Ratings (was Re: Hastings)

Andy Walker <anw@xxxxxxxxx> wrote:
David Richerby <davidr@xxxxxxxxxxxxxxxxxxxxxx> wrote:
... the increase in standards, as measured by ratings and by
numbers of titled players...

Except that neither ratings nor the number of titled players lets
you measure an increase in standards. Ratings are only valid for
comparison within the pool of players in which they're measured

Yes; nevertheless, you hear players saying "I played well last
season, and my rating has gone up, but the way I'm playing now,
it'll go back down again", and such like.

Well, of course they do. If playing well means winning and playing
badly means losing, that's exactly what will happen to their rating.
What they can't do is say `Ten years ago, I was rated X; now I am
rated Y>X; therefore I am a stronger player.'

A rating system is really not much use unless it has a certain
stability. Not just between the players in one club or one league
over one season, but across a country or the world, and over at
least several years, preferably several decades.

Not really. Equivalence across the world is desirable -- we want a
French 2000FIDE player to be expected to score 50% against an
Australian 2000FIDE player because that's what the rating is supposed
to mean -- but there's no need for stability over time. Yes, it would
be nice to be able to say that Fischer peaked at 2785 and Karpov at
2780 so Fischer was slightly the better player but the rating system
has never promised to be able to do that.

The rating system would in no way be compromised if, for example, FIDE
added 1000 points to everybody's rating on January 1st each year[1].
As long a the system drifts in a consistent way across the pool of
players, it's doing what it was designed to do. Sure, it would be
nice if the rate of drift was zero but that's by no means essential.

[1] Well, they should also change the rating requirements for titles
and so on but they're extraneous to the rating system.

Otherwise, rating-limited tournaments, grading prizes and the like
are a nonsense [even more so than logically they are in the first
place],

The rating bands of the tournaments can always move with the ratings.
Maybe that doesn't happen so much with Elo ratings, where multiples of
100 are such tempting boundaries. However, I'm pretty sure I've
played in under-100, U110, U120, U125 and U130 sections of English
tournaments and, while some of those were the bottom section and some
of them the second-from-bottom, it does show that the boundaries are
flexible and can react to any drift in the ratings. If, for example,
a tournament doesn't have many participants in the U100 section, they
can have the bottom group be U110 the next year.

and when a player joins your club after moving from far away or
after not playing for a few years, you cannot get any idea from his
distant or ancient rating what his standard might be. [Admittedly,
in such a case, there is the method we always *used* to use in
pre-grading days, of actually giving the chap a game, and observing
how well he plays.]

Somebody who only has an old rating probably hasn't played much chess
in the time since he was last rated. Therefore, any rating-based
clues to his strength are always going to be highly approximate
because he's out of practice. Even if ratings were constant over
time, knowing that somebody was rated X five or ten years ago doesn't
tell you much about how strong he is today compared to players
currently rated X. Of course, it could be that he was playing under a
different rating organization so he might not be out of practice.

That is actually what we had, across the world and the decades.
There are players in my team who stayed at, say, 150 [BCF/ECF, 1800
FIDE], give or take a few points, for 30+ years, until they decayed
away; and for decades, 2600 FIDE was the sign of a "super GM", one
of the *handful* of players likely to get into the Candidates'
Tournament and perhaps challenge for the world title.

FIDE has only been issuing ratings since 1970. People have been
muttering darkly about rating inflation for at least ten years so `for
decades' can't mean much more than `for at least 25 years'. Was
everything really so rosy in the period 1970-1995? It's a simple
observation that there are many, many more FIDE-rated players now than
there were twenty or thirty years ago. Even under the assumption that
the statistical parameters of the distribution have remained the same,
you would expect the number of players above any given rating point to
have increased and you would expect the highest-rated player of the
current population to have a higher rating than the highest-rated
player of thirty years ago.

Meanwhile, FIDE ratings have [it seems to me] inflated badly, so
that even 2700 is no longer the mark of a top-10 player, and there
are well over a hundred players over 2600.

In 1960, the men's world 100m record was 10.0s. Nowadays, it's not
uncommon for several athletes in any given race to run faster than
that. Time has inflated badly so that even 9.9s is no longer the
mark of a top-ten sprinter.

and the increase in the number of titled players is surely to a
large extent because more people are trying to achieve titles.

Not sure I believe this. It's much easier to get lots of titled
players if the absolute standard needed to become a GM is declining
....

Yes and it's much easier to get lots of titled players if titles are
available for $4.99 from Walmart. The knowledge that X implies Y and
that Y is true does not allow you to conclude that X is true.

If ratings are inflating then, yes, it's easier to become a GM because
it's easier to get a 2500 rating and it's easier to achieve a 2600
performance rating for a norm. But an increase in the number of GMs
doesn't tell you that ratings are inflating.

It is certain that those GMs rated 2600+ at Hastings are strong
players; it's much less clear that they are the peers of the
half-dozen or so strongest players of 1950, or 1936, or 1895. Nor
that they outclass the Birds and Albins to the extent that they
would expect to win *every* game.

I don't know enough about the strength of people like Bird and Albin
to comment on that. It doesn't seem outrageous to me to claim that
somebody like Bird or Albin would score at least a few draws in a
tournament against against modern 2600 players


Dave.

--
David Richerby Solar-Powered Accelerated Projector
www.chiark.greenend.org.uk/~davidr/ (TM): it's like a 16mm film projector
but it's twice as fast and it doesn't
work in the dark!
.

Relevant Pages

  • Re: Finals time limits (again)
    ... Rating points and prizes are exactly ... the same for places 5-2 in the tournament final. ... players 2-5 lose nothing at all by ...
    (rec.games.trading-cards.jyhad)
  • McMakko (was: Ranking system in European countries)
    ... rating system. ... Everybody enters the tournament with their current EGF rating ... players are paired against players with similar ...
    (rec.games.go)
  • Re: New Rating System Proposal
    ... >>consistently compete against 7 other players should deserve a rating ... which awards any form of ranking to people who don't play. ... Games have winners and losers. ...
    (rec.games.trading-cards.jyhad)
  • Re: V:EKN ranking issues
    ... The Rating System page (http://www.white-wolf.com/vtes/index.php? ... my players show up at the website right after their first tournament ... appearance is ranked. ...
    (rec.games.trading-cards.jyhad)
  • Re: New Rating System Proposal
    ... one reason why it's irrelevant to talk about what players in the boondocks ... could be doing if they wanted rating points badly enough. ... It may be reasonable to suspect that a player who doesn't regularly play ... >> what one person has to go through to make tournaments happen for themselves ...
    (rec.games.trading-cards.jyhad)