Difference between FIDE and CFC Ratings

[Fred McKim]

There some discussion about this on a thread earlier, where I tried to suggest that the difference was much less than 200 points as had been suggested by someone.

A good starting point is the analysis on the CFC site.
http://www.chess.ca/CFCvsFIDE.shtml

I had done an analysis back in Nov of the top 20 Canadian based players, having played in a Canadian event in the past 3 years.

Sambuev, Bator Montreal, QC 2725 2497 228
Kovalyov, Anton Verdun, QC 2638 2623 15
Bluvshtein, Mark Toronto, ON 2634 2587 47
Gerzhoy, Leonid Toronto, ON 2629 2497 132
Charbonneau, Pascal Outremont, QC 2585 2511 74
Samsonkin, Artiom Toronto, ON 2576 2402 174
Noritsyn, Nikolay Richmond Hill, ON 2556 2408 148
Hansen, Eric Calgary, AB 2537 2412 125
Porper, Edward Edmonton, AB 2517 2437 80
Roussel-Roozmon, Thomas Montreal, QC 2500 2500 0
Hébert, Jean Longueuil, QC 2486 2414 72
Cheng, Bindi Burnaby, BC 2485 2376 109
Krnan, Tomas Burlington, ON 2470 2390 80
Tayar, Jonathan Toronto, ON 2460 2356 104
Cummings, David Toronto, ON 2456 2369 87
Hartman, Brian Caledonia, ON 2453 2392 61
Yoos, John C. Jack Vancouver, BC 2452 2382 70
O'Donnell, Tom Ottawa, ON 2451 2358 93
Panjwani, Raja Kitchener, ON 2450 2416 34
Piasetski, Leon Vancouver, BC 2416 2301 115

92.4

I performed the identical analysis for the top 20 US players at this level and found the US ratings inflated by 70.8

I'll try to give this a six month update, shortly, unless someone beats me to it.

[Roger Patterson]

Re: Difference between FIDE and CFC Ratings

It might be more meaningful if you restricted the list to those who are currently active in both systems.

[Fred McKim]

Re: Difference between FIDE and CFC Ratings

It might be more meaningful if you restricted the list to those who are currently active in both systems.

Are you talking about the CFC list or my list? I think all the players on my list are active in both systems.

The numbers Gerry automatically (I assume, anyway) churns out would be difficult to define activity, as we don't carry a FIDE activity indicator in the CFC data.

[Roger Patterson]

Re: Difference between FIDE and CFC Ratings

Are you talking about the CFC list or my list? I think all the players on my list are active in both systems.

The numbers Gerry automatically (I assume, anyway) churns out would be difficult to define activity, as we don't carry a FIDE activity indicator in the CFC data.

Well then, how about a regression of the form

FIDE = cfc + a + b*n

Where n is number of games played in cfc system in the last year or two, and a,b are the regression constants and test for statistical significance.

[Fred McKim]

Re: Difference between FIDE and CFC Ratings

Well then, how about a regression of the form

FIDE = cfc + a + b*n

Where n is number of games played in cfc system in the last year or two, and a,b are the regression constants and test for statistical significance.

I'll try to do some more work on this over time. My main purpose was to illustrate that the CFC system was not 200 points out of line with FIDE. Obviously, there is a mean difference and a standard deviation from it like any set of data.

[Felix Dumont]

Re : Difference between FIDE and CFC Ratings

Sambuev, Bator Montreal, QC 2725 2497 228
Bluvshtein, Mark Toronto, ON 2634 2587 47
Gerzhoy, Leonid Toronto, ON 2629 2497 132
Samsonkin, Artiom Toronto, ON 2576 2402 174
Noritsyn, Nikolay Richmond Hill, ON 2556 2408 148
Hansen, Eric Calgary, AB 2537 2412 125
Porper, Edward Edmonton, AB 2517 2437 80
Cheng, Bindi Burnaby, BC 2485 2376 109
Krnan, Tomas Burlington, ON 2470 2390 80
Tayar, Jonathan Toronto, ON 2460 2356 104
Cummings, David Toronto, ON 2456 2369 87
Hartman, Brian Caledonia, ON 2453 2392 61
Yoos, John C. Jack Vancouver, BC 2452 2382 70
O'Donnell, Tom Ottawa, ON 2451 2358 93
Panjwani, Raja Kitchener, ON 2450 2416 34
Piasetski, Leon Vancouver, BC 2416 2301 115

105.4

And what happens when we take off the players from Quebec who don't really play in CFC tournaments? Now the average difference is of 105,4! And some of the players above aren't that active... Basically the current difference between the CFC and the FIDE rating of a very active CFC player should be of at least 125-150. That's way to much...

[John Coleman]

Re: Difference between FIDE and CFC Ratings

Not quite the same subject, would it be possible for tdlist.zip to show the date of the most recent activity of the player? For regular ratings; I don't pay any attention to active ratings.

[Fred McKim]

Re: Difference between FIDE and CFC Ratings

Not quite the same subject, would it be possible for tdlist.zip to show the date of the most recent activity of the player? For regular ratings; I don't pay any attention to active ratings.

