The Pros and Cons of Going Long

This is the first post in a series of posts about The Pros and Cons of Going Long. This blog will discuss the merits and pitfalls of trying to predict NFL form long-term.

As a quick preface, I’d like to direct you to an article published by my colleague, Michael Lopez, on the subject of NFL win rates over a single season. His sample includes every regular season game played between 2003 and 2013.

The most salient point from his analysis was that the league was quite balanced over this time frame: there were no teams which won more than 65% of their games during any one season (a .650 winning percentage corresponds roughly to an expected record of 11-5) and no teams which won less than 35% of their games (.350 is 3-13).

As a numbers-driven industry, NFL gambling is constantly searching for new ways to beat the game. One of the most interesting and potentially fruitful areas of study is in predicting long-term team performance. For example, can we use preseason records, offseason transactions, coaching changes, or draft picks to help us determine how well teams will do the following year?

There are two main problems with this approach. First, there are many variables at play. Teams go through myriad changes over the offseason. Some teams may have an incredible draft but lose their top receiver to injury or suspension. Others may have a mediocre draft but get a string of injuries to their opponents and ride that wave all season. Some teams have good records in preseason but are full of backups and 3rd stringers trying to make the squad. Others have bad records in preseason but play their starters for extended minutes and get their timing down for the regular season. It’s very hard to predict which of these factors will win out and which ones will be relegated to background noise by the end of September.

Second, we don’t have any good way to measure long-term performance. The closest thing would be preseason win/loss predictions by oddsmakers or pundits in August, but those are notoriously inaccurate (as

There are two types of NFL betters. One type is the most common: bettors who wager on individual games. They watch all the games and look for edges in the matchups, finding trends and statistics that might point them toward a game they believe will result in a win.

Another type of better watches the games and makes long-term predictions about which teams will be good or bad for the entire season (or multiple seasons). These betters often start their work before the season even starts, trying to get an early edge on what might happen over a 16-game slate.

What if you could predict the future? What if you had a crystal ball that was all-knowing and all-seeing? What decisions would you make? Would you bet on every team to win their next game or would you spread your bets out across a season? Well, what if I told you, you don’t need a crystal ball to do this.

In this blog we will attempt to predict the results of NFL games over the course of the 2016 season with no knowledge of form or injuries. The methodology for doing this is simple. Look at how many points each team scored last year and estimate how many points they will concede.

This method has been used in soccer for about 5 years now and it has proved pretty successful. Obviously, it won’t be accurate every time but over a larger sample size (like an entire season), it should prove to be pretty successful.

At the beginning of each NFL season, I submit a number of predictions to a number of publications. These predictions aren’t sexy, but they’re not boring either. I predict how many games each team will win and lose.

I also have a prediction for how this set of predictions will fare. This one is boring and not sexy at all: it’s very likely that my predictions will be wrong.

How likely? Over the past four years, my prediction accuracy has fallen between 47% to 53%. At the end of the 2017 season, my predictions were roughly 51% accurate, which was close to average over the previous four seasons.

So why do I keep doing this? Because on average, I’m doing better than you are. And because every once in a while, I do something right and get some recognition for it.

It’s amazing how much conventional wisdom gets tossed around regarding the NFL. In particular, I find it interesting how many people are willing to make definitive statements about the NFL that contradict each other. The following are a few of my favorite “always” or “never” statements:

* Always take home teams coming off a bye week.

* Always take underdogs on Thursday night.

* Never take road teams coming off their bye week.

* Teams never win back-to-back games on the road in the NFL.

What makes these statements interesting is that they were all true at one point, but in today’s NFL, the only one of them which still holds true is the first one. It turns out that there are very few always/never statements that work in today’s NFL and still fewer which work over an extended period of time.

The NFL season is a marathon, not a sprint. The marathon started on September 5th when the Kansas City Chiefs beat the New England Patriots 42-27, and it will end on February 3rd with Super Bowl LII. In between, there are 256 regular season games and 10 postseason games. There will be hot streaks and cold streaks; injuries, trades, and free agency; coaches being fired and new ones hired. It’s a lot of football to take in and make sense of.

This is where analytics can help. Analytics can help you identify which teams are actually good and which ones are due for a regression to the mean; identify which players are good bets for regression and which ones are worth trading for or picking up in free agency; identify potential sleepers on both sides of the ball; even identify coaching changes that could impact teams either positively or negatively. Analytics can give you an edge over other fans and fantasy owners who may not be paying as much attention to the details or who may be blinded by traditional narratives.

Analytics, however, can only help you so much. Consider Week 1&

Similar Posts

Leave a Reply

Your email address will not be published.