Fun_People Archive
8 Sep
Kasparaging

Content-Type: text/plain
Mime-Version: 1.0 (NeXT Mail 3.3 v118.2)
From: Peter Langston <psl>
Date: Fri,  8 Sep 100 16:31:16 -0700
To: Fun_People
Precedence: bulk
Subject: Kasparaging

X-Lib-of-Cong-ISSN: 1098-7649  -=[ Fun_People ]=-
X-http://www.langston.com/psl-bin/Fun_People.cgi
Forwarded-by: magister poti
From: "Joshua W. Burton"

Andrew A. Gill writes:

> This was exactly opposite Jon's point, vis.:
>
> | He's not even that good at chess.  People seem to groove to the notion
> | that playing chess is like sprinting a mile: over time mankind just
> | naturally gets better, thanks to inexorable Progress in health,
> | steroids, and space-age insoles.  So of course today's world champion
> | is the fastest human being ever.  The time to run a mile is
> | quantifiable; chess isn't.  There's no way to pit Kasparov against
> | long-dead chess masters, but many chess experts agree that many past
> | champions could have handily spanked Garry.
>
> So, my question still stands.
[I.e. how would Kasparov have done against earlier chess masters?  -psl]

This is a question that interested me briefly but intensely in college
(early 1980s) when my roommate was a rising International Master with a
Kasparov number of 2, and Kasparov himself was the butt of a shaggy dog
joke by the Soviet powers-that-checkmate, who had kept him from the world
championship for several years by the simple expedient of never letting
him play Anatoly Karpov.  At that time, before Kasparov's formal ascension,
I bet Jonathan a ruble that Kasparov would be the last human world champion.
Unfortunately, the ruble didn't last out the bet.

Anyway, it is _not_ true that chess is unquantifiable; the USCF and FIDE
rating systems are large statistical universes, whose properties have been
extensively studied.  There is definitely a caveat when you talk about long
periods of time (say 40 years or more) over which the state of the game,
especially opening theory, has actually advanced.  Alekhine or Capablanca
would be swept off the board by any modern IM or GM, simply because they'd
be out of their book while the opponent was still on firm ground.  But this
caveat is a bit like the common sense limit on how far back you can use
price inflation numbers, which like chess ratings are well understood over
any short range of years.  Jane Austen's pound sterling was about $100
US-Y2K for buying things, or $1000 for hiring people, because England was
a third-world country back then, but you can't use that to price a headache
cure or an evening's world-class music in your drawing room across two
centuries.

So...given that raw chess ability will only compensate for primitive opening
and endgame theory to a limited extent, how accurate _is_ the chess rating
system in comparisons where that effect is minor?  I don't have the
citations at my fingertips this decade, but basically the only systematic
inaccuracy that anyone has identified is a steady ratings deflation of
between 1 and 1.5 points a year, caused by the fact that points are
conserved in tournaments.  You see, most new players come in at a low
rating, take points from other people as they advance, and then steal them
out of the system by dying or retiring.  That's a small effect:  a 200-point
underdog can still expect to win roughly a game out of five, so getting a
rating right to within 10 or 20 points is a fairly impressive statistical
achievement.  By the way, in case anyone is wondering, my former roommate
and physics lab partner, Jon Yedidia, outranked me by over 1100 points, so
there was literally no way we could interact meaningfully over the
chessboard.  He could beat 24 players of my caliber simultaneously,
blindfolded.  We tried playing with a 30-second clock for him, but he did
all his thinking on my time, so it came down to mere physical coordination
snatching my pieces from me.  Very depressing, especially since I never
thought him a much brighter physicist than myself.

So, Jon with a rating in the mid-2400s knew exactly where he stood with
respect to the giants of the age, and what his (slim) odds of beating them
were.  I recall that at that time Kasparov was somewhere around 2740 (all
ratings here are FIDE, not USCF), and Karpov and Korchnoi were in the very
low 2700s.  Fischer had retired with a never-surpassed rating of 2830 or
so, and there was a perennial beer argument about whether it was conceivable
that in his reclusive state he could have stayed current with the whole
world's rising tide of theory, AND added enough private innovations to stun
the world upon his hypothetical return.  The flavor of this romantic
speculation is exactly captured in the movie "Searching for Bobby Fischer".
Of course he _did_ finally come back, not as insane as feared nor as
brilliant as hoped, but still clearly among the half-dozen best players in
the world after a 20-year vacation...a feat that I can only compare to Mark
Spitz's Olympic comeback try (when he actually broke one of his 1972 world
records, without making the team).

Two more amusing asides.  (1) There have been only three American world
chess champions in history, and every one of them has retired undefeated
by refusing to defend the title.  Paul Morphy, Bobby Fischer, Deep Blue.
(2) Like chess, Go has a finely tuned rating system (and a much more useful
one than chess, since players of wildly different abilities can play
meaningfully with enough stones of handicap).  Anyway, during the
three-century isolation of Edo Japan, the island of Okinawa was open to
Portuguese and Dutch traders on a limited basis, and was therefore
considered treif by the rest of Nippon.  Came Commodore Perry in the 19c,
and for the first time since the Ashikaga period the Go masters of Okinawa
got to return to the home islands to play.  And...they discovered that
there had been ratings drift, such that an Okinawan player claiming to be
9-dan strength was really only about 7.5-dan in Tokyo.  Unfortunately, the
Nipponese lacked the conceptual framework to attribute this to a flaw in
the rating system, as opposed to a flaw in the players.  In a period of a
couple of years, about a hundred of Okinawa's top Go players committed
seppuku.  If you think that game ratings are by their nature completely
subjective, the shades of those unfortunate players would be the ones to
take it up with.

``Gary Kasparov, you've just beaten the +---------------------------------+
world's most powerful chess computer -- | Joshua W. Burton  (847)677-3902 |
what are you going to do next?'' ``I'm  |      jwb@post.harvard.edu       |
going to http://www.disneyworld.com!''  +---------------------------------+

© 2000 Peter Langston