NFL Line Predictions By the Numbers

In 1980 I began to rank NFL teams as a hobby. My intent was to rank NFL teams strictly by numbers, without emotions or taking things like injuries into account. My first rankings were done using paper and pencil and my algorithm was very crude. Over the years I refined and computerized my rankings. After a while I discovered the rankings were very useful as a platform to a predictive algorithm designed to wager against the Vegas line.

My idea is very simple really. Just as it is far easier to predict general stock market trends than it is to predict individual stocks, it is easier to predict the general trend of a season than it is to predict individual games. This is an imperfect analogy but it is the best I can come up with. Translated into wagering terms, it means that if you bet the same amount on every game of every week, and if you use the same system for each game regardless of the situation, then over time, you will come out ahead.

For instance, in 2007 of last year I bet $100 against the line for each game. Thus, when all 32 teams played, I placed 16 individual bets of $100 one way or the other against the line, depending on what my algorithm told me to do. After the super bowl, I made a little over $1,120 in profit for that season. I did the same for the 2007 – 2008 NBA season, and made a profit of just over $6,000, betting $100 on every game based solely on the algorithm's numbers. Again, my idea is that over time, general trends will work for you and will cover up some of the problems caused by betting on only one game. In real terms, if you bet on all the games, and get just over 52% of them right in terms of line wagering, you will come out ahead.

My algorithm assumes nothing at the start of the year. All teams are considered equal. It does not matter who won the Super Bowl last year or what happened during free agency. This means my algorithm is virtually useless for the first two or three weeks. I do not actually start using it for wagering until week 5 when there are enough numbers to build a good picture of who is good, who is middle of the road, and who is downright terrible. As time passes, the better it is at predicting.

Starting from week 11 I will post my algorithm's predictions. I will consult the same online sporting book so as not to find the best odds in favor of my algorithm. And because this is my blog, I will add personal comments. I will also publish how well the algorithm did the previous week.

I will also publish my algorithm's over/under predictions, but it is not really designed for this although it staggered to a very small profit last year. This algorithm is designed to predict against the line.

I will publish predictions later on in the week.

I will briefly discuss some of the philosophy I employ for designing this algorithm. As I said earlier, it is much easier to predict long term general trends than it is to predict individual games. This is why sports books are rich and most of us are not. Everyone knows, however, that sometimes, catastrophic events happen to a team. Take Tom Brady's injury, for example. What we normally do when something like this happens is get emotional and assume a team will fall apart afterwards. To continue with this example, my algorithm would say that over time, we will see how the Patriots react because over time, their general trends will smooth out and we will have a much clearer picture of where they stand relative to other teams. Eventually, at least to this point, we now know the Patriots are still pretty good. So I might have lost some money for the first couple of weeks after the Patriots lost Brady (although this happened in week 1 so I was not using the algorithm as yet), but eventually, the algorithm will right itself. Over time, I come out ahead because wagers based on true objective evaluations win out over wagers on individual games.

Of course, nothing is perfect. A couple of years ago, my algorithm just could not get a handle on the Jets, and this helped bring me down in what was a brutal year. But for most years, this algorithm does quite well. Last year, for instance, the Algorithm consistently said to pick against the Patriots covering their huge spreads during the month of December. It was right more times than not. It pretty much did the same thing for the Cowboys. As a Cowboys fan, I, along with just about everyone else, expected the Cowboys to climb out of their funk, but the algorithm basically said: "Until they do, assume nothing." Sadly, they never did.

So stay tuned for this year's predictions. We'll see whether this philosophy has any merit. So far, my algorithm is up $564.76 but can it keep this up? We'll see.