From: Larry Kaufman comcast net> Date: 18 aug 2007 Subject: Re: PanAtlantic and Dojo24 [was: Call for data] There is an obvious flaw in simply using a linear equation to make=20 these comparisons. Neither rating system allows for negative ratings.=20 Therefore, any player with a Pan-Atlantic rating less than your intercept= =20 value (1190 overall, 1215 for Europe) would automatically appear underrated= ,=20 even if his Dojo rating is just one (or zero). Since all of the American= =20 players in your sample are over 1500, this strongly biases the result of th= e=20 comparison to make Europe appear underrated. I would suggest redoing the calculation with the intercept value=20 assigned. It's probably not fair to use zero, as a beginner in Europe is= =20 weaker than a typical beginner in Dojo, but maybe an intercept of 400 might= =20 be reasonable (this used to be the minimum, and now might be considered the= =20 rating of a novice but not a beginner). I should also tell you to remove the datapoint that shows a Dojo ratin= g=20 of 2350, as I happen to know that this handle is often used by visiting pro= s=20 and so is grossly inflated. This is of course one problem with Dojo ratings= ;=20 you never know who is really playing the games. Larry Kaufman ----- Original Message -----=20 From: "Daniel Toebbens" gmx de> To: topica com> Sent: Saturday, August 18, 2007 1:35 PM Subject: PanAtlantic and Dojo24 [was: Call for data] It's me again and I got me some more data points: I saw a swedish player on= =20 Dojo24 this afternoon, did run his nickname + 'shogi' through google and a= =20 swedish list 'Who is who on Shogi Dojo'. It was kind of old, but about half= =20 of the players are still active in both rating systems. With this I have no= w=20 15 american players and 20 european ones. The distribution over strength is= =20 also quite good, though I still have only one american player in the lower= =20 regions. Now we finally start to get some interesting results: 1) The relation is approximatly linear over the whole range. Only one=20 outlier appears at too low a PanAtlantic rating - with 200 points in Dojo24= =20 said player should easily make a rating above 1000 with a few more officia= l=20 tournaments. Best fit is PanAtlantic =3D 1190 + 0.47 * Dojo24. About everybody falls within +-250 PanAtlantic points of this line. 2) But it gets really interesting, when you do a cross-atlantic comparison:= European data alone fit to PanAtlEurope =3D 1215 + 0.38 * Dojo24 American data alone fit to PanAtlAmerica =3D 1380 + 0.39 * Dojo24 Thus over the board the American ratings are about 160 points higher than= =20 the European ones. That's more than one full rank. Results must still be= =20 taken with care; the scattering of especially the American data is quite= =20 high, and I have data of two really high rated European players only. Here is the full list of data points, again rounded to 5ers: European ratings: PanAtl Dojo24 650 215 1305 695 1325 135 1325 730 1340 645 1355 685 1385 50 1390 365 1420 800 1470 1000 1490 780 1550 500 1575 895 1655 1300 1755 1380 1785 1055 1815 1365 1845 1325 1855 1910 2065 2030 American ratings: PanAtl Dojo24 1515 610 1555 1235 1740 1175 1895 1220 1930 1285 1970 1420 1995 1660 2000 1555 2000 1745 2030 1265 2075 2350 2095 1480 2140 1455 2160 1780 2280 1760 ------------ --=20 Der GMX SmartSurfer hilft bis zu 70% Ihrer Onlinekosten zu sparen! Ideal f=FCr Modem und ISDN: http://www.gmx.net/de/go/smartsurfer --^---------------------------------------------------------------- This email was sent to: = bobd science uva nl EASY UNSUBSCRIBE click here: http://topica.com/u/?a2i6Ys.= a7fLI9.= Ym9iZEBz Or send an email to: shogi-unsubscribe topica com For Topica's complete suite of email marketing solutions visit: http://www.topica.com/?p=3DTEXFOOTER --^----------------------------------------------------------------