clock menu more-arrow no yes mobile

Filed under:

How does Jacoby Brissett compare?

NFL: Indianapolis Colts at Tennessee Titans Christopher Hanewinckel-USA TODAY Sports

There has been plenty of discussion surrounding Jacoby Brissett and what value he brings to the Indianapolis Colts passing game. Opinions range from him being a top 10 QB to not even being a starting caliber talent. I’ve read posts and articles comparing him favorably to Andrew Luck and I’ve seen analysis that calls him just a game manager.

So what’s the truth? Well, not to get philosophical but there is no truth. Football is complex game. We can talk about intangible variables, differing frames of reference, game theory and even nonlinear feedback systems all day and get nowhere closer to the “truth” of even a single football game.

About the only thing we can do is develop metrics to try and approximate the truth of a system and attempt to measure players against that truth.


With QB stats, we often do that by measuring correlation to wins. Now, as every statistician warns, correlation is not necessarily causation. Kneel-downs are highly correlated to wins but aren’t a great measure of QB play. Similarly, minimal passing yards is often more about a leading team running out the clock than it is a poor passing game.

Therefore, a better test for a stat is to see how repeatable it is by measuring it against future wins. This doesn’t completely solve the correlation-causation problem but it weeds out a lot of poser stats. The following chart shows a set of passing stats and their predictive correlations(1).

Mouseover Definitions: att, ttt, avg _yac, aDOT, ayd/cmp, cmp, y/cmp, sck%, sck_yds, sack, int, int%, yds, 1st, cmp%, td, ypa, cpoe, td%, (td-i)%, nya, psr, 1st/db, wPSR, epa/db,

The chart is sorted from lowest correlation to highest so, the stats on the left side of the chart aren’t very good at measuring QB skill and the ones towards the right are the best at doing that. They have the greatest truthiness.

Using some of the better stats, I will compare Jacoby Brissett to QBs from 2018 and the first 11 weeks of 2019 (2). I highlighted Brissett, Andrew Luck and some other QBs for comparison (Mitchell Trubisky, Blake Bortles, Dak Prescott & Baker Mayfield).

This is not meant to be an “is Brissett better than QB X” analysis. I think all but the most biased fan feels that the final year of Andrew Luck’s career was better than Brissett’s second year under center, which is in turn is better than the season Bortles had when he was benched.

Instead, I am just trying to give an idea of where Brissett fits in to the spectrum of stats most used to measure QBs. Along the way, I will directly compare 2019 Brissett to 2018 Luck as they are playing in the same system with many of the same players and coaches.


The go-to metric for accuracy tends to be completion rate.

Luck had the 11th highest completion rate of all QBs in 2018, while Brissett ranks 18th as of week 11. Completion rate however is not a great stat in isolation as it is dependent upon how far a QB throws the ball.

It’s pretty clear that shorter passes tend to have higher completion rates over longer passes and thus QBs that throw farther should be expected to have lower completion rates.

Now, average depth of target (aDOT) in isolation is not a measure of QB value. If you refer to the correlation chart there is almost no correlation to wins, meaning a short or long aDOT is not inherently a good or bad thing. However, when used in conjunction with completion rate it can make a pretty powerful stat.

Using aDOT, an expected completion rate above/below the actual completion rate can be determined. This is called completed percentage over expected (CPOE) and is a much better measure of accuracy than un-adjusted completion rate.

Both Luck and Brissett didn’t throw the ball far under Frank Reich (both rank 26th), which again is not necessarily a bad thing. However, when adjusting for that length on each pass, Luck’s completion rate was 2.3% higher than expected, which ranked 16th and Brissett is -1.9% below expected in 28th place.


One of the most popular stats for QB efficiency is yards per attempt.

Luck had the slightly higher measure but both rank poorly among their peers at 21st. To find a root cause, YPA can be deconstructed into average yards per completion multiplied by completion rate:

  • Yds / Att = (Yds / Cmp) * (Cmp / Att )

Since we already know that both Luck and Brissett had decent un-adjusted completion rates, their poor YPA must be driven by low avg completed yards.

Luck (25th) and Brissett (22nd) don’t measure well in this metric, which makes sense because we already knew that they had low aDOTs. We can further deconstruct completion yards into air yards and yards after the catch.