I doubt that Gerry can change the format of that list, as he'd have to look at tournament data.

However the new web top player search function will allow you to select a list of players using last activity date as a filter.

[Egidijus Zeromskis]

Re: Difference between FIDE and CFC Ratings

FIDE = cfc + a + b*n

Where n is number of games played in cfc system in the last year or two

Counted games (n) starting 2010

correlation (diff vs n)	-0.792477233				
intercept (a)	-43.5991989				
linest	(b) -1.315385475	 -43.5991989			
average	-92.4		
Name      CFC  FIDE      n     diff

Samb.	2725	2497	107	-228	http://www.chess.ca/memberinfoSQL.asp?CFCN=146462
Kova.	2638	2623	9	-15	http://www.chess.ca/memberinfoSQL.asp?CFCN=142449
Bluv.	2634	2587	0	-47	http://www.chess.ca/memberinfoSQL.asp?CFCN=125709
Gerz.	2629	2497	44	-132	http://www.chess.ca/memberinfoSQL.asp?CFCN=142916
Char.	2585	2511	15	-74	http://www.chess.ca/memberinfoSQL.asp?CFCN=108065
Sams.	2576	2402	86	-174	http://www.chess.ca/memberinfoSQL.asp?CFCN=146305
Nori.	2556	2408	99	-148	http://www.chess.ca/memberinfoSQL.asp?CFCN=132534
Hans.	2537	2412	69	-125	http://www.chess.ca/memberinfoSQL.asp?CFCN=132475
Porp.	2517	2437	44	-80	http://www.chess.ca/memberinfoSQL.asp?CFCN=136955
Rous.	2500	2500	9	0	http://www.chess.ca/memberinfoSQL.asp?CFCN=122701
Hébe.	2486	2414	9	-72	http://www.chess.ca/memberinfoSQL.asp?CFCN=101280
Chen.	2485	2376	45	-109	http://www.chess.ca/memberinfoSQL.asp?CFCN=141968
Krna.	2470	2390	16	-80	http://www.chess.ca/memberinfoSQL.asp?CFCN=132215
Taya.	2460	2356	0	-104	http://www.chess.ca/memberinfoSQL.asp?CFCN=133658
Cumm.	2456	2369	51	-87	http://www.chess.ca/memberinfoSQL.asp?CFCN=123161
Hart.	2453	2392	36	-61	http://www.chess.ca/memberinfoSQL.asp?CFCN=102700
Yoos.	2452	2382	40	-70	http://www.chess.ca/memberinfoSQL.asp?CFCN=107994
O'Do.	2451	2358	36	-93	http://www.chess.ca/memberinfoSQL.asp?CFCN=106245
Panj.	2450	2416	18	-34	http://www.chess.ca/memberinfoSQL.asp?CFCN=121202
Pias.	2416	2301	9	-115	http://www.chess.ca/memberinfoSQL.asp?CFCN=101297

What does that mean? :?

[Gary Ruben]

Re: Difference between FIDE and CFC Ratings

What does that mean? :?

Hmmmmmm...... Probably either the file or system itself is corrupted. (corrupted = containing errors).

Probably the way to attempt to standardize the file is to go back to before the bonuses and rate the events without bonuses. Then see if the results are closer to the FIDE ratings.

[Fred McKim]

Re: Difference between FIDE and CFC Ratings

Counted games (n) starting 2010

correlation (diff vs n)	-0.792477233				
intercept (a)	-43.5991989				
linest	(b) -1.315385475	 -43.5991989			
average	-92.4		
Name      CFC  FIDE      n     diff

Samb.	2725	2497	107	-228	http://www.chess.ca/memberinfoSQL.asp?CFCN=146462
Kova.	2638	2623	9	-15	http://www.chess.ca/memberinfoSQL.asp?CFCN=142449
Bluv.	2634	2587	0	-47	http://www.chess.ca/memberinfoSQL.asp?CFCN=125709
Gerz.	2629	2497	44	-132	http://www.chess.ca/memberinfoSQL.asp?CFCN=142916
Char.	2585	2511	15	-74	http://www.chess.ca/memberinfoSQL.asp?CFCN=108065
Sams.	2576	2402	86	-174	http://www.chess.ca/memberinfoSQL.asp?CFCN=146305
Nori.	2556	2408	99	-148	http://www.chess.ca/memberinfoSQL.asp?CFCN=132534
Hans.	2537	2412	69	-125	http://www.chess.ca/memberinfoSQL.asp?CFCN=132475
Porp.	2517	2437	44	-80	http://www.chess.ca/memberinfoSQL.asp?CFCN=136955
Rous.	2500	2500	9	0	http://www.chess.ca/memberinfoSQL.asp?CFCN=122701
Hébe.	2486	2414	9	-72	http://www.chess.ca/memberinfoSQL.asp?CFCN=101280
Chen.	2485	2376	45	-109	http://www.chess.ca/memberinfoSQL.asp?CFCN=141968
Krna.	2470	2390	16	-80	http://www.chess.ca/memberinfoSQL.asp?CFCN=132215
Taya.	2460	2356	0	-104	http://www.chess.ca/memberinfoSQL.asp?CFCN=133658
Cumm.	2456	2369	51	-87	http://www.chess.ca/memberinfoSQL.asp?CFCN=123161
Hart.	2453	2392	36	-61	http://www.chess.ca/memberinfoSQL.asp?CFCN=102700
Yoos.	2452	2382	40	-70	http://www.chess.ca/memberinfoSQL.asp?CFCN=107994
O'Do.	2451	2358	36	-93	http://www.chess.ca/memberinfoSQL.asp?CFCN=106245
Panj.	2450	2416	18	-34	http://www.chess.ca/memberinfoSQL.asp?CFCN=121202
Pias.	2416	2301	9	-115	http://www.chess.ca/memberinfoSQL.asp?CFCN=101297

