Page 1 of 2 12 LastLast
Results 1 to 15 of 17
  1. #1
    CC Candidate Master
    Join Date
    Dec 2015
    Posts
    124

    Comparison of ACF Quick and Classic ratings

    I've run a regression of the June 2017 ACF Quick and Classic ratings for 81 members of our club who have both ratings:

    ACF_Quick_vs_Std_ratings_June2017.JPG

    This suggests that ACF Quick ratings are roughly 100 points higher than Classic ratings in the 1000-2000 range which is important for rating prizes. However, this relationship may not be robust because of the small sample size, so has anyone run this regression for all players with both an ACF Classic and Quick rating?

    I would do so myself but the ACF IDs are only in the ACF Masterfile (which includes the Classic ratings but not the Quick ratings) but not in either of the rating lists.

  2. #2
    Reader in Slood Dynamics Rincewind's Avatar
    Join Date
    Jan 2004
    Location
    The multiverse
    Posts
    21,425
    I haven't but you could do something quick by looking at unique names in the quick and classic and ignoring the master file. The probability of a false match would be pretty small I would think.
    So einfach wie möglich, aber nicht einfacher - Albert Einstein

  3. #3
    CC International Master William AS's Avatar
    Join Date
    Jun 2008
    Location
    Adelaide
    Posts
    1,810
    I suspect that here in South Australia the difference is even greater than that.
    “A wise prince will seek means by which his subjects will always and in every possible condition of things have need of his government, and then they will always be faithful to him.” — Niccolo Machiavelli (1469–1527), Il Principe (The Prince)

    A wise prince will also realise that this can be achieved more successfully by means other than rape, pillage, greed, lies, deceit or misinformation.

  4. #4
    Monster of the deep Kevin Bonham's Avatar
    Join Date
    Jan 2004
    Posts
    35,602
    I ran a similar comparison using ze sekret Ratings Officer herbs and spices a few months ago. I looked at the whole sample and also at only adults aged 25-55. I just looked at averages, not regressions. In both cases Classic ratings were much higher in NSW, SA, Tas, marginally higher in Qld and much lower in Vic. For ACT much lower for the whole sample but about the same for the age-restricted sample. For WA not enough data. The biggest overall differences were NSW and SA which were both over 100 points on both lists. The age-limited difference in Vic was 140 points. This includes inactive players.

  5. #5
    Reader in Slood Dynamics Rincewind's Avatar
    Join Date
    Jan 2004
    Location
    The multiverse
    Posts
    21,425
    Attached is comparison for all players with both active quick and classical ratings broken down by state (1083 obs). Both NSW and SA seem to have the regressions with x-coeff closest to 1. Most other states are less than one. All have positive y-intercepts. It appears that VIC ratings would probably have a negative y-intercept if I excluded those with ratings < 1000 but I haven't done that analysis. VIC appears to have the lowest x-coeff of all states.

    ratings.jpg
    So einfach wie möglich, aber nicht einfacher - Albert Einstein

  6. #6
    CC Candidate Master
    Join Date
    Jun 2014
    Posts
    201
    Quote Originally Posted by Rincewind View Post
    Attached is comparison for all players with both active quick and classical ratings broken down by state (1083 obs). Both NSW and SA seem to have the regressions with x-coeff closest to 1. Most other states are less than one. All have positive y-intercepts. It appears that VIC ratings would probably have a negative y-intercept if I excluded those with ratings < 1000 but I haven't done that analysis. VIC appears to have the lowest x-coeff of all states.

    ratings.jpg
    What does this mean?

  7. #7
    Reader in Slood Dynamics Rincewind's Avatar
    Join Date
    Jan 2004
    Location
    The multiverse
    Posts
    21,425
    Quote Originally Posted by Frank View Post
    What does this mean?
    Quick ratings correlate well with classical ratings but tend to track a bit higher for players who have a rating that is less than 1000 and those from NSW and SA of any rating. Quick ratings tend to track lower than the classical rating for middling to stronger players from Victoria. The variance also looks worse for lower ratings across the whole dataset but I haven't calculated this it just appears that there are more outliers in that range.

    Note that there were only 10 datapoints for CAWA and so I wouldn't read too much into that. The next smallest group for foreign with 30 datapoints.
    So einfach wie möglich, aber nicht einfacher - Albert Einstein

  8. #8
    CC Candidate Master
    Join Date
    Jun 2014
    Posts
    201
    Quote Originally Posted by Rincewind View Post
    Quick ratings correlate well with classical ratings but tend to track a bit higher for players who have a rating that is less than 1000 and those from NSW and SA of any rating. Quick ratings tend to track lower than the classical rating for middling to stronger players from Victoria. The variance also looks worse for lower ratings across the whole dataset but I haven't calculated this it just appears that there are more outliers in that range.

