🚀 go-pugleaf

I have done (part of) the CHESSTEST with GnuChess
with many different timings from 1 second to 720
seconds. The results show that the test is rather
insensitive to this large increase in computing
effort.

Remarkable is the (positional) bellin test; the
program hardly improves it score with increasing
time. Note that on the tactical larsen1 test set,
GnuChess gets a 90% score!

If we plot total score against time, we get (very)
roughly a straight line.

Specifications:
GnuChess 4.0 pl 74 + patches, with default
Makefile options, with a bug in chesstest
resolved.

Running on a 50 Mhz Sparcstation LX with 32 Mb
memory. During the test GnuChess was the usually
the only running program, but was used at some
occasion.

 
  test         		1 sec.  2 sec.	4 sec.	8 sec.	22 sec. 60 sec.	240 sec.720 sec.	total
CHESSTEST/bellin          5       5	  6	  7	  7	  7	  6	  6		 33      
CHESSTEST/finecomb        7       7	  8	 10	 10	  9	 13	 14		 50      
CHESSTEST/jenoban        87      94	101	102	120	121	133	138		202     
CHESSTEST/larsen1        58      60	 66	 74	 77	 86	 87	 90		100     
CHESSTEST/larsen2         7       9	  9	 13	 14	 16	 12	 12		 48      
CHESSTEST/openings       29      29	 30	 29	 33	 38	 38	 41		110     
CHESSTEST/positional     34      34	 35	 42	 43	 52	 56	 62		154     
CHESSTEST/sherev          9	  8	  8	 11	 11	 12	 12	 14		 60      

total           	236	246	263	288	315	341	357	377		757


-- 
Michel Grimminck, Computational Physics, University of Amsterdam
http://www.fwi.uva.nl/~grimmink