2004 Football Study

After a sub-par football season in 2003, I decided to take a look at my methods during the summer prior to the 2004 season. I had previously done research on the predictability of my situational angles, but I wanted to do a more detailed analysis of that aspect of my handicapping. At the time, I also had 4 years of results from my math models in the NFL and College and wanted to find the relationship between my math prediction, the line, and the chance of each team covering the pointspread based solely on the math. I had also incorporated fundamental indicators into my analysis a few years prior and wanted to test the validity of using statistical profiles to predict future results.

Researching Situational Analysis

In studying the predictability of my situational angles, I needed to go back in time and track the results of each angle based on the win-loss record of the situation heading into each season, the number of parameters used in deriving the angle, the recent results of the situation (the record over the last 3 years), and how many years the situation went back. I then recorded the record of that situation for that year and, of course, I only charted an angle starting with the season after I discovered it – so as to avoid back-fitting. I was pleased to discover that all of situational angles that I use had a combined 56% win percentage in both the NFL and College Football. I also discovered that the more parameters (or factors) used in deriving an angle, the less predictive it was relative to its record. In other words, a 100-40 ATS situation with 5 parameters was more predictive than a 100-40 ATS angle that used 9 parameters. In fact, the 100-40 ATS angle with 5 parameters had a 57.5% chance of winning the next time it applied while the 100-40 ATS angle with 9 parameters had a 55.2% chance of winning in the future, in general. I also discovered that the longer an angle went back in time, the more it held up in the future – so finding an angle that has worked great the last 5 years, but was only 50% the 15 years prior to that is pretty much worthless. My findings were all pretty much what I expected, but I also learned that there are certain types of parameters that are more predictive than others.

One of the keys to my success in the 2004 season was the knowledge of exactly what to expect from every situational angle that I use. There are some services that may get you to think that they have a game that can’t lose because of some 24-0 ATS angle they’ve come up with. An angle that is 24-0 probably has more than a dozen factors and a 24-0 ATS situation with 12 parameters would have a 58.5% chance of covering in the next game – which is hardly a “Lock”.

Researching my Math Model

Prior to the 2004 season, I had 4 years of results for my math models in the NFL and College, and while I knew that my models had produced winning results, I needed to know the relationship between my math prediction, the actual pointspread, and the success of the model. After studying the results, I can now determine a team’s chance of covering the spread, based solely on the math, at any given pointspread given the mathematical prediction of the game (after adjusting for injuries and current personnel). My model has worked better in the NFL than in College, so a 4 point difference between my math model and the line in the NFL is more significant than a 4 point differential in a college game. I can now assign a chance of covering the spread for each team in every game based solely on the math based on the actual past performance of my math model. It is important to note that none of the data used in the 4 year study was used in the derivation of the model (i.e. there was no back-fitting the data), so the future predictability of the models should be in line with the past performance. Of course, the more years of data I have, the more accurate I can be in predicting the future results of the models and I have updated my data to include last year’s results.

Researching Fundamental Indicators

The research that I did to find the predictability of my fundamental indicators and statistical profile analysis is very similar to the study I did for the Situational Analysis, but I found that fundamental indicators actually are more predictive than the situational angles.

Putting It All Together

The purpose of all the research that I did last summer was to know exactly how predictive each element of my handicapping is and to find a way to combine those elements to give an overall assessment of each game based on a combination of Situational Analysis, Fundamental indicators, and my Math Model. An example of combined analysis is a game in which Team A applies to a 140-60-5 ATS situation that uses 6 parameters. Team B applies to a statistical profile indicator with a record of 86-28-4 ATS and my NFL math model favors Team A by 10.4 points when Team A is a 7 point favorite in reality. As discussed above, a situation with a record of 140-60-5 and 6 parameters has a 56.8% chance of winning if the line is fair. The fundamental indicator favoring Team B has a 58.2% chance of winning given a fair line and my math model would give Team A a 54.4% chance of covering at a line of -7 points. The trick is assigning a point value to the situation and the fundamental indicator based on their chance of covering at a fair line. In this case, the situation favoring Team A is worth 5.2 points while the fundamental indicator favoring Team B is worth 6.3 points. My math model favors Team A by 10.4 points, so adding the value of the situation and the indicator would result in an overall prediction of Team A by 9.3 points (+5.2 – 6.3 + 10.4 = 9.3), which would give Team A a 53.0% chance of covering at the line of -7 points. Obviously, things can become a lot more complicated when there are multiple situations and indicators applying to a particular game - which is most often the case, but my years of studying probability theory at Berkeley has given me the tools to sort through it all and come up with an accurate measure of the overall affect of the situations and indicators.