CFC ratings database: some analysis for ratings & activity

[Roger Patterson]

Re: CFC ratings database: some analysis for ratings & activity

The database has now been uploaded with historical CFC information back to 1996.

[Roger Patterson]

Activity back to 1997

A chart for CFC member activity (total, rated<1200, and rated >1200) going back to 1997 is shown below. Graphs for individual provinces can be found at http://victoriachess.com/cfc/distribution/provinces.JPG (stats are those calculated by activity in the province rather than by the province of the players - as a result, the numbers are kind of bumpy when a Canadian Open rolls into town....)

[Roger Patterson]

Re: average rating back to 1997

average rating of the top 100 back to 1997. It seems clear that the last few years have induce inflation. Maybe there was some deflation that needed to be corrected but the correction was over the top.

[Duncan Smith]

Re: average rating back to 1997

average rating of the top 100 back to 1997. It seems clear that the last few years have induce inflation. Maybe there was some deflation that needed to be corrected but the correction was over the top.

We attended events locally for most of the decade and contrary to some recent posts from Steve Karpik and others the local trend in players is pretty close to the charts. Early in the decade, tournaments here were getting somewhat more popular and some of the higher end veteran players came out of partial retirement and played more as well. I would say during that rating deflation period the level of play was actually quite high as well due to an influx of strong juniors ( more so then normal - CMA and imports ) and those veteran players. The rating deflation should not be understated it was real. I could illustrate but I don't think that's necessary.

Around 2006 ( not sure exact timing ) there was a huge drop in local interest which clubs like SCC are now improving on that lower base. I am also hearing about drops in attendance in CMA and school events, but that is second hand information I haven't researched it. There are some strong new entrants but not as many of those last group of stronger veteran players are playing regularily anymore. Things are recovering I'm not sure how much.

Its pretty obvious overall that CFC ratings have overcorrected. The top end players have skyrocketing ratings and even class players I used to be in tournaments with have done a kind of V shaped recovery of their rating ( eg Bob Armstrong ). What should the true ratings be who really knows. Doing anything about it is problematic, because you have players from one era or another and that divergence would only work itself out over many tournaments ( will many of the veterans even come out and play ? ).

[Ed Seedhouse]

Re: average rating back to 1997

average rating of the top 100 back to 1997. It seems clear that the last few years have induce inflation. Maybe there was some deflation that needed to be corrected but the correction was over the top.

Sorry, but as far as I can see your graph shows nothing of the sort. The average rating of the top 100 today remains below the peak figure in the late 1990's.

This will depend on where you think the correct line "should" be, but you have provided no rationalization as to where that would be. Nor does the graph seem to me to show any long term trend. What does a regression line show and what's the r^2?

I just don't think that data supports a claim of "overcorrection" on it's own.

[Egidijus Zeromskis]

Re: CFC ratings database: some analysis for ratings &amp; activity

historical CFC information back to 1996.

It is interesting to see players' actions, but maybe could generate a list of people who made that possible - most active TDs over the period and provinces? :)

[Roger Patterson]

Games per year per player by rating

A pretty interesting and somewhat surprising graph.

a) the correlation between strength and games/year is pretty strong...

b) how many times have you heard or been part of a conversation about 'how can we get the stronger players to play?' That conversation is backwards - it's the weaker players you need to work on.

I haven't posted graphs for the provinces - you can get the data at: http://victoriachess.com/cfc/activit...e=distribution if you are interested.

[Vlad Drkulec]

Re: average rating back to 1997

average rating of the top 100 back to 1997. It seems clear that the last few years have induce inflation. Maybe there was some deflation that needed to be corrected but the correction was over the top.

How are you determining the top 100? Is it the top 100 active players? It appears so because using the current top 100 list that can be downloaded from the CFC including inactive players yields an average of 2400 and not 2350 or so shown on your graph.

