Author: Howard Exner
Date: 11:25:59 11/16/98
Go up one level in this thread
On November 16, 1998 at 11:58:48, Komputer Korner wrote: >On November 15, 1998 at 20:07:05, Howard Exner wrote: > >>Although Rebel doesn't have fancy names for different chess playing >>personalities it does offer lots of flexibility in the various settings, >>AntiGM, Selection, Combination and Playstyle. Some settings I notice can >>detect best moves a ply earlier, usually when one sets Selection lower to say, >>3 (for endgames) and 4 for middlegame positions. These positions for example >>are solved much quicker when selection is set to 4, AntiGM=Strong and >>playStyle = Active or Aggressive. >> >>4r1k1/p1pb1ppp/Qbp1r3/8/1P6/2Pq1B2/R2P1PPP/2B2RK1 b - - id Ren5 - Qxf3; bm d3f3; >>rr3qk1/1b3p2/4pPp1/1pppP3/p6Q/PnPP2P1/1P3PB1/R3R1K1 w - - id Kh2 - Cmb; bm Kh2; >>8/p5Nk/3B2pp/8/2PP1pn1/4p3/P3P1Bq/R2R1K2 b - - id Nf2 - Cmb; bm Nf2; >> >>Overall my favoutite choice is AntiGM=Strong, PlayStlye=Agressive and >>Selection=Normal. I like Rebel's use of Normal Selection setting as it >>adapts itself to the position. Sometimes it misses as in the above 3 examples >>but overall it seems to hit on the right choice based on the positions I've been >>testing. >> >>I'm wondering as Computers get faster will the Selection settings >>get smaller. Too high a selection setting sometimes makes Rebel not consider >>certain moves for more than a brief passing. Will added speed allow the program >>to spend a bit of extra time on a wider range of moves so as to not miss >>something (as in the trend of Rebel 10, which according to the description on >>Ed's homepage that the selection settings where reduced overall)? >> >>Anyone else experimenting with the different parameters that chess programs >>offer? I am aware of the CM series alternate settings. How about other >>programs? > >Yes, added speed will allow selective searchers such as Rebel to consider moves >(which are razored) to a deeper full width level. The controversy is whether >brute force type programs will gain as much from the increased speed. I believe >they won't. Rebel 10 isn't as selective as Rebel 9 but with increased computer >speed, I believe that it should benefit more in comparison to the brute forcers. My observation on Ed's making Rebel 10 less selective than Rebel 9 is positive so far in the test positions I've been running (I should mention that one aspect of chess I've always liked are solving chess combinations. Most of those books are in my chess library so when I collect these positions for entering in the computer I am able to study them also while the computer churns away in the background). For the machine I use, a K6-233, I'm finding that in general Rebel 10's solution times are moving into the tournament time control range (2-7 minutes) from both ends. "Both ends" meaning that many quick solutions (under 30 seconds) for Rebel 9 take a bit longer for Rebel 10 while at the other end many longer solutions (7-15 minutes) for Rebel 9 are found quicker. >However there is a limit to this increased benefit. At around 16-20 ply full >width , I believe that the 2 styles of programming converge in obtaining >benefits from increased speed. Note that we have a long way to go before the >selective searchers like Rebel 10 reach 16 ply full width so the next couple of >years should see that type of program really benefit from faster computers. >-- >Komputer Korner
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.