As always, here’s a quick primer on some of the stats and data I’ll be using this week (feel free to skip this part if you’ve read this column before):
Expected points added (or EPA)
EPA is a metric showing how many expected points a team added on a given play.
Here’s the short version: it captures the true value of a play by accounting for the context in which it occurred.
The longer version?
EPA is calculated by first determining how likely the next score of the game is — a touchdown, field goal or safety — based on a variety of factors that include field position, down, yards-to-go, score differential and time remaining in the game.
More precisely, if a team is first-and-1 at the opposing team’s goal line, the model will predict an expected points value near 7; it’s very likely that the team will score a touchdown. So if a player in that situation then runs the ball in for a touchdown, the expected points added (EPA) won’t be very large.
On the other hand, if a team is at fourth-and-19 at their own 20-yard line, there could be negative expected points in that situation; it’s more likely the other team will score next. So if a quarterback throws a deep bomb for a touchdown, that play will have a very high EPA value.
A play is a success if it had a positive EPA value. So if a team has a 60% success rate, that means 60% of their plays had a positive EPA. A run of two yards on first-and-10 is often not a success; a team tends to be less likely to score after such a play.
Average depth of target
This metric quantifies how far a quarterback’s average pass is thrown down the field. It is measured vertically (straight north/south) from the line of scrimmage to where the receiver catches (or doesn’t catch) the ball. Obvious throwaways are recorded as 0 air yards, to prevent a QB from gaining credit for just throwing the ball out of bounds 20 yards downfield. This metric is a useful way to quantify and compare the gunslingers against the check-down artists.
Win probability model
There’s something very important to note about the win probability model I will be using for this series. It does not take into account the strength of either team in the match-up. Both teams begin the game with a 50% to win. All the model is doing is looking back over the past decade to say historically, given the current score, possession, time remaining, home-field, etc., how often has this team ended up winning. If you want to see an actual look at how likely the Chiefs are to win the game, just check the vegas odds before the game and at halftime.
So, if this model does not perfectly predict who will win, what good is it? Well, it’s an excellent way to contextualize just how important a play was. We can say that before Mahomes did _________, the Chiefs had a 30% chance of winning, and after, they had a 70% chance. Thus, we can say that the amazing thing Mahomes just did added 40% to the Chiefs win probability — also known as Win Probability Added. This makes our model very useful because it treats all teams equally and thus lets us compare plays across games to see which had the biggest influence on the game.
With that out of the way, let’s get to analyzing Thursday’s game:
Chiefs vs. Denver Broncos, October 17
Ahh... back to the familiar this-game-was-over-well-before-it-was-over feel that Chiefs fans experienced in the first few weeks of the season.
There is a string of three plays here — all within a few minutes of game time — that are worth noticing.
With about 10 minutes left in the second quarter, the Chiefs were deep in Broncos territory and needed to convert a fourth-and-1 at the five to have a chance to add to their 10-6 lead.
As a stats guy, I was (of course) thrilled when I saw that head coach Andy Reid had decided to go for it instead of settling for a field goal.
There have been countless papers, blog posts and tweets dedicated to the fact that NFL coaches are far too conservative on fourth down, missing out on as much as an entire win per season by doing so. In fact, the authors of my favorite paper on the subject now both work in the league. One works for the Baltimore Ravens and the other heads of the NFL’s statistics team.
The benefit in going for it on fourth-and-short is particularly pronounced when a team has a high-powered offense — as the Chiefs do.
As anyone who watched the game will likely never forget, the play Reid chose was a quarterback sneak. Again, I was thrilled.
On fourth-and-short, quarterback sneaks have the highest success rate of any play-call. If you don’t trust the analytics, you should at least trust New England Patriots head coach Bill Belichick, who calls more quarterback sneaks with the face of his franchise (and arguably the NFL) than any coach in the league. In the short term, the on-field result corroborated the probabilities: the play worked. The Chiefs converted the fourth down with at least a yard to spare, adding over 10% to the Chiefs’ chance to win the game.
However, most Chiefs fans would certainly not consider this play to have increased the team’s chance of winning. Patrick Mahomes was injured on the play — and is now likely to be out for at least two weeks. A few plays later, his backup Matt Moore missed Damien Williams on a third-and-1 play, forcing the the Chiefs to kick a field goal — thereby eschewing the advantage they had gained with Mahomes’ sneak. With the field goal, the team’s chance of winning went right back to where it was before the sneak — essentially wasting the play that injured the reigning MVP.
Chiefs fans know this story. An important player gets injured, the offense fails to score a touchdown, the other team gets the ball back and begins to slowly dismantle any lead the Chiefs had built — and any hope for a Chiefs win.
