Rated players under 1000

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #46
    Re: Rated players under 1000

    For clarification purposes, the correct spelling of Jack's last name is Maquire, not Miguire. # Jack Maquire prolific chesstalk blogger and chess coach.

    For retro purposes Sunny Burnett's name on Miami Vice was spelled Burnett # previous Miami Vice episode # Don Johnson

    Comment


    • #47
      Re: Rated players under 1000

      Originally posted by Fred McKim View Post
      So yes 1400 vs 1600 and 1600 vs 1800 and 1800 vs 2000 and 2000 vs 2200 will all yield the same averages.
      Should yield the same averages.

      Comment


      • #48
        Re: Rated players under 1000

        Originally posted by Ed Seedhouse View Post
        Should yield the same averages.
        It does, actually. The percentage of drawn games might go up a bit, but the overall result remains surprisingly constant.

        Comment


        • #49
          Re: Rated players under 1000

          Originally posted by Mathieu Cloutier View Post
          It does, actually. The percentage of drawn games might go up a bit, but the overall result remains surprisingly constant.
          Not consistent with the data I've seen. The variance is not very significant but it exists. It would be surprising if it didn't.

          Comment


          • #50
            Re: Rated players under 1000

            Originally posted by Ed Seedhouse View Post
            Should yield the same averages.
            Has this ideal 75% been studied?
            I looked at a sample event, the Toronto Labour Day for games 200 rating points apart plus/minus 25:
            Open +4=2-0 83.3%
            U2000 +16=4-3 81.8%
            U1800 +15=9-6 65%
            U1400 +14=4-4 72.7%

            Strange junior went up from 415 to 770 with a 1006 performance. 994 drawing master to 1141 not so bad. The 933 to 1020 and 860 to 991 are still underrated.

            Comment


            • #51
              Re: Rated players under 1000

              Originally posted by Erik Malmsten View Post
              Has this ideal 75% been studied?
              I looked at a sample event, the Toronto Labour Day for games 200 rating points apart plus/minus 25:
              Open +4=2-0 83.3%
              U2000 +16=4-3 81.8%
              U1800 +15=9-6 65%
              U1400 +14=4-4 72.7%

              Strange junior went up from 415 to 770 with a 1006 performance. 994 drawing master to 1141 not so bad. The 933 to 1020 and 860 to 991 are still underrated.
              it has been looked at. Results do not match the theoretical curve but do not seem to depend on strength (although data is noisy). Results for the CFC that I derived from data 1996-roughly 2010 below. Jeff Sonas has similar results (that results actually scored do not match the expected result) for FIDE.



              Comment


              • #52
                Re: Rated players under 1000

                Originally posted by Roger Patterson View Post
                it has been looked at. Results do not match the theoretical curve but do not seem to depend on strength (although data is noisy). Results for the CFC that I derived from data 1996-roughly 2010 below.
                This is super-interesting, thanks for posting Roger.

                How do you explain a spike to 0.25 at the 1500 mark? What was the total sample size of games played at this rating differential? Are we simply looking at a situation of only four games played and a fluke win in one of them by a lower-rated opponent? :)

                Comment


                • #53
                  Re: Rated players under 1000

                  Originally posted by Vadim Tsypin View Post
                  This is super-interesting, thanks for posting Roger.

                  How do you explain a spike to 0.25 at the 1500 mark? What was the total sample size of games played at this rating differential? Are we simply looking at a situation of only four games played and a fluke win in one of them by a lower-rated opponent? :)
                  noise. The number of games played with a 1500 point rating differential is small and you can see the curves picking up noise at high rating differential. The full data set is all games rated in the CFC database but I would have to go back to the original data (and possibly I don't have it any more) to tell you how many at each rating differential point. But I would guess the spike is due to one or two upset wins.

                  All of this stuff was previously posted. If you go back and look and my old posts, you can find the original discussion and reaction by others to it.

                  Comment


                  • #54
                    Re: Rated players under 1000

                    Also the data is old enough that there will be many rated forfeits in there that were never actually played. Unless Roger had a way of filtering those out...
                    Christopher Mallon
                    FIDE Arbiter

                    Comment


                    • #55
                      Re: Rated players under 1000

                      Originally posted by Roger Patterson View Post
                      it has been looked at. Results do not match the theoretical curve but do not seem to depend on strength (although data is noisy). Results for the CFC that I derived from data 1996-roughly 2010 below. Jeff Sonas has similar results (that results actually scored do not match the expected result) for FIDE.
                      This is to be expected except in exceedingly large samples. In any event you can't just analyze this stuff by eye. A chi-square test should be done at the very least. I expect that the observed variances would not be statistically significant, but that's just a guess until such a test (or another equal one) is made.