    Note that there were only 10 datapoints for CAWA and so I wouldn't read too much into that. The next smallest group for foreign with 30 datapoints.
    Thanks!

  9. #9
    CC Candidate Master
    Join Date
    Jun 2014
    Posts
    201
    Quote Originally Posted by Rincewind View Post
    Quick ratings correlate well with classical ratings but tend to track a bit higher for players who have a rating that is less than 1000 and those from NSW and SA of any rating. Quick ratings tend to track lower than the classical rating for middling to stronger players from Victoria. The variance also looks worse for lower ratings across the whole dataset but I haven't calculated this it just appears that there are more outliers in that range.

    Note that there were only 10 datapoints for CAWA and so I wouldn't read too much into that. The next smallest group for foreign with 30 datapoints.
    Does this mean we need only to have one (a composite) rating?

  10. #10
    CC International Master
    Join Date
    Jan 2004
    Location
    Wynyard,Tas
    Posts
    2,026
    Quote Originally Posted by Frank View Post
    What does this mean?
    Not a lot of itself. For instance if a player is rated 2000 normal and 2020 rapid it would be unlikely (though possible) that the player plays slightly better with five or fifteen minutes on their clock than ninety. However how the difference between ratings correlates with other things (State, age, rating) may reveal something.

    Quote Originally Posted by Frank View Post
    Does this mean we need only to have one (a composite) rating?
    It depends whether you want the rating to be simple or meaningful.

  11. #11
    Reader in Slood Dynamics Rincewind's Avatar
    Join Date
    Jan 2004
    Location
    The multiverse
    Posts
    21,425
    Quote Originally Posted by Frank View Post
    Does this mean we need only to have one (a composite) rating?
    Well we are looking at a population statistic There is a lot of individual variability which is the key reason why we have a rating system (track individual performance).

    The strong correlation is not unexpected and indeed it would be disturbing if it was absent. The variations between states is interesting and perhaps more a function of rating practices in those states. For example the small number of CAWA datapoints is probably due to that state rarely rating allegro events. The NSW/CV differences are interesting but again probably explained by tournament rating practices.
    So einfach wie möglich, aber nicht einfacher - Albert Einstein

  12. #12
    Monster of the deep Kevin Bonham's Avatar
    Join Date
    Jan 2004
    Posts
    35,602
    My understanding is the CV difference is down to having rated a high proportion of mixed adult/junior quick events.

  13. #13
    CC Candidate Master
    Join Date
    Dec 2015
    Posts
    124
    Thanks very much Rincewind, as that's exactly what I was looking for!

    Can you please send me the dataset or specify the linear regression coefficients and correlation coefficient for the NSWCA subset?

  14. #14
    Reader in Slood Dynamics Rincewind's Avatar
    Join Date
    Jan 2004
    Location
    The multiverse
    Posts
    21,425
    The regression statistics for the NSW dataset are

    SUMMARY OUTPUT
    Multiple R 0.911456643973753
    R Square 0.830753213843896
    Adjusted R Square 0.830201921380847
    Standard Error 185.095735316568
    Observations 309

    ANOVA
    df SS MS F Significance F
    Regression 1 51627702.6472578 51627702.6472578 1506.91922985669 1.8874759409523E-120
    Residual 307 10517952.388341 34260.431232381
    Total 308 62145655.0355987

    Coefficients Standard Error t Stat P-value Lower 95% Upper 95%
    Intercept 268.829490615712 36.8883044738045 7.28766188770252 2.68414633442234E-12 196.243588618645 341.415392612778
    X Variable 1 0.925871711995182 0.0238509580136257 38.8190575601301 1.8874759409526E-120 0.878939674168868 0.972803749821496

    Dataset attached.
    NSWCA_SQ_JUN2017.zip
    So einfach wie möglich, aber nicht einfacher - Albert Einstein

  15. #15
    CC Candidate Master
    Join Date
    Dec 2015
    Posts
    124
    Thank you Rincewind, much appreciated.

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Similar Threads

  1. Skill differences (comparison of chess and go)
    By Schu in forum Games Australia
    Replies: 28
    Last Post: 28-01-2016, 11:58 AM
  2. SP vs Tornelo test comparison
    By Kevin Bonham in forum Arbiters' Corner
    Replies: 78
    Last Post: 12-06-2011, 05:14 AM
  3. Replies: 2
    Last Post: 07-11-2010, 11:09 AM
  4. Results from Bled Olympiad 2002 for comparison
    By News Bot in forum Chess Australia
    Replies: 0
    Last Post: 16-10-2004, 07:04 AM
  5. AMD 2500+ Comparison
    By Byrone in forum The 1337 Lounge
    Replies: 12
    Last Post: 17-09-2004, 07:50 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •