Author: Robert Hyatt
Date: 11:34:15 01/21/01
Go up one level in this thread
On January 21, 2001 at 13:03:47, David Rasmussen wrote: >On January 21, 2001 at 10:44:15, Robert Hyatt wrote: > >>On January 21, 2001 at 06:27:57, David Rasmussen wrote: >> >>>On January 20, 2001 at 20:28:28, Robert Hyatt wrote: >>> >>>>On January 20, 2001 at 18:30:45, David Rasmussen wrote: >>>> >>>>>On January 20, 2001 at 10:34:47, Robert Hyatt wrote: >>>>> >>>>>>On January 20, 2001 at 02:38:57, Mark Longridge wrote: >>>>>> >>>>>>>Some of the programs, crafty and gandalf come to mind, let their clocks run down >>>>>>>pretty low (say as low as 30 seconds) near where the game would normally be >>>>>>>close to over. But if the other player is just shuffling wood back and forth and >>>>>>>is playing with an inc, that player can build up a huge time advantage. Crafty >>>>>>>tries too hard to avoid the 50 move rule, and all of a sudden it's got 25 >>>>>>>seconds left and a lost position. >>>>>>> >>>>>>>I bet a lot of GM's and some programs do this on purpose. I don't see why crafty >>>>>>>shouldn't go for the 50 move rule instead of a silly pawn push, especially when >>>>>>>it's time is so low. Now the silly draws are becoming silly losses. >>>>>> >>>>>>If I saw this happen I might be concerned. However, crafty does _not_ let the >>>>>>human get way ahead on time. It has specific code to prevent this by speeding >>>>>>up itself. And it _never_ loses on time, ever... >>>>>> >>>>>> >>>>> >>>>>I've seen you claim this before, but still, it has lost several times on time to >>>>>my program, playing 8 moves a second on the same machine through winboard. >>>> >>>> >>>>Then there is a big problem on your machine. To stress-test crafty, I play >>>>games with the time control 999 moves in 1 minute. The games go to 100-200 >>>>moves with _no_ time problem at all. >>>> >>>>I notice you said one machine. That's not a reasonable test. Make one >>> >>>I know, that could be the problem. >>> >>>>program's hash too big and that ends it... paging a program in and out >>>>is not going to work... >>>> >>> >>>Well, that's certainly not the problem because >>> >>>1. Both programs have little hashtable and overall memory consumption, about 10 >>>MB in all, and there is 192 MB of RAM >>> >>>2. There is no disk activity at all >>> >>>3. This is not Windows98 we're talking about. I've tried it on NT, Win 2000, >>>Solaris, and Linux, all with the same result. Maybe winboard is the problem. >> >> >>No. If you are seeing it lose 2 8 games on linux, you _definitely_ have >>something broken. I haven't seen Crafty lose a game on time in years, >>unless I accidentally break something while testing. In a 1 0 game it >>can _easily_ play over 1000 moves with no problem. I have seen it play >>250+ move 2 0 and 3 0 games on ICC. > >2 8 ? I said 8 moves a second, that is > >level 8 0:01 0 > >in winboard lingo. > >I don't believe that I have anything broken as I've tried this on at least 10 >vastly different machines running different versions linux. Crafty plays "1 0" all the time on ICC against humans. It _never_ flags. I would think 1 0 is far more difficult than what you are playing above. I know of several crafty clones on ICC that play 1 0 all the time. The _only_ ones that have problems try to use 95/98. I am unaware of _any_ linux version that has any problem with time in any way. I assume you are not doing something funny in the .craftyrc file? You should try removing it completely and see how it does. Of course, back to the two programs on one computer... anything can happen there as playing a very fast game makes no sense in such a set-up unless you are just trying to stress-test things. However, to confirm what I say, I'll be happy to let you play crafty or scrappy a series of games at any time control you want... to confirm that it will _never_ lose on time unless it is unintentionally screwed up...
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.