6 monthly grades

[Sean Hewitt]

In which case, you can tweak the k-factor accordingly. Arguably k is too high if your rating can change by 300 points in, say, a 5-round weekender. If you had it such that:
A grade: k = 15
B grade: k = 20
C grade: k = 25
D grade: k = 30
E grade: k = 35
Otherwise: k = 40
Then you might negate the problem. I don't know if k = 40 is too high, but if so, find out the maximum practical k, and allocate a k to each class such that they're roughly equally distributed.

This approach is broadly correct, subject the determining the correct values of k. I assume that juniors would have a higher k factor.

[Roger de Coverly]

This approach is broadly correct, subject the determining the correct values of k. I assume that juniors would have a higher k factor.

In a national rating system, you have to be very cautious about assuming that just because you haven't seen any games by a player in your particular subset of the world's rating pool, that this means they have been inactive. As far as I'm aware, Glicko makes the assumption that absence = inactivity.

It's as if the NCCU had their own rating system and if players ventured to Blackpool every so often, they were given a massive K for their Blackpool results. This might be appropriate if they'd hardly ever played in the North, but if at some time in the past they had established a reliable Northern rating, it would be totally silly.

[Sean Hewitt]

This approach is broadly correct, subject the determining the correct values of k. I assume that juniors would have a higher k factor.

In a national rating system, you have to be very cautious about assuming that just because you haven't seen any games by a player in your particular subset of the world's rating pool, that this means they have been inactive. As far as I'm aware, Glicko makes the assumption that absence = inactivity.

It's as if the NCCU had their own rating system and if players ventured to Blackpool every so often, they were given a massive K for their Blackpool results. This might be appropriate if they'd hardly ever played in the North, but if at some time in the past they had established a reliable Northern rating, it would be totally silly.

I agree entirely. Hence the need to determine the correct values of k for each group, and of course determing the appropriate groups!

[Richard Bates]

By varying K-factor with activity aren't you effectively trying to implement a rating system whereby an individual's rating aligns much closer with performance over the rating period, with far less weight given to prior recorded strength? Which is basically what the ECF system does, in a much less complicated way!

[Rob Thompson]

By varying K-factor with activity aren't you effectively trying to implement a rating system whereby an individual's rating aligns much closer with performance over the rating period, with far less weight given to prior recorded strength? Which is basically what the ECF system does, in a much less complicated way!

What actually is the weighting system used by the ECF?

[Roger de Coverly]

What actually is the weighting system used by the ECF?

There's no single answer and the explanation will depend on where you start.

In the simplest form where everyone plays exactly 30 games over 3 seasons, each game contributes 1/30th of the total. Where it's less than 30 it will be 1/n where n is the number of games. This cuts off at 9 for publication. Where more games than 30 are played, if they are in a single season, only that season counts and the weighting is again 1/n. If you have to go back into previous seasons to count up to 30, games are discarded on a first in, first out basis. Thus the oldest are ignored in preference to the more recent. The weighting will be "about 30", because games are discarded on an event by event basis. If the date of the game is known, as opposed to the date of the event, this will be used. So if you play two games in a day, either both will count or neither.

The practical effect is that the method can "forget" earlier performances quite readily for people who play a lot, but remembers them for players who don't play much. After 36 months, it has no memory whatsoever.

So if you had a 175 player who was absent for three years, on their comeback they would be ungraded for rating purposes.
If they made a 150 performance in their come-back 5 round tournament, that would be potentially be a 150 grade but wouldn't be published. To get their 175 back, they would have to play at 200 in their next tournament. Elo systems would work the same way provided both tournaments were in the same rating period. Without knowing the detail, I'd suspect that Glicko systems would give more weight to the first event.

[Rob Thompson]

So if i have understood correctly there is no mechanism for weighting games played in the most recent season more (as they are more relevant) than games played in previous seasons, if a player is fairly inactive?

[Sebastian Stone]

What effect would this have on relatively inactive players (those who play less than 30 games a year)?

I have played 12 rated games (3 of those against unrated players) since September and in my club only 4 out of 23 of our players managed to play above 30 games last season (I was not one of them).

[Roger de Coverly]

So if i have understood correctly there is no mechanism for weighting games played in the most recent season more (as they are more relevant) than games played in previous seasons, if a player is fairly inactive?

