Week 7 rankings notes

> Week 7: #3 Williamsburg 7 @ #4 Solon 22

Given the entire season of data, computers think they are about even. Computer believes Solon played a great game, 18 points better than expected. Regina game seems to be really weighing on computer’s evaluation of Solon. Computer really thinks they should have beaten Regina by 13 points. Williamsburg is a parallel analysis: a horrible game against Solon and a great game versus Grinnell.

>Why are Akron-Westfield, Northwood-Kensett, and Woodbury Central all ranked so high with losing records? Just wondering. Thanks!

#6 Akron-Westfield A-1(3-4)
#13 Northwood-Kensett A-2 (2-5)
#14 Woodbury-Central A-1 (3-4)

I looked at all three. The common theme is that the computer really likes Northwest Iowa.

Results against Class A competition:
#6 Akron-Westfield: wins #14; loss #9
#13 Northwood-Kensett: wins #7; loss #2, #8, #10, #12
#14 Woodbury Central: wins #10, #12; loss #3, #6

Northwood-Kensett has played 5 games against the top 12 teama in Iowa. They also played the 1A #2 team, Mason City Newman.

> Bedford rolls their last 5 opponents and now drop from 8 to 16?

In week 6 the computer thought Bedford’s 40 point win over #25 (4-2) Southeast Warren was a competitive game and should be counted.

In week 7, Southeast Warren lost to Guthrie Center (6-40). This dropped Southeast Warren sufficiently that the computer now predicts that Bedford is 30 points better than Southeast Warren. And thus, computer now thinks the Southeast Warren vs Bedford game was not competitve and should be ignored.

Bedford’s ranking is now based on two games:

(13-6) win over 1A #40 Central Decatur (3-4)
(0-12) loss to 1A #19 Mount Ayr

A 7-point win over Central Decatur is not impressive at all.
Computer think Bedford’s 12-point loss was an upset; Bedford should have won that game.

So consequently, computer now ranks Bedford at #16.

Playoff games will provide better competition and more data.

> Are computer’s expected game margins an in input to the model?

The spreads are not inputs to the model. They are simply a way to understand what the model is thinking.

Model’s inputs are: actual game margins and home-field advantage. Model’s outputs are relative strengths of teams. Finally there is a process to ignore games that are expected to be non-competitive.

After the teams’ relative strengths are calculated, past games can be examined to see how closely the computer models reality. For example, the Resid column in team reports (expected spread – actual game margin) can tell you if the team had a good game or a bad game (positive Resid = team did better than expected, negative Resid = team did worse than expected).

So all predictions are strictly an output of the model and are never an input. However, they can give you insight into what the computer is thinking.