2017 NFL Season Predictions

With the 2016 NFL season days from its close, its time to start looking into 2017.  Speculation is everywhere when it comes to sports betting, especially the American sport with the largest fan base.  You can’t flick on ESPN, click on a web link or hear a radio show without finding subjective guessing from the “prognosticators”.  Prognosticators, that if you ever research, are statistically as successful as a monkey throwing a dart at a board (see this article for data on how bad it truly is).

However, no matter how much the prognosticators tell us about their hunches, and gut feelings, there are certain clues we can get from objective empirical data to make more accurate predictions.  Collectively, along with your own intuition, you can put together a relatively strong argument for upcoming NFL team performances.  So whether you’re trying to impress the boys at the bar, hoping to lay some money on next year’s outcomes or just looking for some solid data, I’ve put this piece together for you.

There are 3 key areas we look at to statistically determine which teams are most likely to rise/fall from one year to another:

  • Net Close Wins/Losses (avid readers have seen this before in NHL NCW/Ls and NBA NCW/Ls)
  • Pythagorean Win Theorem (fully explained for those that don’t want to admit they have no idea what it actually means)
  • Net “Luck” from Turnovers gained (with a little Newmanomics twist)

These data sets help paint a picture of what should have happened “on paper” in 2016, and consequently, give us some insight into 2017.  Each one, independently, provides quality information, but I take it another step at the end and combine all three to give you a complete look at what the numbers are telling us.

Net Close Wins/Losses

Newmanomics’ bread and butter: Net Close Ws/Ls.  In this set of data, statistic’s “tendency towards the mean” helps us get an idea of who’s likely to come out with more overall wins and losses for next year, based on how well each team performed in close games – the ol’ “luck’s gonna run out theory”, in other words.  “Close Games” in this sample, are represented by any contest won/lost in which the difference in final score was 7 points or less.  The “net” part is simply taking the respective team’s sum of close games won (+) vs. close games lost (-).  Long story short (feel free to peruse the older blogs if you’re looking for the long story), with the parity in the NFL, and chaotic/sporadic nature of the game, luck plays a huge part in determining games won or lost, especially when the margin of victory is so small.  Think of it this way, the “better team” is less likely to win a close game relative to games won by larger margins, where one fumble recovery or bad call could influence the outcome.  Hence, it’s fair to say (and I have the data to support in previous blogs) teams that won the majority of their close games in 2016 (lucky) won’t have so much fortune in 2017, and vice-a-versa.  Here’s how the data shaped out:

2017 NFL Season Predictions

Teams like the Chargers, Jaguars, Eagles and Bears all (net) lost a large number of their games, and are likely to not be as “unlucky” in 2017.  And in the same light, Raiders, Texans and Dolphins, who “lucked out” in 2016 most likely won’t be so fortunate in 2017.

Pythagorean Win Theorem

We’ve all seen Moneyball, and whether you did or didn’t it’s unlikely you truly understand what the PWT means.  Truth be told, until recently I only knew what it represented, not how the formula got you there.  There are plenty of great places to get a tutorial on how it works (like this one), but here’s the general idea:  Similar to the NCw/l indicator, PWT uses basic math to interpret data using the logic that “the best team doesn’t always win” to predict how performance will change in an upcoming year.  Astute readers, probably already know there is a strong correlation between NCw/l and PWT, and you’re right, but the PWT digs a little deeper because it takes into account every game, and every point NOT just games won/lost by 7 points or less.

Honestly, PWT is really just a fancy-pants way of looking at one team, how many total points were scored in games they played (their points AND their opponents’ points) and what the ratio of those points are “owned” by said team to get an idea of how many games they should have won.  The only reason A squared + B squared = C squared has anything to do with it, is simply a known geometric relationship.  Think of that right triangle, A is points scored, B is points given up, and C is the total amount of points scored by either team in one team’s season.  So the Pythagorean Theorem shows us a proportion of “ownership” of points one team has, and how that should relate to win percentage.  Now using this data, looking forward, it can give us some quality insight (a step further than NCwls) on teams that put up more points than their opponent in the aggregate (over the season), yet just got unlucky on certain occasions, or individual samples, like game(s), to determine their “luck factor”.

Imagine this extreme example:  Team A goes 6-10 on the year, but every game they won, they did so by 20 points, and every game they lost was by 1 point.  This means they would have a Net Points scored for the year, of +110.  Whereas Team B goes 10-6, with all one point wins, and 20 point losses, meaning they would have a -110 NPS.  Team B is likely in the playoffs after an insanely lucky year, while Team A is looking at a top 10 draft pick.  The smart money would be on Team A moving up next year with how well they played relative to unlucky outcomes.