Well there used to be. For example if you always played 20 games a season, your grade would have been two-thirds of this season's performance + one-third of last season's.

The way they now do it, is that they take this year's 20 games and add back the last 10 from the previous season. That might work for improving (or declining) players but to my mind just introduces a biased weighting towards end-season events.

I have played 12 rated games (3 of those against unrated players) since September and in my club only 4 out of 23 of our players managed to play above 30 games last season (I was not one of them)

If you didn't play any more this season , the grading system will go back in time for a maximum of two seasons to attempt to find another 18 games. The more games you play this season, the fewer will be retrieved from previous years. This assumes you are an adult. Junior players are treated as ungraded every year.

[Sebastian Stone]

How will this effect season long tournaments? Would the club be expected to send in their games halfway through a competition or will they send them at the end of the season as per usual making the 6 month period irrelevant?

And how would this effect people who play both in season long tournaments and congresses?

[Roger de Coverly]

How will this effect season long tournaments? Would the club be expected to send in their games halfway through a competition or will they send them at the end of the season as per usual making the 6 month period irrelevant?

This applies to leagues as well of course. If you want grades to change more frequently and in particular establish grades for new players, then the results have to be processed more frequently which in turn implies they will have to be reported more frequently. It ought not to be difficult provided you capture the results somehow, somewhere on a computer and have a painless way of exporting the captured data to the grading system. In practice I expect there will be competitions, particularly the minority which still have adjudications and adjournments which will report annually.

And how would this effect people who play both in season long tournaments and congresses?

If you had leagues or club competitions which only reported annually, you would have a six-monthly grade which only reflected your congress results. Internationally rated players suffer this already with the 4NCL. For example Hastings Masters results will be in the 1st March list, but 4NCL results, even those from 2010 won't be included until July 2011.

The rules as to whether an annually reporting league should re-base the calculations for later games to the six-monthly list have yet to be established.

[Richard Bates]

If you had leagues or club competitions which only reported annually, you would have a six-monthly grade which only reflected your congress results. Internationally rated players suffer this already with the 4NCL. For example Hastings Masters results will be in the 1st March list, but 4NCL results, even those from 2010 won't be included until July 2011.

The rules as to whether an annually reporting league should re-base the calculations for later games to the six-monthly list have yet to be established.

Apologies coz this is old ground, but memory's not good. What IS the reason for 4NCL reporting so late? Purely administrative?

[Roger de Coverly]

What IS the reason for 4NCL reporting so late? Purely administrative?

Norms may be an issue, but I think it's part-ratings. You need 9 games for a first international rating, but it has to be at least 3 per event. That's something difficult to do if you rate each weekend separately. The 4NCL is promoted as a means to get an international rating, so it's important for many players in the lower divisions.

I saw a proposal somewhere, probably from Stewart Reuben, that the minimum number should be 1.

The French league (and presumably others) rate each separate weekend. The French have nearly all their players FIDE rated already.

[Richard Bates]

What IS the reason for 4NCL reporting so late? Purely administrative?

Norms may be an issue, but I think it's part-ratings. You need 9 games for a first international rating, but it has to be at least 3 per event. That's something difficult to do if you rate each weekend separately. The 4NCL is promoted as a means to get an international rating, so it's important for many players in the lower divisions.

I saw a proposal somewhere, probably from Stewart Reuben, that the minimum number should be 1.

The French league (and presumably others) rate each separate weekend. The French have nearly all their players FIDE rated already.

Presumably though, games are rated as as date of game ie. games rated in October are rated on a different basis from games in February. Since I'm sure norms definitely aren't an issue (4ncl always used to be rated in stages) then if the splitting of ratings allows events still to be considered as 'whole' then FIDE regulations must be very deficient if they do not allow part-ratings to be treated likewise. It certainly seems bizarre that the regulations should actively discourage games from being rated in a timely manner, especially as the late rating of games arguably fundamentally undermines the basis for the whole system (I think i am possibly right that one English GM many years ago managed to 'secure' their title by 'finding' an error in an old rating calculation that, when added at the later date, gave them the five pts they needed to reach 2500 - when in reality the 5pts could well have been filtered out for a reasonably active player pretty rapidly).

[IM Jack Rudd]

No, games are rated as at date of event-start-date. This is normal for an event that straddles rating periods (eg Hastings); it's just that the 4NCL straddles rather a lot of them.