Schwamb@ICS.UCI.EDU ("Karl B. Schwamb") (11/10/90)
Does anyone out there know of any empirical results that give a feel for how search ply may be related to USCF rating? I realize, of course, that this is not too useful without considering the static position evaluator and peculiarities of the search method. Still, I was wondering if there were some general qualitative statements that could be made about this relationship. -Karl Karl B. Schwamb schwamb@ics.uci.edu Information and Computer Science University of California, Irvine 92715
tjbryce@amherst.bitnet (11/10/90)
In article <9011091027.aa03480@PARIS.ICS.UCI.EDU>, Schwamb@ICS.UCI.EDU ("Karl B. Schwamb") writes: > Does anyone out there know of any empirical results that give a > feel for how search ply may be related to USCF rating? I realize, > of course, that this is not too useful without considering the > static position evaluator and peculiarities of the search method. > Still, I was wondering if there were some general qualitative > statements that could be made about this relationship. > > -Karl > > Karl B. Schwamb schwamb@ics.uci.edu > Information and Computer Science > University of California, Irvine 92715 Each additional ply of search adds about 200 points to a programs plying strength. (sources: Scientific American, October 1990 and the book Chess Skill in Man and machine, as well) APPROXIMATE PROGRAM PLY RATING Belle 5 1550 Belle 6 1860 Belle 7 2080 Belle 8 2200 Hitech 9 2380 Deep 10 2580
kpc00@JUTS.ccc.amdahl.com (kpc) (11/16/90)
In article <11155.273b8b99@amherst.bitnet> tjbryce@amherst.bitnet writes:
Each additional ply of search adds about 200 points to a programs
plying strength. (sources: Scientific American, October 1990 and
the book Chess Skill in Man and machine, as well)
How about in go? (N.B. I think that some approaches use shallow or
even no search, other than perhaps for ladders.) Any interesting
patterns for kyu ratings?
P.S. What's the status for PD go program frameworks? What are the
good ones out there, and are there any multilingual ones (e.g. C or
C++ and a high level language combined)?
--
If you do not receive a reply from me, please resend your mail;
occasionally this site's mail gets delayed.
Neither representing any company nor, necessarily, myself.dsa@dlogics.COM (David Angulo) (11/21/90)
In article <KPC00.90Nov15174201@JUTS.ccc.amdahl.com>, kpc00@JUTS.ccc.amdahl.com (kpc) writes: > In article <11155.273b8b99@amherst.bitnet> tjbryce@amherst.bitnet writes: > > Each additional ply of search adds about 200 points to a programs > plying strength. (sources: Scientific American, October 1990 and > the book Chess Skill in Man and machine, as well) Well, this is a rule of thumb and as any such generalization goes, take it with a grain of salt. In the good, brute strength programs, the same program when an additional ply of search is added (without taking into account time factors), might increase in strength by this number. There are a lot of other factors, however. There is a rec.chess group (I think) that talks about these things a lot - if you want more info, I'd suggest you post there. > > How about in go? (N.B. I think that some approaches use shallow or > even no search, other than perhaps for ladders.) Any interesting > patterns for kyu ratings? > Well, the same thing would apply. Yes, many programs do no minimax tree searching at all. One would expect that if such searching would be added, the ratings would go up similarly on a similar scale. Note, however, that the scales are NOT similar. The internation and American chess rating systems are both based on statistical normalized distributions (bell curves). The kyu system is based on handicaps between two players of different strengths. > P.S. What's the status for PD go program frameworks? What are the > good ones out there, and are there any multilingual ones (e.g. C or > C++ and a high level language combined)? There is GNUgo (see the GNU... news groups) as well as a GNUchess program. I believe these are written in g (?) a c lookalike language. -- David S. Angulo (312) 266-3134 Datalogics Internet: dsa@dlogics.com 441 W. Huron UUCP: ..!uunet!dlogics!dsa Chicago, Il. 60610 FAX: (312) 266-4473