2013 Regular Season College Football Back Test

I'm going to reiterate that this blog is an ongoing real-time experiment designed to establish whether or not it's possible to successfully transfer my basketbal handicapping skills into college football. It is widely documented that the "best" football handicappers rarely exceed a 58% accuracy mark. My efforts in college basketball have yielded a 80% success rate so I felt challenged to see whether or not I could transfer my basketball skills into football.

My analytical background as a trader and analyst made it natural for me to want to back-test my football handicapping theories. About one month ago I set about constructing a database of all college FBS teams for the 2013 college football season. I programmed my analytics into the 2013 database based upon efficiency, explosiveness, ability to create turnovers and performance variance. The specifics of those analytical formulas will remain proprietary. I excluded any games that had a line in excess of 28.5 points. Any match-up that contained a non-FBS team was also automatically excluded.

The database was constructed using real-time data. For example, I used only information that was available before the match-up was actually played. If a team had played only three games before the match-up I was analyzing, then I only used those three games of data. For my analytics to have any predictive value, however, I need a given team to have played at least three games. As a result the first football handicapping selections aren't available until the fourth week of the college football season. This will also be true for the 2014 college football season. It takes at least three games for my analytics to have predictive value.

The results were only mildly surprising to me based upon my previous basketball success and the fact that I'm using the same general principles to analyze both football and basketball. Nonetheless I am quite sure that there will skeptics out there who believe that such a success rate is impossible. This group will be quite sure that I have cheated somehow in order to achieve such a phenomenal handicapping success. At this point I have only my word that the back-test was legitimate. As they say, that and $3.00 will buy you a cup of coffee!

Nevertheless here are the actual results of my 2013 back-test. It will be interesting to see if I can come close to duplicating these results during the 2014 season. I categorized each game as either a 1) olive pick (modest edge on outcome), 2) yellow pick (significant analytical edge on outcome) or 3) orange pick (strong significant edge on outcome). Each pick is sorted by whether is was an RD (road dog), RF (road fav), HD (home dog) or HF (home fav).

As you can see the yellow picks were the strongest category. I theorize that the orange picks flagged other analytical metrics that other handicappers use, particularly the sportsbooks, which made made the lines larger and thus harder to overcome That aside, the yellow picks were 80% successful against the consensus sportsbook lines in Las Vegas. Not bad!

Lastly, a couple of basic observations. In college football each team will recieve an average of 13-14 offensive possessions per game. A great offensive team will score on about 50% of their possessions. The average is closer to 35%, however. If one teams loses the turnover battle by one turnover they can still cover the line. A two turnover deficit, however, will ensure that a team will not cover the line. It makes sense if you think about it. Assume team A suffers a two turnover deficit against team B. Team B has 16 chances to score while team A only has 12. A "pick-six" or a special teams touchdown (punt or kick-off return) is just as significant as an extra turnover.

Again, it will be an interesting challenge to see if I can come close to duplicating these results in real-time during the 2014 football season. The first picks won't be available until the fourth week of the upcoming season, sometime in late September.