Monthly FIDE rating lists

[Christopher Kreuzer]

(Thanks to John McKenna for pointing this out)

http://www.chessbase.com/newsdetail.asp?newsid=8290

"The big news is that FIDE has new ratings available for Rapid and Blitz; that from July 2012 FIDE will be publishing its rating lists on monthly basis; and that the rating floor for players to be included in all lists will be lowered to 1000."

Will that have any impact on British chess?

[Roger de Coverly]

We had an earlier discussion which touched on the frequency of ratings.

http://www.ecforum.org.uk/viewtopic.php?f=4&t=3135

More frequent rating lists can mean that it takes more games to correct an out of date rating. But is there much practical difference between monthly and two-monthly? Only for players with the time available to play in multiple consecutive events.

The rating floor will make it easier for players to get on the International list. 1000 is about the standard of the experienced but not very good club player. A player more than capable of taking a piece when offered, but less able to actually win with that advantage, particularly if short of time.

[Stewart Reuben]

that from July 2012 FIDE will be publishing its rating lists on monthly basis; and that the rating floor for players to be included in all lists will be lowered to 1000."

I greatly believe more frequent lists leads to more chess activity. Many countries including Russia, India, China do not have a rating system and rely on FIDE. The above was my proposal in 1999. It has taken a long time to get there. We were worried about corrupting the list. Of course you won't see the effect until 1 August 2012.
The USCF updates as the results come in. That would probably be awkward for title norms.

[Roger de Coverly]

We were worried about corrupting the list.

The methods used by the International system are (edit) almost (/edit) unchanged from the days when it recorded only the results achieved by elite players in near-elite competition with the only real difference being the frequency of publication.

Where you can make a comparison with a domestic system, as in England, I'm not sure these methods have worked especially well when extended down the rating scale. As a higher priority than predicting results, the rating system should be able rank players in the correct order. Below 2000 and particularly where Juniors or other new players are concerned, I don't think it's doing that well.

There's a problem with updating the base rating immediately on completion of a tournament, which is that the tournament immediately following doesn't know what rating to use in the lag while the rating reports are collated, calculated and published. A well defined cutoff is a practical necessity. The concept of "expected change" works reasonably well.

(edit) Changes to the 320 point rule have been made

It used to be applied to the tournament as a whole, making it possible to lose points just by playing. The calculations were then applied game by game thereby always giving the highly rated player a reward for winning. 320 points was then increased to 400 points. I would think these changes had little aggregate effect, even if for individuals it enabled them to reach title thresholds. (/edit)

[E Michael White]

Unless changed the K-factors will be too low. Active junior players will be underated compared to their compatriots who are first rated later. Selection policies for juniors are wise to place more importance on TPRs which are independent of a players current rating. However juniors who play primarily against adults will fare better than those who play other juniors. Stewart Reuben is mistaken to believe that this will encourage more juniors to play. Perhaps this is the thinking behind a separate junior rating system.

[Stewart Reuben]

FIDE are considering running a modified Glicko System alongside the current FIDE system with a view to changing over in due course if it proves superior.
We acknowledge that the Sticko (modified Glicko) predicts results 'better'. That was not what interested us. It seems to be substantially superior at handling new players and ones who don't play much.

I find it hard to believe that people don't realise an 8 year old will not be enthused if he is told there is a grading system and he will get such a grade in about 6 months (hitherto 1 year).

[Roger de Coverly]

and ones who don't play much.

FIDE should be very careful about equating "not played in an Internationally rated event" with "not played at all". Experience with the Australian Glicko system suggests that players who are active outside of Australia can return and have the rating effect of their first result on return greatly magnified.

[Stewart Reuben]

'FIDE should be very careful about equating "not played in an Internationally rated event" with "not played at all". Experience with the Australian Glicko system suggests that players who are active outside of Australia can return and have the rating effect of their first result on return greatly magnified.'

That is perfectly true. My own rating in the US showed such a tendency after a 26 year absence from their system. We were shown numbers for Matthew Sadler which were absurd.
However, the advantage of Glicko is that the reliability of the player's rating partially determines the effect his rating has on the opponents. That reliability is measured by the amount of activity within the system. It is a separate number from the effect o n his own rating (as per k factor).
FIDE would need a system where the player's rating, if based on few games, would not determine qualification for an event such as th World Cup.
The great diasdvantage of the Glicko system is that it is difficult to understand and impossible to work out rating changes without a computer. The Australians don't publish the Reliability quotient, thus making it totally impossible for amateurs to work out new ratings.