But this time... something different happened. Just two plays later, Anthony Hitchens sacked Joe Flacco, knocking the football lose for Reggie Ragland to recover it. Stride untouched, Ragland ran into the end zone. For the Chiefs, that play marked the largest boost in win probability in the game: an increase of almost 20%.
As you can see from the chart, the game was over after that play. The Broncos arguably tried their best — but playing with a big lead, the Chiefs were able to relentlessly attack Flacco and the Broncos offensive line, picking up a total of nine sacks on the night.
We’ll come back to that.
Patrick Mahomes — and Matt Moore
Up to this point in the season, this column — and the analyses and code that build it — have been fairly straightforward. We look at how Patrick Mahomes performed, his receivers, the running game and so on.
But now there are two Chiefs quarterbacks to analyze. Expect another in-depth article detailing what the advanced metrics tell us we can expect from Moore, but for now, we’ll just plug him in to the standard analysis framework we’ve been using.
Mahomes against Denver
Prior to his injury, Mahomes was back to the efficiency the Chiefs have been accustomed to seeing. As I (and many others) have noted, Reid drew up more short (but effective) passes to compensate for the injured offensive line and Mahomes’ lingering ankle injury — which was reflected in Mahomes’ depth of targets against the Broncos.
This game plan resulted in Mahomes’ best performance since Week 3 against the Ravens — and one of the best performances of the week.
QB Performances - Week 7
|QB Hit |
|QB Sack |
|G. Minshew II||0.03||6.00||44%||6%||6%|
If you’ve been following along with this column, you’ll probably notice that there were some really high EPA per play numbers for quarterbacks in Week 7 — and really low ones, too. In fact, this is one of the wider distributions we’ve seen this season; as you can see here, the good performances were better than we normally see, and the bad performances were worse.
Buoyed by the iridescent performance of Aaron Rodgers and the... "performance" by Sam Darnold, Week 7 featured the widest distribution of QB EPA/play of any week so far this season. pic.twitter.com/8nPt1OcR4x— ArrowheadAnalytics (@ChiefsAnalytics) October 22, 2019
Mahomes was on the right side of this spectrum — he had a top-10 quarterback performance as evaluated by EPA per play.
But Matt Moore? Not so much.
Moore against Denver
Here’s how Moore’s depth of targets compared to the Chiefs’ average through Week 7:
Despite throwing shorter than usual when he first came into the game, Moore eventually started targeting deeper down the field. He threw fewer passes over 10 yards than the Chiefs’ average, but that isn’t a bad depth distribution for a guy who supposedly hadn’t taken a single first-string rep.
The EPA, however, was not great. As you can see in the table above, his performance ranked him fifth-worst among quarterbacks in Week 7 — just above Sam Darnold, Matt Ryan, Flacco and Daniel Jones. That’s not exactly where the Chiefs are used to seeing their quarterback.
Once again, the Chiefs gained a majority of their receiving yards after the catch. This week, Mecole Hardman was the most efficient receiver on the team, with a sizable gap between him and the rest of the team’s receivers. Tyreek Hill was also more efficient than average, while the rest of the receivers were just about average. But the two Williamses — Darrel and Damien — were horribly inefficient.
Chiefs rushing attack
The Chiefs finally ran the ball enough we could see this chart again. Oddly, not a single run outside of the middle was successful for Kansas City on Thursday.
Another interesting note on the running game: Anthony Sherman was 100% successful on his runs — meaning that both of his two attempts put the Chiefs in a better position to score than before. With seven attempts, he is now at 100% on the season, compared to a 35% rate on 14 attempts last season. In other words, Sherman could gain no yards at all on seven more carries and still be as effective he was in 2018.
The Chiefs play calling was a bit different this game; with Moore in the game — and with a big lead — the Chiefs ran a lot. Even controlling for a neutral game script, the Chiefs only passed the ball 57% of the time. That’s their lowest passing percentage of the season — far lower than the past two weeks when they were above 80%.
Other than the Mahomes injury, the dominant storyline from this game was the resurgence — or rather the resurrection — of the Chiefs defense. As noted above, they responded to Mahomes’ injury in a huge way. It’s likely no surprise that this was the defense’s best performance of the season. But just how good was the defense by our advanced metrics?
Of every single game in our play-by-play database — that includes every NFL game since 2009, both regular and post-season — this was the eighth-best performance a Chiefs defense has produced, limiting the Broncos offense to an astounding -0.35 expected points added per play. And only one of those seven better Chiefs defensive performances occurred after 2013: the Chiefs 30-0 stomping of the Houston Texans in the 2015 wild card game. Here’s how that performance looks on a chart compared to all games for all teams over the past three years (bottom left is better, but lower in the Y axis is more important than left on the X axis).
That’s a damn good defensive performance.
That’s it for Week 7 of Stacking the Box Score.
Leave a comment or reach out on Twitter (@ChiefsAnalytics) if you have questions or would like to see something new for next week. Thanks!