/cdn.vox-cdn.com/uploads/chorus_image/image/65864168/1187385770.jpg.0.jpg)
Welcome to this week’s Stacking the Box Score, where I’ll be quantifying what went down between the Kansas City Chiefs and their most recent opponent — in this case, Sunday’s 23-16 win over the New England Patriots.
As always, here’s a quick primer on some of the stats and data I’ll be using. If you already feel comfortable with these metrics, skip the link and keep reading!
Chiefs vs. New England Patriots, December 9th
Win probability
:no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/19482297/wp.png)
While digging through the win probability data from the Patriots game, what jumped out most to me was near the end of the game, the model was confident the Chiefs would win the game. Trailing by seven, Tom Brady led the Patriots down the field, scrambling for a first down at the Chiefs 12-yard line. But despite being just a dozen yards away from tying the game, the model only gave the Patriots a 32% chance of victory. That proved to be accurate; it was the last first down the Patriots would earn.
Another interesting thing is how low the Travis Kelce’s fumble ranked. Called a “huge swing in momentum,” it was only the eight-most impactful play of the game behind the plays noted on the chart — including Brady’s interception and the Patriots’ flea-flicker touchdown.
Analysis
The Chiefs defense made a statement this week.
Despite giving up a passing touchdown after just a few plays in the opening drive, they particularly excelled against the pass, limiting Brady to an EPA per passing play of negative 0.40 — meaning that on average, when the Patriots’ quarterback dropped back to pass (whether the play resulted in a scramble, a sack, a penalty or an actual pass attempt) he made his team likely to score 0.4 fewer points. With 41 drop backs recorded, this adds up to the Chiefs defense limiting Brady to a total 16.6 fewer expected points added than the average offense should be expected to score in similar situations (down, distance, etc.).
:no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/19445071/brady.png)
Make no mistake: this defensive performance was historic for a team playing the Patriots. In games against Brady since 2009 (when our data set begins), this was the second-best defensive performance by total EPA, the third-best by EPA per drop back and a top-10 performance by Completion Percentage Over Expectation (CPOE).
The Chiefs pass defense is beginning to look quite good. Football Outsiders now rates the Chiefs as the sixth-best pass defense through their Defense-adjusted Value Over Average (DVOA) metric. But can we use our advanced stats — and perhaps some Next Gen Stats passing charts — to break down their performance even further?
Thanks to a former Carnegie Mellon statistics student (now employed by the Baltimore Ravens), we can.
Sara Mallepalle and her team at CMU created a tool that not only allows us to scrape data from the Next Gen passing charts, but also gives us a statistical model (in nerd-speak, a generalized additive model) with which we can estimate how likely a pass is to be completed — given its location on the field.
The methods behind this tool and model are open sourced — and if I’ve piqued your interest I highly suggest you check it out.
But I’ve taken this tool a step further, allowing us to use it to measure defensive performance.
:no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/19445185/nfl_cpoe.png)
Here we see the 2019 completion percentage of all NFL teams at all points of the field. The blue areas are where the completion percentage is 50% or lower; the orange areas are where it is higher. This makes sense, right? Passes closer to the line of scrimmage are easier to complete.
:no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/19445188/kc_cpoe.png)
Here is the same information — but only for passes thrown against the Chiefs.
Now we’ll subtract the Chiefs data from the data for the whole league — which leaves us with a defensive version of Completion Percentage Over Expectation (CPOE).
:no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/19445146/chiefs__defensive_CPOE.png)
If you’d like to create graphs like this for yourself, I’ve posted the code online. Feel free to dm me on twitter if you have any questions
On this final chart, the blue areas represent where it’s been good for the Chiefs (fewer completions), while orange areas show where there are been more completions — that is, it’s been good for the opposing offenses.
You can see that the Chiefs have been better than average in multiple areas of the field — in particular at the seams and on the right side of the field in general. The Chiefs have, however, been somewhat vulnerable to the very deep ball, allowing more completions very deep in the middle of the field and along the left sideline.
How did the pass defense play against New England?
:no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/19445228/kc_cpoe_ne.png)
The main takeaway here is that the Chiefs defense really limited Brady’s ability to throw the ball deep. This likely came through a combination of stellar pass rush and tight coverage. Whatever the cause, Brady only completed one pass beyond 20 yards.
This season, the Chiefs offense came out of the gate hot. But through advanced analytics, it’s been the defense that has looked more favorable in recent weeks. This is comforting to see — but for the team to have a shot at anything more than an AFC West championship, they’ll need both units playing well. That’s something we have rarely (if ever) seen in the Mahomes era.