[Joey Stewart]

I dont like the idea of this floor of 1000. That could mean we will have absurdly low rated playing floating about, nicking hundreds of points from us hard working adults when they finally figure out how to play well.

It would be nice if there was some sort of rake back system where points were retroactively given back to 'consistent' players (I.E. those who have maintained a relatively steady grade for a long period of time) when a fast improving low rated player finally hits his peak - perhaps once somebodies rating improves by 100 points or more and then they reach a plateau whereby they stick around plus or minus 30 points of a certain rating then their past results are recalculated for any opponents they had back on their way up so that they can get a few back too.


Is that a workable sort of idea? I am sure it would encourage more established players to keep going, instead of what tends to happen where they play a mere handful of games for the rest of their chess career for fear of running into under rated opponents who will nick their points.

[Paul McKeown]

I tend to think that this fear of juniors is something that belongs to the bottom half of the grading list, adults below 130 or so, terrified of kids, basically because the law of averages say that piles of kids are going to go past them. By the time you get anywhere near 200 ECF, the idea that vast swathes of junior players are underrated and will beat gazillions of grading points out of you, somehow doesn't seem real. There will always be a few highly talented young players, but most juniors play in rather predictably one dimensional ways. The few highly talented teenagers that you might lose to are exceptional and make no difference to your grade. As for FIDE ratings, below 2300 or so they are pretty meaningless, so no one should worry about them, except FIDE who need to sort out the ridiculous k-factors that keep almost everyone static, excepting for those that play many scores of FIDE rated games per year.

[Roger de Coverly]

'
That is perfectly true. My own rating in the US showed such a tendency after a 26 year absence from their system. We were shown numbers for Matthew Sadler which were absurd.

One idea that's straightforward is that after a period of x years, an inactive rating expires. So returning players are just treated as new players. A modified version would just put the K factor up. That would mean their opponents in their come-back event would be playing unrated opposition unless you are prepared to tolerate asymmetric rating.

[Stewart Reuben]

Roger>One idea that's straightforward is that after a period of x years, an inactive rating expires. So returning players are just treated as new players. A modified version would just put the K factor up. That would mean their opponents in their come-back event would be playing unrated opposition unless you are prepared to tolerate asymmetric rating.<

That surprises me. You treasurer historical information. Why not treasure historical ratings. It is how the information is used that is important. The ECF system of discarding data afer a few years is hugely flawed.
Glicko tries to solve this. I don't know whether it is successful. But the basic premise is that, if your rating is highly stable it has a substantial effect on your opponent's rating when you play him. If your rating is highly unstable, then it only has a small effect on your opponent's rating when you play him.

Joey >I dont like the idea of this floor of 1000.<
But you are living in a country where there is an established internal rating system. Many countries, including Russia, don't have that and FIDE is international.

I have no idea why Paul thinks FIDE Ratings under 2300 are meaningless. Does he also think ECF grades under 210 are meaningless?

[Paul McKeown]

Does he also think ECF grades under 210 are meaningless?

No. The ECF system has been stable and has worked down to grades of -100 or thereabouts for decades. FIDE ratings have been tinkered with for about twenty years, dropping the rating floor every few years, so that when one player would have been assigned an initial rating of 2205, that player some years later would have received an initial rating (with a less outstanding achievement) of 2005. Similarly with 2005 and 1805, 1805 and 1605. Now the process continues all the way down to 1000.

Consequences:

a) some players are dramatically overrated compared to their peers, because they won the jackpot a few years before their peers
b) some improving players (typically but not necessarily juniors) are given (correctly at the time) low ratings, but because of FIDE's moronic k-factors they never (or extremely slowly) progress later towards much higher ratings more representative of their later greatly improved standards
c) FIDE ratings below 2300 tend to be based on a small sample of player's activity and are therefore inherently unreliable.

For these reasons I tend to think that FIDE ratings below 2300 are rather a lottery.

[Roger de Coverly]

c) FIDE ratings below 2300 tend to be based on a small sample of player's activity and are therefore inherently unreliable.

Whilst this is a statement of fact for many British players, it doesn't apply in countries where there is little chess played outside of time limits suitable for FIDE rating. So the ratings for French and German players would reflect all their results.

At present, FIDE will now rate games of nominally up to two hours duration and those of four hours or more, but not the durations in between. There's a provision about being allowed shorter time limits for lower rated players, but events taking advantage of this are few and far between.