                      Comment


                      • #56
                        Re: Rated players under 1000

                        Regarding the play-up comments, I had a look at the Toronto Labour Day Open as representative of a large Ontario tournament.
                        There appeared to be pretty strict control over play-ups. I found only one player allowed to play up by more than 100 points (Paul Panayotou in the U1800 with a 1218 provisional rating).
                        John Brown, or someone similarly rated would play in the U1800 Section and would not play anyone rated below 1300.
                        My advice to players who do not wish to face very low rated players is to play in tournaments with sections such as the Toronto Labour Day Open that have strict controls over playing up.
                        Paul Leblanc
                        Treasurer Chess Foundation of Canada

                        Comment


                        • #57
                          Re: Rated players under 1000

                          Originally posted by Ed Seedhouse View Post
                          This is to be expected except in exceedingly large samples. In any event you can't just analyze this stuff by eye. A chi-square test should be done at the very least. I expect that the observed variances would not be statistically significant, but that's just a guess until such a test (or another equal one) is made.
                          by eye, a chi squared test is not required. the curves are smooth out to +-500 point or so and the difference between expected and actual is clearly real. The total number of data points is the total number of games played in the CFC, at a guess without looking it up, of the order of 75000 games. This is more than enough - the variance for small rating differences is probably less than the width of the line on the graph. It is only at the large rating differences that there are not many games.

                          The Sonas data (http://en.chessbase.com/post/the-son...tter-than-elo- for example - he has published other articles on this as well) also makes it clear that the distribution of actual is different. He plots the full set of data points he has and it is clear they are not centered on the theoretical expected result.

                          Comment


                          • #58
                            Re: Rated players under 1000

                            Originally posted by Roger Patterson View Post
                            by eye, a chi squared test is not required.
                            But it would be nice to have it.

                            The Sonas data (http://en.chessbase.com/post/the-son...tter-than-elo- for example - he has published other articles on this as well) also makes it clear that the distribution of actual is different. He plots the full set of data points he has and it is clear they are not centered on the theoretical expected result.
                            Going by memory I believe the expected distribution of performance in a single game is not "normal" as assumed by Elo. More modern variants on Elo use the Logistic distribution for that reason.

                            Comment


                            • #59
                              Re: Rated players under 1000

                              Originally posted by Ed Seedhouse View Post
                              But it would be nice to have it.
                              all right, I found the file. It's a little over 111000 games. A chi square test yields 0 to 30 decimal places which is the most Excel will let me do. At a rating differential of 125 points, the curves are a little over 7 standard deviations apart.

                              If you can't tell by eye that two distributions are different when the statistics are that overwhelming, you need to rethink how you approach this type of question.

                              [edit: test done on rating differential range +-525 points]
                              Last edited by Roger Patterson; Friday, 22nd September, 2017, 04:20 PM.

                              Comment


                              • #60
                                Re: Rated players under 1000

                                Originally posted by Bob Armstrong View Post
                                Hi John:

                                I am not directly answering your question, but the Scarborough CC facts below may be interesting:

                                1. The U 1400 Section (There are three sections) at SCC, has just this year become an exclusively junior section. As I understand it (I was not at last Thursday's SCC AGM where this was decided), now all adults under 1800 play in the middle U-1800 section (Even if their rating is 1399 or lower - even 800).

                                2. Of the currently posted pre-registration list for the first new tournament of SCC's 2017-8 season, the Howard Ridout Memorial Swiss, starting tonight, of the 22 pre-registrants in the U 1400 junior section, I would bet 50% are rated 999 or lower or unrated (Not sure because ratings not given on the pre-reg list posted here on Ct). And I would expect when Rd. 1 starts tonight at 7:30 PM, the total juniors in the U 1400 section will be at least 30 registrations, and my 50% figure will still hold. I cannot report after the round on this, because SCC usually does not post the players rankings in the sections at the start of the round, including ratings.

                                But I will advise once the standings have been posted on the SCC website later in the next week (They include current rating).

                                Bob A
                                Hi John:

                                I said I would follow up with some facts:

                                1. U 1800 section - all adults U 1800 must play in this section......even if their rating is U 1400 (Since our U 1400 section is now juniors only). In this section, there is only one adult under 1000 (And this person is at 988).
                                2. U 1400 (Juniors Only) - out of 50 juniors in this section as of Rd. 2, 21 are rated U 1000 (This does not include unrated's, since we were discussing only those with a CFC rating).

                                Bob A

                                Comment

                                Working...
                                X