From: Ross Dickson <rdickson <at> ACHILLES <dot> NET>
Date: 28 may 2003
Subject: Re: Dual ratings

Larry
A linear least squares fit gives the following:
Pan Atlantic Score = -53.6 x ShogiClub24 - 2017
Pearson product moment correlation coeficient = 0..83

There seems to be a strong periodic component to the relationship, which I
ignored.
Ross

     The following is a listing of all ratings of players rated on both the
Pan-Atlantic and ShogiClub24 rating systems fulfilling the criteria given
below.  Of course, only those players whose identities on ShogiClub24 are
known to me could be included.  The first column is Pan-Atlantic, second is
ShogiClub24.

      2422  2284
      2285  1993
      2255  2013
      2195  1753
      2174  1925
      2118  1455
      2101  1465
      2094  1555
      2074  1064
      2044  1181
      2012  1453
      1980  1513
      1926  1378
      1924  1124
      1908  1161
      1852  1377
      1809  1298
      1784   993
      1769  1149

     All players listed have at least 80 games on ShogiDojo, which reduces
the effect of starting rating to under 5%.  All players listed have at
least 30 games on Pan-Atlantic, where starting rating plays only a tiny
role and is in any case based on known rank.  Player must be listed on 2002
or 2003 Pan-Atlantic rating list.  No names are given in case some players
prefer anonymity.
     There are 19 players on this list.  The average spread is 557.5
points, more than I expected.  We need to do a linear regression on the
above numbers.  If anyone can perform this for us easily, we would
appreciate it, so I don't have to rack my brain trying to remember how to
do it on my calculator or how to work downloadable software.

     Larry Kaufman