CFC rating system broken?

[Kerry Liles]

Re: CFC rating system broken?

On this topic, I did notice something (probably a typo) in the CFC Handbook.

Regarding the estimated ratings of unrated players used to calculate the performance of unrated players against other unrated players, I think the formula for this estimated pre-rating of unrateds should be AR-200+400(W-L)/N, not AR-200+400(W-L) as stated in the handbook.

AR in this formula is a sort of modified average rating of the section. Then I assume the -200+400(W-L) is supposed to give the unrated player an estimated rating somewhere between 200 above and 600 below this modified average, based on performance - which it does with the /N at the end. Otherwise, the estimate drops 400 pts for each loss, which could rapidly generate a negative estimated rating!

Here's the text:

717. Order of Calculation. When an event is rated, performance ratings under equation 714a are obtained, first for unrated players, then for provisionally rated players. For each rated player is calculated: Ro-400(W-L)/N. The mean of these is AR. When unrated plays unrated in this first pass, the opponent’s rating is estimated as AR-200+400(W-L). After this, the previously unrated players are rerated on the basis of the ratings of their opponents just calculated. Finally, the players having established ratings are rated, first those with ratings below 800. A game with a previously unrated or previously provisionally rated player enters the rating of a player with an established rating on the basis of the opponent’s post-event (new) rating. In the case of a player having an established rating below 800, the following is used to calculate the rating of opponents having established ratings 800 and above: the higher of the player’s pre- and post-event ratings.

Thanks for pointing this out...
I will verify this with someone who likely knows and look into sorting it out.

[Fred McKim]

Re: CFC rating system broken?

On this topic, I did notice something (probably a typo) in the CFC Handbook.

Regarding the estimated ratings of unrated players used to calculate the performance of unrated players against other unrated players, I think the formula for this estimated pre-rating of unrateds should be AR-200+400(W-L)/N, not AR-200+400(W-L) as stated in the handbook.

AR in this formula is a sort of modified average rating of the section. Then I assume the -200+400(W-L) is supposed to give the unrated player an estimated rating somewhere between 200 above and 600 below this modified average, based on performance - which it does with the /N at the end. Otherwise, the estimate drops 400 pts for each loss, which could rapidly generate a negative estimated rating!

Here's the text:

717. Order of Calculation. When an event is rated, performance ratings under equation 714a are obtained, first for unrated players, then for provisionally rated players. For each rated player is calculated: Ro-400(W-L)/N. The mean of these is AR. When unrated plays unrated in this first pass, the opponent’s rating is estimated as AR-200+400(W-L). After this, the previously unrated players are rerated on the basis of the ratings of their opponents just calculated. Finally, the players having established ratings are rated, first those with ratings below 800. A game with a previously unrated or previously provisionally rated player enters the rating of a player with an established rating on the basis of the opponent’s post-event (new) rating. In the case of a player having an established rating below 800, the following is used to calculate the rating of opponents having established ratings 800 and above: the higher of the player’s pre- and post-event ratings.

Marcus - A+

Yes there is a typo as you have described. You are probably the only person who has figured out what this section even means. Jonathan Berry and I designed it to handle unrateds vs unrateds, in the first pass, years and years ago. As the program is currently written, if there are more than 50% unrateds, there is a different method for calculating AR.

[Gary Ruben]

Re: CFC rating system broken?

Hi Gary

Back in 1985 I had a 2300 range CFC rating and a high 2200 range FIDE rating. Not sure if this is what you have in mind for a *high* CFC rating. :)

I think in 1985 the CFC membership was probably higher and the club and open tournaments had a larger player pool. At least in Ontario and probably other provinces. The correspondence club had around 700 members and the CFC surely must have been triple or quadruple that.

I don't know how many FIDE games you had played or even if your FIDE rating was fixed. My ICCF rating is based on probably at least 150 games, mostly played in invitational and national team events so would be a more accurate indication of how I stack up against the worldwide pool of players than would be my CCCA rating which has now been inactive for several years. It's not that I don't want to play CCCA but they use me for international and I don't want too heavy a game load.

[Egidijus Zeromskis]

Re: CFC rating system broken?

3) The only thing preventing there being more input data is that, understandably, people feel they don't want to pay to have the same games rated twice. They can't rate games with only FIDE because the CFC won't let them!

I think, the Canadian tournament can be FIDE rated and no CFC rating fees are charged if all players are CFC members without (with some special exceptions for Quebec tournaments).
Do you have untold stories?

[Tom O'Donnell]

Re: CFC rating system broken?

I think, the Canadian tournament can be FIDE rated and no CFC rating fees are charged if all players are CFC members without (with some special exceptions for Quebec tournaments).
Do you have untold stories?

Sorry, I don't think I understand what you are writing. I tried running a tournament rated FIDE without also being rated CFC and this was not allowed. I had to pay the CFC rating fee for the event (and of course the FIDE fee for the event, too). It happened in early Sept. 2009. The CFC is the gateway to FIDE and their toll is three bucks a head. ;-)

I would love to have a Governor propose a motion whereby for some processing fee tournaments could be submitted by the CFC to FIDE but not rated by the CFC. Why pay twice?

[Fred McKim]

Re: CFC rating system broken?

Sorry, I don't think I understand what you are writing. I tried running a tournament rated FIDE without also being rated CFC and this was not allowed. I had to pay the CFC rating fee for the event (and of course the FIDE fee for the event, too). It happened in early Sept. 2009. The CFC is the gateway to FIDE and their toll is three bucks a head. ;-)

I would love to have a Governor propose a motion whereby for some processing fee tournaments could be submitted by the CFC to FIDE but not rated by the CFC. Why pay twice?

