Author: Rolf Tueschen
Date: 03:58:02 02/20/03
Go up one level in this thread
On February 19, 2003 at 20:29:53, Robert Hyatt wrote: >On February 19, 2003 at 18:53:47, Sune Fischer wrote: > >>On February 19, 2003 at 18:25:53, Robert Hyatt wrote: >> >>>>>they may improve _faster_ but the overall chess population won't improve. It >>>>>_can't_. >>>> >>>>Why not? >>> >>>Because if _everybody_ improves, ratings can _not_ change, since ratings are >>>based >>>solely on the probability of X beating Y. :) >> >>So what you are saying is, that although the strength on average has gone up, >>the Elo rating shouldn't, by construction? > >correct, unless the ones at the top get better, while the ones at the bottom >stay the same, which doesn't seem reasonable. > >This is very much like an electrical voltage. Which is defined as the >difference in potential between two opposite poles. For example, with pure >ground and +12 volts, the potential is +12 volts. But if the ground is >"floating" and the potential between the floating ground and the actual ground >is +2, then the potential between the floating ground and +12 volts is only >+10. But if you need ten volts to run a motor, it doesn't matter if ground=0 >and +=10, or if ground=10 and +=20, it is the _difference_ that does the work. > >And the two voltages can float all over the place just so their difference is >10, and nothing changes and nobody notices unless they try to compare the >absolute voltage and say "hey, this year you are running at 36 volts while >last year you were running at 32." Even though this year ground=26 and last >year ground=22... > >That's why the comparison simply won't ever work. > > > > >> >>I have to say I'm not sure about that, in practise you have a flux in and out of >>players, it is never a constant pool so you can't expect to conserve average. > >I agree. Which was my point. The "pool changes" so the average rating will >also change, and it tends to move up because of various "tweaks" federations >apply to the ratings from time to time. But the "average" is meaningless, >as in my voltage sample. The difference in ratings (or voltages) is what >makes things happen. Now, Bob, I would really prefer that you could abstain from such true statements because what will you do if there are no differences in top players? Of course, I know, then we enter the parapsychological sphere or should I say paramathematical? Or homeopathical? If you have nothing (differences) and then try to filter something out of it then you get the SSDF ranking list. And this list is older than the history of human chess. Because look, you have new progs on a 6-month base, so that makes 60 generations over the last 30 years. While human chess only has a history of say 10 generations! So SSDF, also on the base of their incredibly high games collection of 60 000 [!!!] games, 1000 games per generation, is able to get results of invisible differences. Note that in human chess we only have more than a couple of million games per generation! So by definition the differences are larger and more inaxact. Rolf Tueschen > > > >> >>>I don't follow. Players don't start at 1000. They might be _seeded_ as if they >>>are 1000 >>>in a tournament, but their initial rating is the rating of their first opponent, >>>based on the >>>result. And this continues for the first 20 or 24 games... >> >>Not where I come from, here we get seeded, and if there is no prior knowledge of >>your strength, then you get 1000. At least that was the way it was when I >>started. > > >That is different from setting your initial rating to 1000. At least FIDE >and USCF don't do that. For your first 20 or 24 games (I think it is 20 now, >but when I played actively it was 24 with USCF) your rating is a "TPR" type >rating, then it is "frozen" after the provisional period passes and the normal >formula is applied from that point forward. > > > >> >>>>inflation comes through the bonus points scored by those moving >>>>up the ranks and the fact that there are no good ways to balance things, no >>>>absoluteness in the scale so it tends to drift (up _or_ down). >>> >>>That was my point... Except it _never_ drifts down unless there is an overall >>>downward >>>manual correction. The current systems only drift up. >> >>That can't be so, the opposite must be the case and the idea is very simple. >>Players enter as weak and low rated, and exit (when they die?) as higher rated, >>well presumably very few end up worse than were they started. >> >>So the pool loses rating and the average goes down. >>You need to pump points into the system to keep it going at the same level. > >You miss the fact that _lower_ rated players also die... > >And there are more of those to boot. :) > > > >> >>> But if everyone would >>>get off the >>>"2850 means XXX is the best player ever" and only use it to say that "XXX is 50 >>>points >>>better than the next best player" things would be a lot saner, because the >>>former is wrong, >>>while the latter is perfectly accurate. >> >>We can't compare ratings over time, but since all the GMs acknowledge Kasparov >>to be the greates ever, it is only natural if he also has the highest rating >>ever. I don't see any real hard evidence the scale is drifting unwarrented, then >>again there is no reason why it should remain the same, but whether it has gone >>up or down I can't say. I believe average strength has increased a little bit, >>so even if the scale has "drifted" a bit upwards I wouldn't call it inflation >>but rather a natural adjustment. :) >> >>-S.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.