NFL Betting Project


Indicative 2017 first week picks.
These are based on information as of February 2017, and will be updated at the end of August 2017. Picks may change.

Please do not use these picks without reading the information below. They are produced for demonstration purposes, and do not necessarily represent my best opinion of the games. I do not use these for betting myself.
NFL betting project

Since 2006 I've been posting NFL picks based on a simple quantitative system that has delivered 522 wins versus 414 losses in 11 seasons, for a 55.8% winning percentage. What it will do in the future is anyone's guess.

In 2016, the system produced a disappointing 43 wins versus 39 losses (52.4%), for a bare 0.1 unit profit to a better paying full 10% juice. In its worst year (2007) it had 37 wins versus 46 losses (44.6%), representing a 13.6 unit loss to the full juice bettor.

The system uses only five simple indicators, and has not been adjusted in any way since it was designed before the 2006 season, using NFL data from the 1983 to 2005 seasons (odd numbered years to design, even numbered years for holdout.
Click here for documentation on the system
Record to date

Back in March 2006, my goal for the system was to have a 55% win rate for three years, without any adjustment. 2006 came in just shy of target, and 2007 was a disaster. After three years, the system was only at 52.5%, so I failed in my original goal.

Then things turned around, and the system delivered seven of the next eight years above 55%, including spectacular results in 2009 and 2014.