Ten or eleven points of the inflation can be attributed to the arrival of Sambuev, Samsonkin and Kovalyov on Canadian shores. No real inflation there, the talent pool just got better and three 2280ish players were knocked out of the top 100 and replaced by three players with ratings that were collectively 1100 or 1200 points higher.

Perhaps another ten points can be attributed to random returns of former top ranked players. I can take credit for one point of the increase around 2008 since I helped talk Ray Stone into coming out of retirement for a team tournament which made him an active player for that year anyway knocking a high 2200s player off the list for a high 2300s player.

[Roger Patterson]

Re: average rating back to 1997

top active players. If you want to see who is on the list on any given date, go to http://localhost/cfc/ranking.php?yea...1&province=CAN

The CFC list includes people who haven't played for 30 years and excludes currently active players who's membership is not current. Not very useful for looking at changes in the rating system [nor is it particularly satisfying as a top player list]

The general problem of noise in the measurement [people coming and going, getting better/worse, etc.] makes determining inflation a hard problem. You need as much data and any many different measurement techniques as possible. Although I note that for the players you mention, their arrival does not coincide with the recent jump in the level.

[Roger Patterson]

Re: CFC ratings database: some analysis for ratings &amp; activity

It is interesting to see players' actions, but maybe could generate a list of people who made that possible - most active TDs over the period and provinces? :)

just for you :-). One of our most active TDs is Anon.

http://victoriachess.com/cfc/tournam...N&datatype=tds

This link gives the most active TDs for the entire rating period (copied some of the top ones below). You can also get:

1) number of tournaments each year (Canada and by province)
2) number of tournament directors each year (Canada and by province)
3) activity of tournament directors each year (Canada and by province)
4) listing of tournaments each year (Canada and by province) with name, TD, Winners's rating, average rating, #players.

Some problems with the lists:
- some missing TD names in the data
- Multiple sections are usually entered as multiple tournaments (but not always)
- Tough to separate out the junior only events from other events.
- Sigh - the tournament organizer is no longer recorded so we forget about them....

Most active TDs 1995-2010
Province ID TD # Rated events
CAN 108387 Pelts, Roman 2249
CAN 102198 Rutherford, John 646
CAN 0 Anon 341
CAN 100317 Dutton, Mark S. 321
CAN 106974 Lamb, Bryan 309
CAN 110253 Putic, Predrag 279
CAN 108715 Wright, Stephen 227
CAN 130181 Nadeau, Denis 220
CAN 107847 Gibbons, Lorne 203
CAN 106295 Keshet, Joshua 197
CAN 103802 Tsui, Alick 188
CAN 109502 Hughey, Micah 187
CAN 127489 McDonald, Patrick 170
CAN 111830 Bond, Hal 162
CAN 128173 Dobrich, Vlad 150
CAN 100145 McKim, Fred 142
CAN 127778 Anon 140
CAN 102718 Duff, Ken 139
CAN 134840 Barron, Michael 136
CAN 109477 Demmery, Steve 128
CAN 100205 Stringer, Lynn 118
CAN 128388 Nadeau, Ellen 115
CAN 106340 Demian, Valer Eugen 107
CAN 101613 Burgess, Doug 103
CAN 120416 Bethell, Tim 103
CAN 111363 Daswani, Benedict 99
CAN 108098 Field, Christopher 94
CAN 107184 Holmes, Michael J 85
CAN 100314 Craver, Lyle 76
CAN 105964 Hiuser, Robert 72
CAN 103841 Thurairasah, Shivaharan 71
CAN 109393 Luiting, Larry 70
CAN 106424 Deline, Toni 69
CAN 133671 Peng, David (Yu) 65
CAN 145502 Anon 63
CAN 110803 Den Otter, Albert 61
CAN 101588 Pradzinski, Tim 60
CAN 106687 Schulz, Waldemar 59
CAN 106052 Quiring, John 58
CAN 108844 Winmill, Ronald 57
CAN 101495 Watson, Walter 56
CAN 102882 Brown, John R. 56
CAN 100355 Litchfield, Gerry 55
CAN 109003 Legacy, Leo 55
CAN 103904 Maund, Chris 54
CAN 106856 Palsson, Halldor Peter 54
CAN 100238 Barnes, Mark 53
CAN 106868 Thompson, Ed G. 53
CAN 121367 Johnson, Ghislaine 53
CAN 103074 Poitras, Luc 51
CAN 111760 Davis, Neil 51
CAN 135013 Cosenza, Aaron 51
CAN 104137 Ficzere, Tony 50

