Author: James T. Walker
Date: 21:37:21 02/18/03
Go up one level in this thread
On February 18, 2003 at 19:43:44, Charles Worthington wrote: >On February 18, 2003 at 19:41:56, Charles Worthington wrote: > >>On February 18, 2003 at 19:31:01, Charles Worthington wrote: >> >>>On February 18, 2003 at 18:14:37, James T. Walker wrote: >>> >>>>On February 18, 2003 at 14:54:48, Charles Worthington wrote: >>>> >>>>>I agree Michael. The engine making the decision hasn't a clue. And the >>>>>performance will only be as strong as the strongest engine in the decision >>>>>making process and then when you take into account that they are all sharing the >>>>>resources of a single system the idea seems unfeasable in all but the longest og >>>>>games. >>>> >>>>This is an interesting topic and I did a small amout of testing of the triple >>>>brain when I got my first Shredder. Nothing spectacular came of it but the >>>>problem was that I didn't have any good engines as a second engine to Shredder. >>>>If the two engines are running at say half speed then they only lose about 50-60 >>>>Elo points I guess. Seems to me that if one can prevent the other from making a >>>>big mistake then it might be helpful. How much I don't know but some testing >>>>needs to be done in this area. Like maybe one engine with null move and one >>>>without? The availability of Ruffian and a few other nice programs might make >>>>some new test worthwhile. >>>>Jim >>> >>> >>> >>> >>>That is an interesting point that you make Jim and is certainly worthy of >>>consideration. >Perhaps I will experiment this triple brain when my Xeon system arrives. If the >setup will allow me to dedicate one processor to each of the brain engines then >i can use the hyperthreading to run two threads per brain to offset the weakness >of running two engines at once a little. However I am still somewhat sceptical >that this will outperform one program running on all four threads. The slowdown >caused by the dual engines will be significant even on a powerful platform like >the Xeons. Plus, lets face it....Shredder 7 is a fine engine but it is not >exactly the fastest thinker around. On Dual AMD systems it performs noticably >better than on single cpu systems but it is still only in the 800-900 kNs range. >The question is can an already-slow engine benefit from the assistance at the >cost of running even slower? Plus does Shredder honestly need the help? Shredder >has a considerable amount of knowledge as it is. I have some other tests to >conduct when the system arrives but I will conduct the triple brain test >afterward and post the results here for you. FYI. I just ran a 20 game match (G/10min) of Shredder 7 vs Ruffian/Yace/Triple Brain on 2 XP2400+ machines with auto232. The results were very disappointing. The Triple Brain group only managed 2 wins & 3 draws out of 20 games(3.5-16.5). This after winning the very first game. It looks to me after only watching a couple of games that the triple brain only selects which engine gives the best score for each move. If that's true then this simple decision making function is not very realistic. It also appears that these two engines don't work well together. :-) Later, Jim
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.