Pigskin Prognosticator – a Slight Return
This football season was like no other for me personally. Mostly that is because I wrote a system to evaluate football games (The Pigskin Prognosticator) and teams before the season kicked off.
This colored how I viewed every week and every game. I found I was way more interested in how my pigskin prognostications went then I was any particular team’s results. I now have a better understanding of how gamblers (and FF players) interact with football.
The tech part of this project is interesting personally. It will be my last PHP project. I have started a serious monogamous relation with Python. The jury is out on how many more times I’ll use MySQL but that is another story…
The database is nothing special as you can see from this early NFL version. I continued to tweak the database as well as the code all year. But I found that with a single NFL season your database can be simple, as long as it’s smart.
My formula is what turns out to be valuable in this exercise. All of the enabling code and database technology is needed, but not what is magic 🙂 That will be extremely important as my plan is to re-write this in Python well before football season next year.
There are 256 regular season games in an NFL season and 256 is the only 3-digit number that is zenzizenzizenzic, but I digress. The main thing to remember about my Pigskin Prognosticator is that it picks every game. Even the toss up games you would never touch in a knock out league. It also picks games against the spread. Picking just the favorites in the NFL will get you above 50%, it is way more difficult when you are trying to be smarter than Vegas.
Overall the prognosticator picked 137 out of 256 games correctly for a winning percent of 53.5%. Remember, that 137-119 record is against the spread. So if I would have auto bet all year with the pigskin prognosticator, invested my winning each week in bitcoin… I probably wouldn’t be writing this right now.
A few notes: As a fan I am all for the NFL “flexing” games. As a programmer it can really mess up timestamps. Plan ahead for overtime games. The PP got smarter all year as I gave it more data. Back in November I ran the schedule out and got:
for the finalists. If you figure a computer program ain’t gonna know Philly is cursed, the NFC is still rock solid. On the AFC side the Bills! Come on, they in the tournament for the first time since Jim Kelly! But of course it gives different results now. PP’s percentage was a full 6 points higher the second half of the season when it had half a season’s worth of data.
College football is much more volatile and vast but I played around with a subset of games.
One thing for sure – UCF is the best team in Florida, by allot. The PP gives Georgia a 51.7% chance at victory in the Rose Bowl this afternoon. It is predicting an all SEC championship as it is giving Alabama 52.75 % in the Sugar Bowl tonight.