On Tue, Jul 8, 2008 at 5:18 PM, Brad Mills (NSA Club 620)
<[EMAIL PROTECTED]> wrote:
> Well, for what it's worth:
>
> http://cross-tables.com/results.php?p=17892
>
> I'm not sure which iteration of Quackle was used in the Computer vs. Human
> Challenge, but my guess would be Toronto Player (which I think has since
> been named Championship Player).  Quackle went 4-2 against David Boys, who
> was rated 1861 at the time and was the only person to play Quackle more than
> once in "rated" competition.
>

Mr. Sapphire Brand here is clearly attempting to goad me into giving
you a straight answer :-)

So here's what it looks like to me ...

>From that cross-tables page, the unofficial NSA rating of Quackle is
1930.  It appears to me that Quackle's rating is unofficial since the
Cross-tables data indicates the games Quackle played in were not part
of a rated tournament - for example Chris Cree was rated 1897 and lost
two games, but his rating afterwards was still 1897.  The NSA's win
expectation curve gives an estimated win expectation of 50% or greater
against Quackle for players rated 1930 or above:   Joel Sherman, Matt
Graham, Adam Logan, Brian Capaletto, and David Gibson.

However the rating system isn't designed to be extremely accurate in
making this prediction; it's designed to efficiently sort players into
divisions with a reasonable degree of accuracy; and Quackle's rating
is based on a mere 38 games - there are many reasons why the actual
outcome can be expected to vary from the prediction.

These remarks are mine, not the NSA's or the ratings committee's.

The ratings committee is working on improving the rating system -
http://www2.scrabble-assoc.com/subitems.asp?id=36&sid=72&ssid=109.

Kevin Leeds.

Reply via email to