2016 was disappointing and there are other reasons to believe the system's age is showing. Nevertheless, as long as it has a cumulative track record above 55%, I'm inclined to keep running it.
  1. Are those real betting picks for NFL games?
    Yes, they're real picks, based on a simple quantitative system I developed in 2006 and have been posting for all NFL games since the first week of the 2006 season. It's not the best prediction system in the world, it's not even the best prediction system I ever designed. It's the easiest to develop one I could come up with that I expected to deliver a 55% win rate in live betting, without maintenance for three years. In fact, it's been running eleven years at a 55.8% clip. That's mostly luck, it's not that good. It didn't make 55% in its first three years, only 52.5%, just slightly above break even for a better paying full 10% juice. That's mostly bad luck, it's not that bad.
  2. Why are you doing this?
    The point of this demonstration is to disprove two common myths. They contradict each other, but a surprising number of people still manage to subscribe to both. The first is that it's easy to make winning bets—or financial trades, or to predict the effect of an increase in the minimum wage or a surge in Afghanistan or anything else—using intuition, expertise, commonsense or rules of thumb. That's not true. Those things are hard, and almost everyone who tries using those tools fails. The second is that it takes some mysterious and rare ability to beat the bookie; or the house or the market or City Hall. In fact it's quite simple, using quantitative analysis anyone can do.
  3. So why isn't everyone rich from sports betting or financial trading? And if the tools are so great, why do people make huge bets on national policy, life savings, the future of the Earth and everything else without using them?
    The hard part about getting rich from sports betting is not finding a system with a 55% win rate. It's getting people to give you money. Retail betting sites will quickly notice that you're making smart picks. The good ones will cancel your account if you're a net winner, the bad ones will keep your money. There are other venues for placing bets, like matching sites, casinos, human bookies and others; but none are anxious to give large amounts of money away; all of them have defenses you much consider. If your goal is to make a living on sports betting, you're going to need to work. Simple does not mean easy. You're going to want at least a 57% win rate, and that's maybe ten times the work of building a 55% system. And you're going to want several of them so you can make a lot more bets throughout the year. You're also going to need to manage your capital and risk. The reason most people aren't rich from sports betting, and have such unjustified confidence in unsupported beliefs, is that they aren't willing to do the work, or to admit how slight an edge we have when predicting the future.
  4. Enough chatter. What's the system?
    You should be asking how I built it, but I'll postpone that part since you're in a hurry. For each team in each game, the system gives +1 point if the team lost the ball on turnovers more than it got the ball on turnovers in its last game, -1 if it lost the ball less than the other team and 0 if it had the turnovers were equal. Next it gives +1 point if the team has covered the spread fewer times than it has failed in the season to date, -1 point if the it has covered more times than it has failed and 0 if the numbers are equal. Finally, +1 point is awarded to the team the line has moved against since the previous week. Subtract team A's points from team B's points. If you get 3 or higher, bet on team A. If you get -3 or lower, bet on team B. If you get a number from -2 to +2, don't bet. Download the spreadsheet using the button above to see all the calculations laid out.
  5. You're kidding, right? That's the secret of the universe? It can't be that simple to beat the spread.
    It's easy to find factors like the ones above that give you about a 52% edge, there are dozens of well-known ones (all the ones I used are well-known) and you can find your own. Some have worked for decades, some come and go, I picked some of the longer-lived ones because I wanted the system to run without maintenance for three years. These edges can persist because each one on its on is not enough to overcome full juice for the retail bettor. But if you combine five, you can get to a 55% edge. The turnover indicator relies on the fact that turnovers are mostly random, so teams with net takeaways in a game are overrated and teams with net giveaways are underrated. The cover indicator is based on bookies liking all their customers to lose a little; not so much that they quit or cause trouble; but not to win. Since lots of people bet the same team every week, the system nudges the lines to even things up. The line move indicator is based on over-reaction. People overinterpret new information and move lines more than they should.
  6. So how do I read the picks?
    The table at the top of the page lists next week's games. "Line" is for the home team, so -7 means the home team is favored. For both teams it shows "LGT" (last game turnovers) which is the team's giveaways minus takeaways in its last game. "STDC" (season to date cover) is the number of times the team has covered the spread so far in the season, minus the number of times it has failed. "Power" is the team's power based on information before last week's games, the difference in power is what I think the spread would have been had the teams played last week at a neutral site. In the last two columns, "#" is the home team's points minus the away team's points. If it's 3 or more, bet the home team. If it's -3 or less, bet the visitor. If you prefer do-it-yourself, download the spreadsheet with all the data and formulae. Make it better.
  7. Okay, it worked in the past. How do I know it will work in the future?
    You don't. That's why it's called gambling. If you scroll to the top of the page and look on the right, you can see it has good years and bad years. If you're smart, you'll also scroll to the bottom, where I show two diagnostic tables. Professionals always keep validation measures. Otherwise the only way you can tell if a system has stopped working is by losing money, and that's expensive. You're going to have to click on the validation tables to download explanations, but the quick message is they've been showing some strain since 2010 (recall that the system was designed for a useful life of three years). For one example, in the 2016 season, when teams played each other twice, the team the line moved against won 24 and lost 32. The over-reaction indicator thinks it should be the opposite, and in fact from 2006 to 2015, the record was 298 versus 243. So maybe that indicator is failing. Keep that in mind if you choose to bet on the picks.
  8. You expect us to believe you're the world's greatest sports bettor?
    Of course not. I'm not even a professional sports bettor. I was at one time, and was only average at it among quants (but near the top overall). I know people who are worlds better than me. However, I'm better at explaining things than most of them, and I have no interest in pretending there's some mysterious secret. I've kept up on the state of the art by funding and mentoring young betting quants. You teach them for a year, then they teach you for around three years, then they go off either to build a betting business or apply their skills in another domain, or otherwise decide they don't need you (that's good, that's how life should work, but it would be nice not to have to keep finding new people to do the work).
  9. So if your picks are so great, how come you don't bet with them? Why not show us the system you do use?
    I bet through professionals I back. They're much better than this system. I am not at liberty to disclose their methods. But I will say that the differences are of degree and not kind. This system has five factors, professionals might use 25, and track another 100 for possible inclusion. This system only does NFL line bets, professionals cover a wider variety of sports and bet types for diversification and to maintain scale. I used only lines, scores and turnovers; professionals use huge amounts of data. I did only some basic validation, professionals do more work on validation than on model-building. It really is a lot of work. If you're the kind of person (like me) who enjoys that work, it can be a good life, although finance is way easier than sports betting for the same dollars. If you're not, you could force yourself to do it, but it's no fun, and you probably won't be able to do it well enough to compete with the best.
Factor validation table, click for explanation
Marginal validation table, click for explanation