would not address the "long time controls" question. There seems to be a huge difference in the way hash tables are used in playing games versus automated post-mortem analysis of games. My impression is that hash tables are much more important in the latter case. Bob D.
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.