Tom is correct. A Canadian tournament will only be submitted to FIDE if the CFC rating fees are also submitted.

Tom now that you've aired the issue, maybe you will find a willing Governor to take up the cause.

[Christopher Mallon]

Re: CFC rating system broken?

Why pay twice?

To support the CFC of course.

Financially as well as in other ways, such as maintaining the activity in the rating pool, etc.

[Jonathan Berry]

Re: CFC rating system broken?

I'd like to thank Kevin for being the person to pick up on this anomaly in the rating system.

Naaah. With all credit to KS (after all, Charles and I
are the co-founders of the Kevin Spraggett Fan Club "Resign
Moron"), Kevin wasn't the first. There must have been others,
but see this from 2008:

http://www.chesstalk.info/forum/showthread.php?t=348

and especially this:

http://www.chesstalk.info/forum/showthread.php?t=183

It's "Case 2".

I guess the subsequent silence was because it hadn't happened yet.

The virtual case 1 relates to the recent addition of points to the rating of one of our
young champions, partly because the bonus point system failed him.
Two years before it happened. Scary? No.

Back in 2011 the shocking thing is that KS didn't at once point out that
his buddy buddy PS was the (co-) architect of the bonus point debacle.

[Tom O'Donnell]

Re: CFC rating system broken?

To support the CFC of course.

Financially as well as in other ways, such as maintaining the activity in the rating pool, etc.

I have no problem with supporting the work Gerry Litchfield does. He is extremely competent and efficient, and that's why I have no problem paying some sort of processing fee. But to support the CFC rating system? Why?

As for maintaining activity in the rating pool, my bet is for most players with both FIDE and CFC ratings that they would prefer to maintain activity in the FIDE pool, which is the world-wide rating standard. So why doesn't the CFC insist that all events in Canada be rated by FIDE to maintain the activity in the world-wide rating pool? ;-)

[Fred McKim]

Re: CFC rating system broken?

Naaah. With all credit to KS (after all, Charles and I
are the co-founders of the Kevin Spraggett Fan Club "Resign
Moron"), Kevin wasn't the first. There must have been others,
but see this from 2008:

http://www.chesstalk.info/forum/showthread.php?t=348

and especially this:

http://www.chesstalk.info/forum/showthread.php?t=183

It's "Case 2".

I guess the subsequent silence was because it hadn't happened yet.

The virtual case 1 relates to the recent addition of points to the rating of one of our
young champions, partly because the bonus point system failed him.
Two years before it happened. Scary? No.

Back in 2011 the shocking thing is that KS didn't at once point out that
his buddy buddy PS was the (co-) architect of the bonus point debacle.

Jonathan: You were the first, unfortunately nobody listened to you back then and it took the practical case of it actually being used to bring the issue to light.

[Marcus Wilker]

Re: CFC rating system broken?

Yes there is a typo as you have described. You are probably the only person who has figured out what this section even means. Jonathan Berry and I designed it to handle unrateds vs unrateds, in the first pass, years and years ago. As the program is currently written, if there are more than 50% unrateds, there is a different method for calculating AR.

Just out of interest, what is the method when more than 50% are unrated? I was trying to predict first ratings for a small children's tournament where only two had previous ratings, and the published ratings were much higher than my estimates.

[Fred McKim]

Re: CFC rating system broken?

Just out of interest, what is the method when more than 50% are unrated? I was trying to predict first ratings for a small children's tournament where only two had previous ratings, and the published ratings were much higher than my estimates.

In the case of more than 50% unrated, the rating program asks for a manual estimate of the average rating of the unrateds, to be used for the first pass. Obviously something realistic has to be used.

I had to get one of my tournaments rerated, as I felt the estimate supplied was 200 or so points too high.

Now that you know this you can supply the probably supply the estimate in the future.

[Kerry Liles]

Re: CFC rating system broken?

In the case of more than 50% unrated, the rating program asks for a manual estimate of the average rating of the unrateds, to be used for the first pass. Obviously something realistic has to be used.

I had to get one of my tournaments rerated, as I felt the estimate supplied was 200 or so points too high.

Now that you know this you can supply the probably supply the estimate in the future.

Since this fact seems largely unknown (until now), who normally supplies the estimate? I would have to presume it is the person who is entering the rating information (Gerry?).

How in the world can he (or anyone who doesn't even know the players) estimate anything?

[Fred McKim]

Re: CFC rating system broken?

Since this fact seems largely unknown (until now), who normally supplies the estimate? I would have to presume it is the person who is entering the rating information (Gerry?).

How in the world can he (or anyone who doesn't even know the players) estimate anything?

Yes, Kerry. I had no idea this provision was in the algorithm, until I wondered why my tournament in the fall didn't turn out anywhere close to the ratings I thought it should have. We adjusted the estimate and I was satisfied.

It should be put in the rules (handbook and website).

I think the estimate should have to be agreed upon between Gerry and the organizer. He can try different values until the results look "right". For example you would edxpect that any rated players would come out with approximately the same average rating they started with.

[Alan Baljeu]

Re: CFC rating system broken?

The problem with the Elo system is too often I hear people say they don't want to play because they might lose points. That's terrible when you're trying to promote chess!

Instead of a zero-sum rating system, we should instead have a system that rewards participation. Win games, get points. Place in tournaments, get points. Have a player of the year race, and prizes at the end. Big tournaments offer more points. Collect stats on player performance. Keep lifetime points and points per event.

That would incentivize participation instead of Elo which incenses only rising newcomers.