  • Yds / Cmp = (AirYds + YAC ) / Cmp

Luck has the higher air yards (6.0 to 4.9) but that is offset by lower avg YAC (4.7 to 5.8). So putting it all together:

  • YPA = Cmp % * (AirYards / Cmp + YAC / Cmp)
  • Luck: 7.2 = 67.3% * (6.0 + 4.7)
  • Brissett: 6.9 = 63.8% * (4.9 + 5.8)

or by ranking

  • Luck: 21st = 11th * (17th + 28th)
  • Brissett: 21st = 18th * (29th + 8th)

In other words, their poor yardage efficiency can be summarized as Luck had a YAC problem and Brissett has a depth problem.


Yards per attempt is a good QB measure but it ignores sacks, which should be part of the metric. The QB shares responsibility with the O-line for sacks partly because of Time to Throw. There is a moderate correlation between how long a QB holds the ball and sack rate.

Luck was notorious for holding the ball a long time and was often one of the most sacked QBs in the league until, under Reich, both his time to throw and sack rate dropped dramatically.

Luck had the 7th quickest release in 2018 while Brissett currently ranks 27th which likely contributes to their differing sack rates (Luck 1st, Brissett 9th).

Its important to note that like aDOT, Time to Throw and Sack Rate are not good measures of QB play in isolation. A QB that holds the ball a long time and takes a lot of sacks can still have very good yardage efficiency (Russell Wilson).

When incorporating the sack data into YPA, you get Net Yards per Attempt.

When accounting for sacks, Luck improved his ranking from 21st in YPA to 11th in NY/A, whereas Brissett only improves 2 spots from 21st to 19th.

To summarize:

  • Luck got rid of the ball quickly (9th), helping to avoid sacks (1st) and managed to achieve about average depth of throws (17th) with poor YAC (28th).
  • Brissett holds the ball a long time (27th), while still avoiding sacks (9th) and doesn’t throw the ball very far (29th) but gets good YAC from his receivers (8th).


The most basic success rate measure is also one of the better QB stats. First Downs per Dropback (1st/db) basically measures the ability of a QB to move the ball down the field with his arm.

Luck finished 10th best last year while Brissett is currently 16th. This is a critical stat and if a QB measures poorly, then it must be made up for with a good rushing game for a team is to advance the ball.

To further improve the stat, all plays, not just first downs are evaluated for success. Analysts typically utilize EPA for this. If you don’t know what EPA is, here is a good intro. Typically, passing success rate is defined as the % of all dropbacks that have EPA > 0.

However, one of the problems with that measure is it does not account for QBs racking up “success” in garbage time. I created a weighted Passing Success Rate (wPSR) in part to account for that. Here’s both.

In both measures, Luck ranks 4th and Brissett 16th. wPSR is a more comprehensive stat than 1st/db and not surprisingly it has a greater predictive correlation.


Other than passing yards -- which is a terrible stat -- probably the most common measure of QB value is passing TDs and INTs.

Both QBs do well in TD rate (Luck 4th, Brissett 8th) but Brissett is better at managing the ball in Reich’s system with a 12th best INT rate (Luck 19th).

Combining those individual stats gives a better overall measure, as a QB could have a high TD rate at the cost of INTs or low INTs at the cost of TDs. Typically, this combination is done with a TD to INT ratio but I actually dislike that math for a variety of reasons and instead prefer TD-INT rate differential:

(TD - I) % = (TD - INT) / dropbacks

Brissett has the better number, but Luck has the slightly better rank (6th vs 7th). Clearly both QBs excel here.

While (TD-I)% is a very powerful stat it is limited to measuring the impact of just a few throws. Incorporating all throws gives a better picture of overall QB value as it measures his contributions to getting the team in position to score and not just the score itself. The best way to incorporate all pass plays is by using EPA per dropback.

If you were forced to pick only one stat to measure a QB by this would be it. Luck finished 2018 7th in EPA/db while Brissett is currently 13th.


Each week I publish Brissett’s numbers in the 4 stats of:

  • EPA/db
  • wPSR
  • 1st/db
  • NY/A

I call them the top 4 stats because they have the best predictive correlations of any non-proprietary stat and are the best I can get to my version of “truth” for QB play. In those stats, Brissett ranks about the middle of all QBs.

For some people (like me), this means he is performing like an average QB and diving into the data reveals weaknesses that he needs to improve upon if the Colts are to have sustained success in the future. Others think that intangibles and unmeasured variables are more critical and that this data means nothing. Many of those people feel he doesn’t need to get any better for the team to continue winning.

Whatever the truth is, I think everyone agrees that these numbers don’t dictate that Brissett can’t get better.


  1. Predictive correlations are derived by comparing season team totals from 8 game random samples against the win totals from the remaining 8 out of sample games.
  2. Each year includes the top 32 QBs by passing attempts.