FIDE Ratings

[David Pardoe]

CJ rules ok...hopefully well supported by the ECF/BCF team. It might be helpful to keep a good arms length from the web side. This can be a very demanding media, and maybe not a bad idea to keep a moderately lowish profile, least you get bombarded with all sorts of demands, which can be bamboozling. But maybe we need the ECF office to take a role in monitoring that and responding to any comments as appropriate. Maybe also better not to go for `instant response` approach. Lets have some considered reflection, where we take time to reach the correct decisions. Less knee jerk politics.

Re the Olympics....not sure that running this in parallel with the World Chess Olympiads would be good. It might get eclipsed by the main Olympic coverage.
Birmingham...fine, I suppose. Holding everything in London is probably not good. Lets share things around.

Also, whats the big deal about FIDE rating. Im not convinced we need to chase that `prize`. So long as a reasonable conversion formula exists with the ECF ratings, that seems fine.

[Alex Holowczak]

Also, whats the big deal about FIDE rating. Im not convinced we need to chase that `prize`. So long as a reasonable conversion formula exists with the ECF ratings, that seems fine.

I agree, but the fact remains that you're not going to attract the best players in the UK to a tournament that isn't FIDE rated. Look at the success of Sean Hewitt's events as a perfect example. The Congresses in the West Midlands are perfectly good for attracting local players (and often the odd GM or two), but there's nothing to attract strong national players to the events for them to play against. Indeed, one Birmingham League club recently took a hoard of players up to Sean's Warrington tournament. This suggests there's enough demand for something similar in Birmingham.

[Alan Walton]

I think that one a the benefits around FIDE rating is the way it is calculated, also having a standardised world system does help for international tournaments, home or abroad

The ECF grading system is basically an average calculation and is very rigid (especially as it is only an annual figure)

Whereas the FIDE system is more dynamic by calculating a predicted score for each game, and being produced bi-monthly is a true reflection of current performance

Having a "reasonable conversion" wouldn't help if the number you are converting is mathematically suspect

I think the best way to go is for the ECF grading system to mirror the same calculations and distribution of the FIDE system, but this would be quite expensive I beleive

[Ian Thompson]

Whereas the FIDE system is more dynamic by calculating a predicted score for each game

Doesn't the ECF system also do this? 10 points difference in grade means the expected scores are 0.6 and 0.4 for the two players.

being produced bi-monthly is a true reflection of current performance

Except that it doesn't work well for annual leagues where all the games are submitted for rating at the end. In that respect its not much different from ECF grades. My next ECF grade will be based on games I played between September 2009 and May 2010 (as I didn't play any graded games last summer). My next FIDE rating will be based on games I played between October 2009 and May 2010. Not much difference there. Most of my opponents in FIDE rated games will have benefited from using my October 2009 FIDE rating, which was higher than my rating when most of the games were played, and I will have been similarly disadvantaged. That can't be a good thing if you want the best possible estimate of performance (for me, this season, I think its made my July 2010 rating 7 points lower than it would have been had ratings current at the time the game was played been used).

[Alan Walton]

Ian,

The basic way of looking at it is this

if you are graded 200 ECF (converts to 2250 FIDE) and play 30 games against 175 ECF (2125 FIDE) and get 50% your new ECF grade would be 175, though if you did the same performance in FIDE your expected score would be approx 20 out of 30, thus losing 75 FIDE Points giving a new rating of 2175 which converts to 185 ECF.

With the FIDE system if you play those 30 games spread over the 6 rating periods then you rating loss will not be as severe

[Roger de Coverly]

With the FIDE system if you play those 30 games spread over the 6 rating periods then you rating loss will not be as severe

It's equally true that a player who has improved from 175 to 200 will get the full benefit after 30 games over one season in the ECF system but rather longer under an Elo style system.

It's a problem with all rating systems though - what is the trade-off between reflecting changes in performance against the long term stability of the rating?

[Ian Thompson]

if you are graded 200 ECF (converts to 2250 FIDE) and play 30 games against 175 ECF (2125 FIDE) and get 50% your new ECF grade would be 175, though if you did the same performance in FIDE your expected score would be approx 20 out of 30, thus losing 75 FIDE Points giving a new rating of 2175 which converts to 185 ECF.

With the FIDE system if you play those 30 games spread over the 6 rating periods then you rating loss will not be as severe

A rating system estimates playing strength. You've given an example where two different system start off with the same estimate and then diverge. The fundamental question, which you haven't addressed, is which of the two new estimates is the better one, and why?

