This was originally a reponse to Mark Hopkin's analysis of 1996 rankings.
Mark has changed "Who's Got the Best Rating System" since I wrote this
but my basic argument still holds.
I was surprised to read Mark Hopkin's "Who's Got the Best Rating
System" and I am curious to see if those who have their own systems
agree with me. I agree with him on the fact that there are two types
of systems. There are Predictive and Retrodictive systems. I have
never heard the term Retrodictive but I think that is a good name so
I will use it here. I think that it is important that we stress to
those who see our rating which type of system it is and what that
means. Otherwise, they will be confused by how certain teams can be
ranked in certain places.
A Predictive system can easily be tested for quality be seeing how
well it does what it says it will do. If it is meant to predict
which teams will beat the spread then we can see how well it does
that. With a predictive system the results justify the system.
A Retrodictive system does just the opposite. It rewards teams for
what they have done in the past instead of trying to predict the
future. An algorithm is developed based on the criteria that awards
points in the manner that is deemed to be the most fair. Then
the algorithm can be applied in the exact same fashion to every team.
In this manner you eliminate the potential bias that it is possible
that some of the voters of the AP and USA Today may have. If you
believe in the criteria on witch the algorithm is based then you must
agree with the results. You shouldn't change the algorithm so the
results come out how you want them. If you do this you are subjected
to the same potential bias as the AP voters. If you judge a
retrodictive system by determining how close the results are to what
you think they should be them you might as well just start with the
results you wanted in the first place and forget the computer. This
is exactly what the AP voters do. With a retrodictive system the
system justifies the results. I don't think you can really decide
who has the best retrodictive system by looking at the results as
Mark Hopkins's is trying to do.
It is clear that there will be differences in opinion on what
constitutes a fair algorithm just as everyone would use different
criteria to rank teams in the AP poll if they had their own vote.
This means that a difference of opinion on the criteria would cause
my system to be different from yours. However, we are still better
off than we are when we just vote on it. At least now we know that
the criteria, which may be different for everyone, are being applied
the same for every team and there is no bias toward local or favorite
For instance, I believe that margin of victory should not be taken
into consideration in a retrodictive system. I think that this is
dependent on the coaches personality, the game plan, an other factors.
This is not what coaches gameplan to accomplish. I can see how
someone else might say that I am not using all of the data available
to me and thus there is a better way. I understand and respect their
I also understand how a system that uses margin of victory could
penalize a team for winning provided that they don't win by the
margin that they should have. I don't understand how one could say
that a team could be hurt by winning in a system that does not take
into account margin of victory since all a team can do is win. I
don't understand that opinion but I still must respect it.
Since the point of a retrodictive system is to eliminate potential bias we
shouldn't judge the system based on the results. If you agree with the
algorithm that a retrodictive system uses then you should believe the results.
If you don't then you shouldn't. I would like to know if everyone else agrees
Mail Jeff Bihl