[net.chess] microcomputer chess claims, Elite, Prestige, etc.

mclure@SRI-UNIX@sri-unix (07/20/82)

Just had an interesting talk with that Dr. Irazoqui fellow ICD has
hired to do their micro-computer chess program testing and has
advertised his ICD consulting services in Chess Life.  The current best
micro is the Elite running on a 6502 at 4 mhz.  A number of conflicting
claims about its strength have been advanced.  Everything from 1771 to
1950 with lots of data to support the claims.  Irazoqui entered Elite
in a normal human tournament (11 rounds) where the players were rated
1680-2100 in its open section.  The time controls were 40/2 and 20/1.
The Elite won the section with a score of 8 pts which resulted in a
rating of 1915.  Irazoqui says 11 games is obviously not enough.
Further, he says that Elite is 1500-1600 positionally and 2000
tactically (all of this at 40/2).  He has other data to support his
claim that the Elite is really 1915 plus or minus 30 and discounts any
claims that it is much lower than this.  First, the Elo ratings seem
highly inflated over just a couple years ago.  Second, he performed a
tournament in which Elite, Scisys Mark V, Champion, 2.5, and Savant
played 10 games against each of the others.  2.5 had been rated in
human tournaments at 1484.  Champion had been rated by USCF at 1771,
but some tests by Irazoqui had concluded 1726 was a more likely figure
so he used that.  The results were:

	Elite	 30.5 pts
	Scisys	 19.5
	Champion 18.5
	MCT	 16
	Savant   15.5

which also results in a rating of Elite of about 1915.  However,
experiments by other unamed investigators have shown Elite should be at
least 100-150 points lower.  Unfortunately I don't think any of these
results have been published.  Irazoqui predicts that if Belle and Elite
were entered in the same long human tournament Belle should score not
much more than 300 points better than Elite.  Also, he feels that
Belle's positional level is qualitatively lower than Nuchess and that
the European micro-program Mephisto has at least the same positional
sense as Belle but probably closer to Nuchess.  He feels Nuchess has
the best positional evaluation of any program.  (an offspring of Chess
4.n)

The next unit coming from Fidelity is something called Prestige, which
Elite owners can upgrade to for a fee (no doubt rather expensive).
Kathe Spracklen claims the Prestige will sacrifice up to a pawn or the
Exchange for certain positional advantages and that it will sacrifice
pieces for strong attacks against the enemy King, and that its endgame
play is much better.  Apparently they are working on the Prestige
program from the ground up rather than just upgrading the program as
was done from Champion to Elite.  They have some Russian master and
teacher named Pochinski(?) working with them on it.  Also, the Prestige
will run on the same hardware as Elite but will have 16K of RAM instead
of 4K.  The intended release date is sometime in September.  She says
the Prestige will be entered in the upcoming U.S. Open and is
predicting a rating of close to 2000.

In this month's Chess Life, ICD has some ads offering a Computer Chess
Digest Annual prepared by Irazoqui available in a few months, with
all this information, game listings, and more. It seems worthwhile ($10).