VT/GT Statistical Comparison

VT versus Georgia Tech: What do statisticians think?

So last week statisticians pretty much agreed that VT should win by about 10 points.  Unfortunately they believed it would take 60 minutes and didn't realize I wouldn't have drank a beer down to the label yet when they had a 10 point lead.  So why the awful predictions?

Predicting sporting events is a very inexact science and typically ignores intricacies such as BC previous performances against blitzing teams.  Aside from that, even casinos - who do take such things into consideration - aren't particularly accurate at setting the spread.  Think back to your Stat 100 class in Hutcheson.  The standard deviation between the spread of games and the final result is usually about 17 points.  So about 2/3 of the time Las Vegas is within 17 points of being correct - a huge margin!  Statisticians are marginally better but not enough to talk about. 

This week's statistical comparison will include the following ratings and, if available, the statistical prediction based on the rating:

• Massey Ratings (http://www.masseyratings.com/).  These are Massey's own ratings which include scoring margin, not to be confused with his BCS rankings which will not post until they achieve some level of credibility during the season.
• Sagarin (http://www.kiva.net/~jsagarin/sports/cfsend.htm), perhaps the most popular of all statistical rankings.  The first is his ELO ranking, which is used for the BCS and fits their criteria; the second is PREDICTOR which takes into account scoring margin and is much more accurate for determining how good teams are.
• Fremeau Index (http://bcftoys.blogspot.com/2009/09/week-2-fei-ratings.html) as suggested by furrer4heisman.  From the site this system is "based on drive-based Game Efficiency data that rewards playing well against good teams, win or lose, and punishes losing to poor teams more harshly than it rewards defeating poor teams" where Game Efficiency is "Measurement of the success rate of a team scoring and preventing opponent scoring during the Competitive Possessions of a game as GE = ((Points For - Points Against)/7) / (Total Competitive Possessions/2)" and competitive possesions are basically possessions that potentially matter.
• FACT (http://prwolfe.bol.ucla.edu/cfootball/rothman.txt) stands for the Foundation for the Analysis of Competition and Tournaments and was created by David Rothman originally to rank chess players worldwide.  Peter Wolfe, whose own rating will publish for the first time next week, is gracious enough to continue publishing the fACT rankings since Rothman's death in 2004.

Here is how we stack up:

System	VT Rating (Rank)	GT Rating (Rank)	Highest Rated	Prediction
Massey	2.425 (3rd)	2.0139 (16th)	2.626 - Alabama	VT by 7.2
Sagarin ELO	89.91 (3rd)	84.25 (12th)	94.07 - Florida	VT by 2.7
Sagarin PREDICTOR	91.46 (3rd)	81.00 (22nd)	93.25 - Florida	VT by 8.4
Fremeau	-	-	-	-
FACT	95.01 (2nd)	74.30 (15th)	105.32  - Alabama	-