Upset Watch: Week 17
29 Dec, 2017 by Shaun Lowrie
I am at heart, a statistics geek, and like many geeks, I have certain things that I hang my hat on more than others. My own NFL fetish (careful...) is for what I have decided to call (primarily because I have only gone and created a situation where I need a snappy name for it) 'situational precursors'.
I'm a big believer that the NFL follows statistical patterns in situations that are common every year. Forget about the teams involved for a moment, think about the situation they find themselves in. Never is this easier for me to convey than in week 17, where teams find themselves invariably in one of three types of game:
Games that seriously matter
Games that matter only to the favorite
There are some small variations (a favorite might - like the Vikings this week - have a very small incentive to win, but is unlikely to risk injuries as if it was a 'must win'). Equally, a team like the Browns may consider not going 0-16 against Pittsburgh this week a big incentive to play hard.
But for the most part, even these situations have precedents in the NFL, and this week, I've gone to the trouble of taking a sample of this data from 2015 and 2016, and using it to guide us through the minefield of this year's round of oddities that is week 17. But first, here's last week's results:
Last week's results:
Top 5 + Bottom 5
Last week wasn't great, going 6-10 vs the spread, our worst score of the season. Why? Well, primarily, favorites won, but didn't cover. There were just 3 upsets (of which we had one), but two pushes (Philly and the Chargers) and a host of teams not covering as favorites. Thankfully, we were still almost even overall and lost just 10 units to leave us in a strong position over the course of the season...
Top 5 + Bottom 5
Yes, it's fair to say we're still in 'resounding success' territory. Let's not get carried away, there's still work to do, but +54 means we'd need to go 2-14 in week 17 to finish with a loss this season. Frankly, I'd like to think that by simply allowing my cat to pick games, I could avoid that, but hopefully with the below working out, you'll see that we're aiming much higher...
On with the show:
Week 17 is slightly different, and as we explained above, rather than the usual 'upsets vs non-upsets', we have three categories this week.
Outright Upsets - Games where we think the underdog will win
Underdog Covers - Games where we think the favorite should win, but the underdog will cover the spread
Favorite Covers - Games where the favorite should win and cover the spread
These three categories were developed by looking at the last two seasons of week 17 games and the situations faced by each team. The factors used are:
Was there an upset?
Did the favorite cover?
The size of the points spread
Was the favorite playing for something?
Was the underdog playing for something?
Was the game a meaningless dead rubber?
Did the favorites play their starters?
Did the underdogs play their starters?
After doing this for previous years, we also wanted to make our best estimate of how these situations will apply on Sunday this year. Of course, we know that some teams will not give out information about starters at this stage, but for the most part, we can make a reasonable estimate of whether they'll do so, and whether they're likely to play the full game or not.
Here are some key stats to bear in mind as you read the column:
The average spread over the last two seasons in week 17 is 5.5pts
There have been 15 upsets in the last 3 years, 9 of which came last season. There were 4 in 2014 and just 2 in 2015
Favorites are 20-27-1 vs the spread over the last three seasons
19 of the games over the last two seasons (32 games) had something on the line for at least one team regarding a playoff place
The other 13 games were dead rubbers
We'll explain the statistical reasoning for each game, so let's start with the upsets...
1 - Dallas @ Philadelphia +3
2 - Washington @ New York Giants +3
The key factors: Spread size under 5.5, dead rubber, Underdog playing starters
These are all the games where upsets have occurred in the last two seasons where the spread has been under the average of 5.5. For reference, the reason we wanted to identify these games was simple - there have been just 3 upsets over the last two years where the spread has been greater than that average, and in fact the other 8 upsets were all in games with a spread of 4 points or less. We want to focus on high statistical probability, rather than those big spread games - and there are many of them this year that I'll come to later.
As you can see, one of the biggest factors is that 6 of 8 upsets have also been dead rubbers. Of those 6 dead rubbers, 5 also featured an underdog playing their starters. Those 3 factors are common in just two games this season, both home underdogs in Philadelphia and the New York Giants.
The Giants were on my list anyway, with a number of jobs on the line heading into the final week of the season. The Redskins have been frankly, a walking casualty unit, so the idea of the home Giants winning is not exactly crazy.
The Eagles are a bit of a strange one in that they have stated the desire to play their starters, and they have a good reason to. With Carson Wentz injured, they want Nick Foles to get as much playing time as possible with the players he will be leading into the playoffs as no.1 seed in the NFC. I think they'll be smart and pull the likes of Foles, Ajayi and Jeffery in the second half, but I like their chances of beating down a Dallas team that has seriously struggled and is now done for the year.
The Underdog Covers
New York Jets (+15.5) @ New England
Chicago (+11.5) @ Minnesota
Cleveland (+10.5) @ Pittsburgh
Cincinnati (+9.5) @ Baltimore
Oakland (+8) @ LA Chargers
Tampa Bay (+7) @ New Orleans
The key factors: Spread size over 5.5, underdogs playing starters, not dead rubbers
Why so many underdogs covering? Well first of all, this will still only take us to 8 underdogs covering, so an expectation of 50% of underdogs covering their spread is fine.
