Converting nonsense ECF ratings to Elo...

[John Upham]

Following the release of the nonsense ECF ratings for juniors, will 4NCL be forced to use the normal conversion from "new money ECF ratings" to Elo for the purposes of deciding board orders?

In one of our local leagues (which comprises of rating limited competitions) some juniors will not be able to play any games since they are their playing strength is very much less than their new rating and the club they play for will not select them on merit : has the world gone finally mad?

[Roger de Coverly]

Following the release of the nonsense ECF ratings for juniors, will 4NCL be forced to use the normal conversion from "new money ECF ratings" to Elo for the purposes of deciding board orders?

One might suppose that they would use the 8*ECF + 650 conversion. International ratings take precedence over domestic ones, so there will only be board order problems if you have a squad with both unrated and rated junior players. The unrated players will be rated above their strength and the rated players probably lagging.

In one of our local leagues (which comprises of rating limited competitions) some juniors will not be able to play any games since they are their playing strength is very much less than their new rating and the club they play for will not select them on merit : has the world gone finally mad?

The indications are that the end August grading list won't just be the usually tidy up of missing results and duplicate players, but they will run the random number generator (aka junior and new player estimation) to generate yet another list. So who knows what will come out the other side?

Border is not a league I have any involvement in, but you might consider reversing the junior enhancement for the purpose of eligibility. Even more radically, you could consider calculating unofficial "local" grades using last season's results and using these for eligibility. Many, many years ago the then grader used to do this for the Berks & Bucks Congress in years when the grading list hadn't appeared by middle/end August.

[Matthew Turner]

I believe one of the Scottish officials ran a calculation using the old 'new grades' and found that 10*ECF + 190 was actually a pretty accurate conversion to ELO (certainly more accurate than previous conversions). Before you say anything, I do appreciate this would make Callum Kilpatrick 2530, so it's not perfect!

[Roger de Coverly]

I believe one of the Scottish officials ran a calculation using the old 'new grades' and found that 10*ECF + 190 was actually a pretty accurate conversion to ELO

I read that as well but they were talking about conversion to their domestic Elo system. Players based in England sometimes retain their place in the domestic Scottish list by submitting their English results. You also get Scottish players in events like Blackpool, the British and Hastings of course.

As more "club" players get international ratings, it will become clearer whether the international system is inflated relative to domestic Elo systems. As a "for example" I noted that the player who was briefly the UK's highest rated eight year old, now has a Scottish domestic Elo of around 1600, an International Elo of around 1800 and a "new ECF" of 150.

Given that the ECF's "new grades" system will be unable to tell them Junior grades until 14 months after the event in some cases and they believe the "junior improvement" factors are far to0 high for many players, it also appeared they'd just decided to stop rating games against English players unless the players had international ratings. Now that's a boost of confidence in the "new grades" system.

The original thread is at http://scotchess.s4.bizhat.com/scotches ... c1199.html.

Heather Lang's digest is worth reading.

[Matt Harrison]

I've being doing a brief comparison between the ECF junior ratings and FIDE.

50 of the top 100 juniors have a FIDE rating
In addition 13 juniors have a FIDE rating, but are just below the top 100 ECF junior list

So we have 63 players with ratings in both systems to compare:

Average ECF: 180
Average FIDE: 1943

Using 8xECF+650:

Average estimated FIDE is 2089
Players whose FIDE is overestimated: 60
Players whose FIDE is underestimated: 3

Average overestimate is 146, but remember that for younger juniors 80 points of this is accounted for by the junior bonus of 10 points (5 points for 16/17 year olds - 34 in this sample).

So 8xECF+600 would give an average error of 96, 61 points of which is junior bonus.

Average estimated FIDE is 2039
Players whose FIDE is overestimated: 58
Players whose FIDE is underestimated: 5

Removing junior bonus from ECF, then using 8xECF+600:

Average estimated FIDE is 1981
Players whose FIDE is overestimated: 43
Players whose FIDE is underestimated: 19

The underestimated players tend to be those who have been less active recently. Therefore their ECF grade is a bit lower than the more active ones.

As a sidenote, the expected gain from the 1/7/09 list to the 1/9/09 list is about 5 points. With 6 lists a year, this would be a 30 point improvement, compared to the 61 points assumed by ECF junior bonus.

[David Shepherd]

Matt, I have some reservations about how accurate the fide ratings are for the English juniors, as I suspect that they play in far less fide rated tournaments compared to some of the European juniors and therefore their rating may not change quite as quickly as they should - despite the fact that the ratings are published more often.

[Jonathan Rogers]

I agree with David. I wouldn't much want to play any of the Barbican 4NCL juniors at their current ratings.

[Matt Harrison]

I agree, the FIDE ratings for many of the players are likely to underestimate their strength considerably. If this is the case, then the ECF grades don't look too bad. There will inevitably be a few outliers (such as Callum Kilpatrick's 234 grade - how much of that was due to his tremendous performance for Dulwich in the London League), but otherwise, I don't think the new grades look too out of kilter at 8xECF+600. Not sure about +650 though.

[Sean Hewitt]

I've examined theg 1364 adult players who have a currect ECF grade and also a FIDE rating. The formula ECF * 8 + 650 = FIDE seems to be a reasonable approximation.

It does slightly underestimate players graded below 185 and overestimate those graded above that but it's probably as good as any other conversion we've had.

Don't forget, no conversion will ever be perfect - it's only an estimate.

[Ola Winfridsson]

It does slightly underestimate players graded below 185 and overestimate those graded above that but it's probably as good as any other conversion we've had.

Don't forget, no conversion will ever be perfect - it's only an estimate.

Exactly, and there will always be cases of players who produce the performance of their lives (or rather the opposite) for their first Elo rating, and then spend a long time acting Father Christmas.