What does that mean? :?

Very good work !!

Here's what it means after a little algebraic rearranging:

FIDE = CFC - 43.6 - 1.3*G

In layman's terms it means that someone who didn't play any CFC games should have a FIDE rating 43.6 points lower. This discrepancy increases by 1.3 for every CFC game that is played.

Now the accuracy of this measurement is measured by the correlation coefficient which is -0.7. The closer this number is to 1.0 or -1.0 measures how significant the three sets of data interrelate. 0 would indicate totally random data with no correlation. I'm just an amateur statistician, so others can expalin this better.

The results would seem to indicate that as you play more CFC games your FIDE rating will lag further behind. We have to note that this set of data, is only those players who were in the top 20 CFC players who had played in the last three years.

A goal of the rating system would be to reduce both 43.6 and 1.3, but the 1.3 is even more distressing as it shows with some certainty that there is ongoing inflation at the top level.

I did some analysis a few months ago to try and determine what the adjusted ratings might be of the players who had received significant performance bonuses (if you don't know what this is see Spraggett Blogspot (at your own risk)). I'll look into that.

I'm also looking at what's happened in the 6 months since I originally did this. We no longer have participation points so this may also make a difference.

[Egidijus Zeromskis]

Re: Difference between FIDE and CFC Ratings

Hmmmmmm...... Probably either the file or system itself is corrupted. (corrupted = containing errors).

Probably the way to attempt to standardize the file is to go back to before the bonuses and rate the events without bonuses. Then see if the results are closer to the FIDE ratings.

Probably the system is corrupted. I suspect there should be no correlation between games (n) and a difference (diff), or maybe there should be a different correlation when n>>(a lot of games), then diff >>0.

There is a picture of a data set.

[Egidijus Zeromskis]

Re: Difference between FIDE and CFC Ratings

disclaimer: as calculations of the games were done quickly, some numbers might be wrong (lets say by -5-10), as I did not looked what kind of the games were Regular or Active.

[Fred McKim]

Re: Difference between FIDE and CFC Ratings

disclaimer: as calculations of the games were done quickly, some numbers might be wrong (lets say by -5-10), as I did not looked what kind of the games were Regular or Active.

As there are not a lot of Active tournaments at this level, I doubt that would make a big difference to these figures.

So my plan is:

1) try to factor out performance bonuses and see what we get.

2) bring the study 6 months ahead to just before the long weekend (up to and including the Can Closed kind of thing - which data I can get from FIDE site) or wait until the June FIDE ratings, depending on how much time I have to work on this.

[Roger Patterson]

Re: Difference between FIDE and CFC Ratings

Very good work !!

Here's what it means after a little algebraic rearranging:

FIDE = CFC - 43.6 - 1.3*G

In layman's terms it means that someone who didn't play any CFC games should have a FIDE rating 43.6 points lower. This discrepancy increases by 1.3 for every CFC game that is played.

Now the accuracy of this measurement is measured by the correlation coefficient which is -0.7. The closer this number is to 1.0 or -1.0 measures how significant the three sets of data interrelate. 0 would indicate totally random data with no correlation. I'm just an amateur statistician, so others can expalin this better.

The results would seem to indicate that as you play more CFC games your FIDE rating will lag further behind. We have to note that this set of data, is only those players who were in the top 20 CFC players who had played in the last three years.

A goal of the rating system would be to reduce both 43.6 and 1.3, but the 1.3 is even more distressing as it shows with some certainty that there is ongoing inflation at the top level.

I did some analysis a few months ago to try and determine what the adjusted ratings might be of the players who had received significant performance bonuses (if you don't know what this is see Spraggett Blogspot (at your own risk)). I'll look into that.

I'm also looking at what's happened in the 6 months since I originally did this. We no longer have participation points so this may also make a difference.

and I would add that both of those coefficients are statistically significant. By itself, the intercept of 43 points or so is not such a big problem. The 1.3 points per game is a huge problem.

An interested reader who does not post sent me the following info broken down by region:

http://victoriachessclub.pbworks.com...9180/dummypage

which suggests that Ontario in particular has a differential for it's top players of close to 170 points.