This means we need to isolate the games this year that have big spreads and look for the common factor. Let's see if you can spot it...
It should immediately jump out that all of these games are meaningful for the favorite, but not the underdog. This is not a coincidence. Of course the favorites are likely to be the teams with the most to play for who are in or around the playoffs.
What is most striking, however, is how rarely this has translated to covering the spread over the least two years:
No, your eyes do not deceive you. Only twice out of a relatively large sample of 12 games, have the favorites covered spreads of -5.5 or above, and in one of those games, there was nothing to play for (Cards @ Rams 2016).
That means that effectively, when underdogs played their starters, just 1 of the 11 meaningful games over the average spread of 5.5 was covered by the favorite over the last two years in week 17.
So all of the above games fall into that exact criteria. Why does this happen? Well, put simply, teams have a tendency to play 'easy' after they establish a lead against inferior opposition. When the spread is so high, this often lets teams back to within a touchdown difference by the end of the game.
With 7 games falling into this criteria, we are going to take Favorite win/Underdog cover in all of them. In particular, watch the Steelers, as favorites have covered just twice in the last 2 years when resting starters. Disrespect is the ultimate incentive, even for a team like the Browns who are disrespected every week.
Favorites Winning and Covering
There are three parts to this section:
The key factors: Dead rubbers, spreads below 5.5, underdogs not playing starters,
Green Bay @ Detroit (-6.5)
Houston @ Indianapolis (-4.5)
Kansas City @ Denver (-3.5)
San Francisco (-3.5) @ LA Rams
The key factors: Spreads below 5.5, underdogs with nothing to play for, favorite with something to play for
Buffalo (-2.5) @ Miami
Jacksonville @ Tennessee (-3)
The key factors: Spreads over 5.5, favorite with something to play for, underdogs resting starters
Arizona @ Seattle (-8.5)
There are some wildly disparate games on this slate, which is why we need to break it down a little further into the two sections. The Titans have plenty to play for, whereas the 49ers are completely dead and buried as far as playoff purposes, but we expect the outcome to be the same, statistically speaking.
In the games of section 1, you're going to see largely unwatchable football, but you'll likely see the Colts, Lions, Broncos and 49ers play their starters against backups for the majority of the game. Sure, the Niners and Lions may pull Jimmy G and Matt Stafford respectively at some point, but they are both likely to come out with their full compliment of starters. Meanwhile, their opponents are all starting backups or have been depleted by injuries and will field weakened teams. The Texans will miss DeAndre Hopkins and the Packers have lost Aaron Rodgers to injuries, while the Chiefs and Rams have both stated that they will go with backup QBs. The Rams have crucially also said that Todd Gurley and Aaron Donald will not play, which Jimmy G will likely be thankful for...
In Section 2 we have the perfect storm in Miami as the Bills needing a win head into town to play the Dolphins with nothing at all happening. Nothing. Just a diabolical football team. The Bills are just 2.5 point favorites which is a godsend of a spread considering they can still make the playoffs with a win, and nobody in Miami will likely want to tackle Shady McCoy with a nice and relaxing offseason ahead of them.
In Tennessee, the Titans must win to secure a playoff place too, while the Jags are in a strange position. They have almost no healthy wide receivers left, and will surely look to avoid a catastrophic injury to Blake Bortles or Leonard Fournette in a dead rubber, right? I mean, they were watchng when Connor Cook played for the Raiders last season, right? You don't want to break a long playoff-less streak and then find yourself lining up with Chad Henne under center.
The Titans have been bad, but I like their chances against a team that has not actually been very good for a few weeks.
Finally, that Seattle game. Guess how many teams win as big underdogs when they start their backups?
In what will come as a shock to literally nobody, well... literally nobody with nothing to play for wins when they play a meaningful game with only their backups against a motivated set of starters. Who. Would. Have. Thought.
Yup, the Cardinals have said they'll start Drew Stanton, they've already lost Adrian Peterson, David Johnson, Carson Palmer... blah blah... this is a team of backups and to compound matters, Patrick Peterson is also questionable to play on Sunday. Regardless, the Seahawks are going to have every incentive to go wild against the Cardinals, knowing they can still make the playoffs if Atlanta loses.
Speaking of which...
The one where both teams have something to play for
Carolina @ Atlanta (-4)
I'm taking the Falcons because they really do need to win, whereas the Panthers? They'd like to win, they'd probably quite like to send the Falcons home too, but are they going to send Cam Newton on designed QB runs at the goal line? Are they going to pound Christian McCaffrey and Jonathon Stewart? Are they going to risk Greg Olden's perma-injury again?
No, the Panthers are in the playoffs, and while they could theoretically still get a first round bye, the Falcons will likely prompt wholesale hooking of starters by Carolina if they go up by two scores.
Shaun Lowrie is the founder and editor of Pickwatch, a site designed to give you a new perspective on experts in the NFL media by tracking their performance when picking games or predicting draft picks, team performance, or player potential. You can contact him at [email protected] or on twitter at @pickwatch