[Tony Ficzere]

Re: CFC ratings database: some analysis for ratings &amp; activity

Interesting stuff, but not sure how accurate it all is. When I go to your link and check out myself as a TD for example, it comes up with 12 events. Yet, when I check the CFC website, I count 16 events in 2009 alone. Hmm.

[Roger Patterson]

Re: CFC ratings database: some analysis for ratings &amp; activity

my site gives 17 for you in 2009. Not sure where you are finding 12 events listed.

[edit: the 12 is the number of events you've directed in 2010 up to September? Lifetime total is 50.]

[Roger Patterson]

Membership retention

http://victoriachess.com/cfc/retention2.php

Jonathan Berry is fond of telling us that back in the day, something like 30% of all memberships lapsed within one year, not to be renewed. The available database does not record historical membership data, only current membership but we can look at some related measures: Probability of activity by year if: a) active in a given year (plotted below for the given year as 2009.07 to 2010.07 and for 2001.1-2002.1) b) if a member as of the date of the database (2010.9).

Somewhat surprisingly, something like 35% of current members are not active in the current year. In another context, there was some indication that life members are by and large not active so that could account for part of that 35% (15% or so - i.e. 20% remaining))

From an organizational perspective, this high turnover speaks to the importance of a) trying to reduce the turnover b) accommodation of the turnover through appropriate membership fee strategies for occasional and new players

[Gary Ruben]

Re: Membership retention

http://victoriachess.com/cfc/retention2.php

Jonathan Berry is fond of telling us that back in the day, something like 30% of all memberships lapsed within one year, not to be renewed.

It sounds like that percentage would have been achievable without really having to work at it.

The Scarborough club used to collect from all the members and send the dues to the CFC. I don't recall if the turnover was quite that large but would assume those who didn't renew their club membership would not have separately renewed with the CFC after they stopped playing.

The CCCA at one time collected an affiliation membership from members who were not CFC members and didn't play over the board. I don't know how the CFC counted that money. The CCCA had a good sized yearly turnover so some of the affiliated members would not have renewed.

I think lots of Canadians play organized chess. They just don't play CFC organized and rated chess.

[Roger Patterson]

Odds of playing a GM &amp; expected score

Some work on the distribution of ratings and results in games played is now posted at: http://victoriachess.com/cfc/opponents2.php. This work is part way along the process of getting a better idea of automatically identifying people who have exceptional results and adjusting their ratings to reflect their new level.

Of interest:
1) a graph of the probability of playing someone rated x points lower/higher than you (below). you can see that the odds of an 1800 player playing a GM strenght opponent is about 1 in 10,000.

2) Actual scores versus expected scores. (2nd graph) It seems that the probability of beating a higher rated player is underestimated by the standard expected score formula used in the rating system. I'm not the first to notice this - for example Jeff Sonas has found similar results in the FIDE system (see http://www.chessbase.com/newsdetail.asp?newsid=562). So, eventually, something needs to be fixed here.

The wiggles in the curve at very high ratings differentials are basically juniors with 300-600 ratings beating 1200-1500 rated players. Not too many games at this level though.

A graph of expected score for various rating levels (see the link at the top) shows that the rating level doesn't really matter.

The sample size is 111,499 games from Swiss tournaments from ~2005-2010.