[Alan Walton]

The main problem around any system/calculation is the frequency of updates

The more frequent the updates the more accurate the rating will be, hence using the FIDE model would be currently better

It would be a personal choice which calculation you use but I think the FIDE model is better, mainly due to the mathematical calculation being more sophisticated (people more mathematically astute than me have done these studies before)

Another question with the current ECF system, is what is the average number of games a player is graded on each year, as everybody is aware, using an averaging calculation with a low volume of data will give inconsistent results. Therefore unless you having an A or B grade, you start experiencing larger swings in peoples grade

[Roger de Coverly]

The more frequent the updates the more accurate the rating will be, hence using the FIDE model would be currently better

That doesn't necessarily follow. For example if a player is a bit variable in their results, they might gain 30-50 Elo points one tournament and give them back the next. So if both tournaments were in the same rating period, their rating (as estimate of strength) is unchanged. If you have a rating list between the tournaments, then they are 50 points adrift potentially.

is what is the average number of games a player is graded on each year, as everybody is aware, using an averaging calculation with a low volume of data will give inconsistent results

C,D and E people are averaged over up to 3 years, so an established club player with 10 games a season has a grade based on 30 games. When such a player faces someone who plays about 30 games a season, they both gain or lose the same number of grading points.

That said, the new ECF policy for junior players of ignoring previous seasons' grades is potentially unstable and might make for some strange outcomes next July.

[David Pardoe]

I`ve always been slightly sceptical about gradings, and what they really represent. The ECF system of averaging over a season has fair merit I think, as a guide.
What I`d like to see is grading ranges, ie to consider a player as being grade `x` + or - `y`.
ie, a player graded say 130 may well win against players graded 20 or 30 points higher. By the same token, that player might also lose to other players graded 20 or 30 points lower.
I think this variability tends to deminish as you rise up the gradings, so that when you go above 175 - 185 the tendancy might be for results to go more frequently to `form`.
I have an issue with the concept of `form`, particularly below the 180ish level. I suspect a great deal of variability applies to player performance, particularly in league chess, where factors like `work pressures`, domestic issues, etc can effect player performance. ...as can the standard of opposition. ie, a player on board 2 for a team, playing opponents graded say 175 on average, might struggle to match the performance of another player, playing board 5 against average opposition graded say 135.
Its very hard to draw comparatives about playing strength from these situations.
For that kind of reason, I favour the averaging effect of rating over a season.
Maybe players should be regarded as `within bands`. Maybe we should have letters for 10 point bands ranging from `A` for 240 - 250, down through `B`, `C`, `D`, etc...to players graded 50 & below...?

[Sean Hewitt]

I think there are two amin reasons that players prefer FIDE ratings. Neither is to do with the maths.

1 - Inidividual rating of games. With FIDE you know the impact the result is going to have on your rating ie win and you gain 10 rating points, lose and you lose 5. With ECF you don't know this until the end of the season and the number of games you have played is known. If you play alot of games, the result of one game can be almost meaningless.
2 - Dynamic ratings. The rating change referred to in one is very evident and your rating will change within a 2 month maximum period. With ECF, results take between 2 and 14 months to filter to your grade depending on when the game is played.

I believe that in particular (2) encourages players to play more often. I know that Stewart Reuben shares this view. The ECF could help itself with more regular lists (something which Chris Majer supported).

[David Pardoe]

Ernie..maybe people are more interested in rating games than rating the President?

Anyway, a point was raised about the recent Warrington Congress & FIDE ratings.
Firstly, yes, it was quite a successful event, and great to see a strong and well supported Open section.
Some good tactics/management/promotion in attracting a good field, not least, the appearance of players like Keith Arkell.
Slightly disappointing that more Manchester players didnt enter. Apparently they prefer cheaper entry fees (maybe cheaper venues?). Also, I believe some were put off by venue and travel issues. ie, not easy to get to unless you have a car?
Not sure about the relative merits of this, but top marks to Sean & the organisers for bringing a quality event to the north.

[Ian Kingston]

Ernie..maybe people are more interested in rating games than rating the President?

Maybe so - but they'll have a better chance of finding the discussion if it's placed in an appropriately named thread.

[Carl Hibbard]

Maybe so - but they'll have a better chance of finding the discussion if it's placed in an appropriately named thread.

A bit rushed but sort of done - please point out anything I missed

[Sean Hewitt]

Slightly disappointing that more Manchester players didnt enter. Apparently they prefer cheaper entry fees (maybe cheaper venues?).

Interesting. What cost do such players want the entry fee to be? Would these same players be happy to see a reduced prize fund as a result of reduced entry fees? Presumably comfortable warm surroundings, playing one game per table, and having fully stocked bar within staggering distance of the playing hall have no value to these players?

Also, I believe some were put off by venue and travel issues. ie, not easy to get to unless you have a car?

The venue was less than a mile from the local railway station (a 20 mins walk according to Google). It was also right next to a motorway junction and provided free parking. I thought this was great for everyone, especially those travelling by train from far afield (as many players did).

How much closer does the playing venue need to be to the train station? In the waiting room perhaps?