BGonline.org Forums

Linearity

Posted By: Timothy Chow
Date: Thursday, 13 May 2010, at 8:57 p.m.

In Response To: Linearity (Chuck Bower)

As far as I know it's just based on anecdotal evidence from people examining the ERs/PRs of their own games and of publicly available games.

The main difference between Snowie/GNU/XG methods of computing these numbers has to do with "forced" moves. I would argue that trying to compensate for forced moves is not going to make that much difference in the long run. This will certainly be true if the percentage of forced moves I find myself making is more or less independent of my skill level. For players who are serious enough about the game that they care about their ER/PR, this seems to me like a plausible assumption.

Now, perhaps I'm wrong and (for example) weaker players turn out to make forced moves more often, because they're on the bar more often or something. It still seems likely to me that this dependence, if it exists at all, can be pretty well approximated by a linear function of skill. If this is the case then ER and PR should be pretty much linearly related as long as you average over enough games to smooth out the noise.

It's possible that differences in playing style could noticeably affect the percentage of forced moves a player has to make. Then ER and PR would not be linearly related. By the way, if this is the case then I'm not sure whether ER or PR is a better measure of what we intuitively think of as "skill." I think Doug Zare thinks that ER actually makes more sense here, and for this reason he was opposed to the GNU programmers' decision to make up a new definition of error rate.

Messages In This Thread

• Elo difference mapping to Snowie ER
  Klaus Evers -- Sunday, 9 May 2010, at 10:03 p.m.
  • Elo difference mapping to Snowie ER
    Leonardo Jerkovic -- Sunday, 9 May 2010, at 10:16 p.m.
    • Elo difference mapping to Snowie ER
      Christian Plenz -- Monday, 10 May 2010, at 2:17 a.m.
      • Elo difference mapping to Snowie ER
        Rich Munitz -- Monday, 10 May 2010, at 2:34 a.m.
        • Elo difference mapping to Snowie ER
          Iancho Hristov -- Monday, 10 May 2010, at 4:57 a.m.
        • Elo difference mapping to Snowie ER
          Neil Robins -- Monday, 10 May 2010, at 5:36 a.m.
          • Elo difference mapping to Snowie ER
            Timothy Chow -- Monday, 10 May 2010, at 3:38 p.m.
            • Elo difference mapping to Snowie ER
              Rich Munitz -- Monday, 10 May 2010, at 5:33 p.m.
            • Linearity
              Chuck Bower -- Thursday, 13 May 2010, at 7:34 p.m.
              • Linearity
                Timothy Chow -- Thursday, 13 May 2010, at 8:57 p.m.
              • Linearity
                Rich Munitz -- Thursday, 13 May 2010, at 11:46 p.m.
                • Linearity
                  eXtreme Gammon -- Friday, 14 May 2010, at 12:48 a.m.
                  • Linearity
                    eXtreme Gammon -- Friday, 14 May 2010, at 1:04 a.m.
                    • Linearity
                      Tom Keith -- Saturday, 15 May 2010, at 12:11 p.m.
  • Reminder: I was interested in Snowie Error Rate difference
    Klaus Evers -- Monday, 10 May 2010, at 7:14 a.m.
    • Reminder: I was interested in Snowie Error Rate difference
      Stick -- Monday, 10 May 2010, at 7:50 a.m.
    • Reminder: I was interested in Snowie Error Rate difference
      Neil Robins -- Monday, 10 May 2010, at 3:32 p.m.
      • Agreed: 100 elo ~ 2.6 Snowie ER
        Klaus Evers -- Monday, 10 May 2010, at 6:24 p.m.