All in all, the PWT gives us another tool to learn about how teams’ record fared, relative to the net points they produced in a season.  Simply, the formula is

16 (Games in a season) x (1/(1+ Points Allowed/Points Scored Squared)) = Games (Should Have) Won

And here’s how 2016 shaped up:

2017 NFL Season Predictions

Browns, Jaguars and Raiders the furthest from the median…

Turnovers tell Turnaround, with a Newmanomics Twist

So our last clue, Turnovers.  The idea here, is turnovers, the most correlated statistic to W/Ls (besides points) are largely a matter of chance.  Essentially, it’s thought that because of turnover’s “lucky” nature, a team with a large positive net TO ratio, won’t be so lucky next year/won’t be able to rely on the increase in odds of a victory that (+) turnovers bring them (and vice-a-versa).  +Phil Steele, who I’ve mentioned on this blog many times, a fantastic handicapper, originally opened my eyes to this concept. However, I have a minor objection with his metric.  Turnover ratio is comprised of the net interceptions + fumbles that led to a change in possession.  If the spirit of the metric is to use “luck-based-indicators” to determine change in fortune, I’m not AS BIG a believer that turnovers (as a whole, Phil’s metric) has a much to do with luck AS DO ONLY fumbles (Newmanomics Metric).  Hear me out:  for one, interceptions can largely be prevented by the skill of a quality QB relative to fumbles.  Yes, some passes picked off are the result of tipped or ricocheted balls, but relative to fumbles the luck factor isn’t close.

First of all, Team Fumbles RECOVERED/LOST has contributors from the entire team.  And the more pieces in place, the more independently thinking/moving human bodies – the more chaos – the more luck.  Second, and more importantly, a turned-over ball due to a pass, comes from a QB INTENTIONALLY moving the ball forward/displacing from his possession, and hence has more control over its landing.  However, when it comes to fumbles, for one, the player losing the ball didn’t intend on it, and for two, when the ball hits the ground (can’t happen with a pass/interception) no one can predict where it goes – often times leading to a “coin-flip” situation for possession of the ball.  So, what all this means to me, is using Net Fumbles Gained vs. Lost, albeit a smaller sample set, gives the prognosticators a better indication of which teams got lucky in the year past – and then more likely to improve next year.  So, without further ado, something you can’t find anywhere else, Net Fumbles Gained/Lost:

2017 NFL Season Predictions

Bears, Browns and Jags the big “winners”, Steelers and Raiders the big losers…might be seeing a trend now, eh?

Each clue, gives us some really good info to work with.  However, what I’ve done is take all 3 of these data points in aggregate to give you the best indication of which teams are rising vs. falling next season.  What I did was simply rank each team in all 3 metrics, where they fall and then took the average among them, to give you: The. Final. Answer:

2017 NFL Season Predictions

SO IF YOU MADE IT THIS FAR, here’s where my money is going next year…

And yes, you’re probably thinking to yourself, isn’t this is just simply an “upside down look” at the 2016 records, and the NFL’s natural tendency towards each team being 8-8?  Yes, that’s true, but there are some teams here that combining subjective intuition can lead to some quality answers:

-(Homer Alert) The Chicago Bears finally have put together what seems to be back-to-back above-average drafts.  Finding a RB, WR, C, DE, ILB, S, and NT to combine with some key free agents picks in the last couple years at LG, ILBs, OLB, but have simply been hammered by injuries and poor QB play makes me think they’re trending up.  The injuries should regress towards the mean, and if the Bears can find a viable QB (tall order, I know), and/or get Jay to stay healthy/play the way he did in 2015, combined with what the numbers tell us,  they will be a playoff team.

-The Chargers were painfully unlucky last year.  With having a veteran QB and budding star at RB (given his health), if they can find some help on the other side of the ball, the Bolts should find some success in their new town.

-The Raiders on the other hand do have a very talented team, but were extremely lucky, and are in a very tough division…I don’t see them in the playoffs next year.

-This may be a little bold, but I see regression from the Cowboys too.  On top of what the numbers say, they had a very lax schedule that benefited their young core.  Additionally, the “sophomore slump bug” has two key players it could nip.

After all is said and done, sports predictions, particularly football is a crap-shoot.  Anyone that tells you different has an agenda.  These metrics are simply to be used as a guide to help.  With that being said, I hope you enjoyed, and good luck!

Only 216 days until the start of the 2017 season!

Leave a Reply