From: Ross Dickson <rdickson <at> ACHILLES <dot> NET>
Date: 29 may 2003
Subject: Re: Dual ratings

Larry
You concluded from my earlier note, "the formula you give doesn't make
sense".

I agree, my response was truly incomprehensible. Please accept my apologies
for the confusion I may have caused.

It turns out that for some reason, the least squares application treated
the Pan-Atlantic data as text and assigned some arbitrary numbers to it.
One second of reflection on the results on my part would have been
sufficient to reveal that error, but in my haste, I was negligent.

With the data treated properly, I get exactly the same constants for the
linear regression that Andy reports.

My earlier comment, to the effect that there seems to be a periodic
component to the relationship does seems valid. Once the linear
relationship is subtracted from the data, there remains an apparently
non-random periodic component to the data, with a period of about 140
Pan-Atlantic points below Pan-Atlantic scores of 2100 and 75 points above
scores of 2100.

If there is some underlying reason, it may be in the way one of the
measures is determined. This, in turn, might indicate a source for the
discrepancies between the two measures. As usual, more data might
illuminate this effect. In the absence of more data to verify the effect,
it would be interesting to examine each of the measures independently to
see if there is a periodic component buried within their calculation.

Best regards

Ross