// LOADING...
// LOADING...
THE NUMBERS BEHIND THE VIDEO
How difficult were the throws? How well did your quarterback execute? Pick two passers and put their results in league context.
2026 regular season · Weeks 1–6 · Frozen video snapshot
Read the Brock Purdy report →Right: more completable attempts in the model. Up: more completions above expectation.
Scroll the chart horizontally, or read the values below.
Brock Purdy · Expected completion: 66.6% · Completion above expected: +5.13 pp · 166 modeled attempts
SF · 191 ANALYTICAL DROPBACKS
EPA / dropback
0.390
#1 of 33 · n=191
Completion above expected
+5.13 pp
#8 of 33 · n=166
Difficult-throw completion
54.5%
#2 of 31 · n=33
Passing grade under pressure
80.2
#1 of 33 · n=59
Difficult throws: 18/33 completed. FTN coverage: 136 targeted throws across 5 games.
BUF · 203 ANALYTICAL DROPBACKS
EPA / dropback
0.365
#3 of 33 · n=203
Completion above expected
+5.13 pp
#7 of 33 · n=160
Difficult-throw completion
35.3%
#23 of 31 · n=34
Passing grade under pressure
45.2
#17 of 33 · n=79
Difficult throws: 12/34 completed. FTN coverage: 159 targeted throws across 6 games.
SCHEME CONTEXT
These splits show contact at the catch point, not yards of separation or how much a route concept created. Lower is not automatically better: target choice, coverage and receivers change the opportunity. Purdy’s FTN data covers five games; Week 6 charting was incomplete.
Bars: 0–100% of charted targets. Minimum 20 per split.
Bars: 0–100% of charted targets. Minimum 20 per split.
Each row uses one provider and one definition. Expand the method below for coverage and limits.
| METRIC | Brock Purdy | Josh Allen |
|---|---|---|
| EPA / dropbackAverage QB EPA on passes, sacks and scrambles; excludes spikes and kneels. Minimum 100 dropbacks. · nflverse | 0.390191 dropbacks#1 / 33 | 0.365203 dropbacks#3 / 33 |
| Success rateShare of dropbacks with positive play EPA. Minimum 100 dropbacks. · nflverse | 54.5%191 dropbacks#2 / 33 | 47.8%203 dropbacks#14 / 33 |
| Expected completionMean nflfastR completion probability. Higher means a more completable set of attempts in this model, not better QB play. It does not observe receiver separation. · nflverse | 66.6%166 modeled attemptsContext, not a best-to-worst rank | 66.1%160 modeled attemptsContext, not a best-to-worst rank |
| Completion above expectedActual minus expected completion rate, in percentage points, on the same modeled attempts. Minimum 50 modeled attempts and 100 overall dropbacks. · nflverse | +5.13 pp166 modeled attempts#8 / 33 | +5.13 pp160 modeled attempts#7 / 33 |
| Difficult-throw completionCompletion rate on attempts with nflfastR completion probability ≤50%. Minimum 15 such attempts. Receivers contribute to completions too. · nflverse | 54.5%33 difficult attempts#2 / 31 | 35.3%34 difficult attempts#23 / 31 |
| Difficult-throw CPOECompletion rate above expectation on attempts with completion probability ≤50%. Minimum 15 such attempts; accounts for differences within that difficulty bucket. · nflverse | +14.86 pp33 difficult attempts#2 / 31 | -0.50 pp34 difficult attempts#22 / 31 |
| Contested-target shareShare of charted, intended targets flagged contested by FTN (contact at the catch). Excludes throwaways and missing charting. Minimum 50 targets. Uncontested does not mean wide open. · FTN + nflverse | 13.2%136 charted targetsContext, not a best-to-worst rank | 10.7%159 charted targetsContext, not a best-to-worst rank |
| Contested completionCompleted passes divided by FTN contested targets. Minimum 15 contested targets. Catch-point contact is not a separation measurement. · FTN + nflverse | 50.0%18 contested targetsT-2 / 26 | 29.4%17 contested targets#20 / 26 |
| Contested share: play actionContested-target share on play-action throws. Minimum 20 charted targets. Comparing this with no play action is descriptive, not a causal estimate of scheme. · FTN + nflverse | 9.7%31 charted PA targetsContext, not a best-to-worst rank | 5.0%40 charted PA targetsContext, not a best-to-worst rank |
| Contested share: no play actionContested-target share without play action. Minimum 20 charted targets. Coverage, receivers, target choice and timing all affect this number. · FTN + nflverse | 14.3%105 charted non-PA targetsContext, not a best-to-worst rank | 12.6%119 charted non-PA targetsContext, not a best-to-worst rank |
| EPA without play actionPFF EPA per dropback without play action. Minimum 30 split dropbacks. PFF EPA and nflfastR QB EPA use different models; compare within a row. · PFF | 0.420145 non-PA dropbacks#1 / 33 | 0.380125 non-PA dropbacks#2 / 33 |
| Passing grade under pressurePFF passing grade under pressure, on its 0–100 scale. Minimum 30 pressured dropbacks. A grade is not a completion percentage. · PFF | 80.259 pressured dropbacks#1 / 33 | 45.279 pressured dropbacks#17 / 33 |
| Pressure-to-sack ratePFF percentage of pressures converted to sacks. Lower is better. Minimum 100 overall PFF dropbacks. · PFF | 8.5%59 pressured dropbacks#2 / 33 | 13.9%79 pressured dropbacks#10 / 33 |
| Quick-game EPAPFF EPA per dropback in its short time-in-pocket bucket (2.5 seconds or less). Minimum 30 split dropbacks. Ties use the precision published by PFF. · PFF | 0.54069 quick dropbacksT-1 / 33 | 0.44075 quick dropbacks#3 / 33 |
| Overall passing gradePFF passing grade, 0–100. Minimum 100 overall PFF dropbacks. · PFF | 84.5192 PFF dropbacks#3 / 33 | 70.4200 PFF dropbacks#18 / 33 |
33 eligible quarterbacks · higher is better
Actual minus expected completion rate, in percentage points, on the same modeled attempts. Minimum 50 modeled attempts and 100 overall dropbacks.
| RANK | QUARTERBACK | VALUE | SAMPLE |
|---|---|---|---|
| 1 | Caleb WilliamsCHI | +9.84 pp | 82 |
| 2 | Lamar JacksonBLT | +8.09 pp | 93 |
| 3 | Jayden DanielsWAS | +7.36 pp | 103 |
| 4 | Joe BurrowCIN | +6.01 pp | 186 |
| 5 | Kyler MurrayMIN | +5.68 pp | 90 |
| 6 | Drake MayeNE | +5.17 pp | 173 |
| 7 | Josh AllenBUF | +5.13 pp | 160 |
| 8 | Brock PurdySF | +5.13 pp | 166 |
| 9 | Geno SmithNYJ | +4.21 pp | 168 |
| 10 | C.J. StroudHST | +3.89 pp | 203 |
| 11 | Trevor LawrenceJAX | +3.43 pp | 178 |
| 12 | Bryce YoungCAR | +2.98 pp | 181 |
| 13 | Patrick MahomesKC | +2.51 pp | 147 |
| 14 | