
Click here to skip to the final results and the compiled data.
The term “clutch” describes a player who is able to achieve success at a critical, high-pressure point in a competition.
The characteristic of being “clutch” in sports is a controversial topic because many statisticians argue that there is no evidence supporting the notion of certain players being inherently more clutch than others. While some players may perform better in clutch situations than others, how do we know that this is because they’re better in high-pressure situations and not because they’re just better in general? In other words, these people would argue that Michael Jordan’s perceived clutchness was just his greatness being more exposed in clutch situations even though he was always great.
There are interesting points from both sides of the debate. I personally tend to think that clutchness exists. It’s hard to watch players like Tom Brady and LeBron James in the postseason and think that they don’t perform at a higher level while under pressure. Maybe statistics are simply not yet able to accurately quantify elevated performance in high-pressure situations.
I tried to determine which quarterbacks performed at the highest level in clutch situations this season, but there’s always room for improvement. I think the results pass the eye test, though.
The Process
We have to start by coming up with a definition for clutch situations.
I toyed around with different options to see what yielded the best sample size for analysis while also exemplifying high-pressure situations. In the end, I defined a clutch situation as any play which takes place after halftime while the difference in points scored between both teams is less than or equal to eight (a one-possession game). Naturally, this is mostly subjective and arbitrary. The conditions which make sense to me might seem illogical to you. I tried to pick the best definition, but some degree of dissent is inevitable.
Brian Burke’s Expected Points metric was used for quantifying quarterback performance. I think it’s best suited for this project because it adds more weight to the plays which deserve it. The most popular statistic for evaluating quarterback performance is Adjusted Net Yards Per Attempt (ANY/A). The issue is that ANY/A views a five-yard pass on 4th and 7 the same way as a five-yard pass on 4th and 4. One of them keeps a drive alive, one of them is effectively a turnover. We have to account for this massive difference for proper evaluation of clutch performance.
Expected Points determines the expected points a team will end the drive with before (EPB) and after (EPA) a play. On first down with five yards to go at the Cleveland Browns’ five-yard line, the New Orleans Saints were given 6.060 Expected Points before the play. That makes sense. Teams will score a touchdown in this situation more often than not. On the next play, Drew Brees threw a touchdown to Micheal Thomas. For this play, the EPB (Estimated Points Before) was 6.060, while the EPA (Estimated Points After) was 7.000. The difference between the two tells us the number of points that a player added. On this play, Drew Brees isn’t given that much credit because he was in such a favorable situation. If he threw an interception, he would have been penalized more harshly because of the number of points he cost his team.
We’ll use the average number of Expected Points a quarterback added on plays which meet our criteria for a clutch situation. Any play which lists the quarterback on the play-by-play log was included except for quarterback kneels and spiked balls. Any attempted pass, sack, or quarterback rush was included.
Finally, I adjusted for strength of schedule. It should be obvious why this is necessary — certain quarterbacks face better defenses than others. I also factored in the number of plays in a clutch situation that a quarterback had against different defenses. Playing 97 plays against great defenses and 3 plays against horrible defenses isn’t the same as a 50/50 split.
Finally, let’s move on to the final data.
Results
I included all 32 starting quarterbacks plus a few big names who didn’t play the full season but played significant time (Fitzpatrick, Foles, Jackson). Clutch Value is defined as the average Expected Points added by a quarterback per play in clutch situations (one-possession game after halftime).
Rank | Player | Clutch Value | Plays |
---|---|---|---|
1 | Tom Brady | 0.468 | 135 |
2 | Patrick Mahomes | 0.438 | 168 |
3 | Matt Ryan | 0.355 | 153 |
4 | Drew Brees | 0.329 | 120 |
5 | Mitchell Trubisky | 0.319 | 142 |
6 | Jared Goff | 0.312 | 136 |
7 | Philip Rivers | 0.305 | 110 |
8 | Marcus Mariota | 0.268 | 112 |
9 | Nick Foles | 0.258 | 52 |
10 | Derek Carr | 0.221 | 106 |
11 | Dak Prescott | 0.215 | 185 |
12 | Ben Roethlisberger | 0.203 | 196 |
13 | Deshaun Watson | 0.159 | 157 |
14 | Russell Wilson | 0.158 | 176 |
15 | Josh Allen | 0.148 | 104 |
16 | Aaron Rodgers | 0.145 | 171 |
17 | Nick Mullens | 0.135 | 71 |
18 | Andrew Luck | 0.116 | 183 |
19 | Carson Wentz | 0.114 | 144 |
20 | Eli Manning | 0.109 | 164 |
21 | Alex Smith | 0.093 | 40 |
22 | Cam Newton | 0.081 | 195 |
23 | Andy Dalton | 0.077 | 109 |
24 | Joe Flacco | 0.061 | 99 |
25 | Sam Darnold | 0.053 | 104 |
26 | Lamar Jackson | 0.05 | 115 |
27 | Case Keenum | 0.038 | 133 |
28 | Ryan Fitzpatrick | -0.006 | 21 |
29 | Matthew Stafford | -0.011 | 94 |
30 | Baker Mayfield | -0.017 | 129 |
31 | Josh Rosen | -0.072 | 90 |
32 | Jameis Winston | -0.101 | 87 |
33 | Kirk Cousins | -0.129 | 104 |
34 | Blake Bortles | -0.165 | 100 |
35 | Ryan Tannehill | -0.243 | 90 |
The first thing I noticed was Tom Brady’s ranking at #1. Nearly every passing metric has him outside of the top-5 this season, but this metric truly displays his ability to elevate his game when it matters. I was also surprised by Josh Allen’s high ranking compared to other rookies. Does Josh Allen deserve more praise for his rookie season or is this metric misrepresenting his true performance?
One thing that bothered me was the fact that some quarterbacks were faced with more clutch situations than others. I believe that performing well over a larger sample of plays should be rewarded. Therefore, I adjusted every quarterback’s Clutch Value based on their Clutch Play Percentage (plays in clutch situation / total plays). A quarterback’s Clutch Value decreased if their CP% was below league-average, and their Clutch Value increased with their CP% was above league-average.
I separated these results from the prior data because some people may disagree with the need for the adjustment.
Rank | Player | Adjusted Clutch Value | Clutch Play Pct. |
---|---|---|---|
1 | Patrick Mahomes | 0.478 | 25.20% |
2 | Tom Brady | 0.444 | 22.00% |
3 | Mitchell Trubisky | 0.372 | 27.00% |
4 | Matt Ryan | 0.343 | 22.40% |
5 | Drew Brees | 0.318 | 22.30% |
6 | Jared Goff | 0.287 | 21.40% |
7 | Marcus Mariota | 0.284 | 24.60% |
8 | Nick Foles | 0.283 | 25.50% |
9 | Dak Prescott | 0.262 | 28.20% |
10 | Philip Rivers | 0.26 | 19.70% |
11 | Ben Roethlisberger | 0.235 | 26.80% |
12 | Russell Wilson | 0.22 | 32.30% |
13 | Deshaun Watson | 0.162 | 23.60% |
14 | Derek Carr | 0.161 | 16.90% |
15 | Aaron Rodgers | 0.155 | 24.80% |
16 | Carson Wentz | 0.152 | 30.90% |
17 | Josh Allen | 0.152 | 23.80% |
18 | Nick Mullens | 0.133 | 23.00% |
19 | Andrew Luck | 0.131 | 26.10% |
20 | Eli Manning | 0.121 | 25.70% |
21 | Cam Newton | 0.114 | 32.40% |
22 | Andy Dalton | 0.09 | 27.10% |
23 | Lamar Jackson | 0.078 | 36.30% |
24 | Joe Flacco | 0.063 | 23.90% |
25 | Sam Darnold | 0.049 | 21.30% |
26 | Alex Smith | 0.041 | 10.20% |
27 | Case Keenum | 0.034 | 20.60% |
28 | Ryan Fitzpatrick | -0.002 | 7.60% |
29 | Matthew Stafford | -0.007 | 15.20% |
30 | Baker Mayfield | -0.018 | 23.50% |
31 | Josh Rosen | -0.061 | 19.50% |
32 | Jameis Winston | -0.083 | 19.20% |
33 | Kirk Cousins | -0.084 | 15.10% |
34 | Blake Bortles | -0.145 | 20.30% |
35 | Ryan Tannehill | -0.277 | 26.40% |
Mitchell Trubisky is the unexpected hero of today’s quarterback analysis. Last time it was Deshaun Watson. Maybe I’ll find a metric that values Nathan Peterman next time.
Comments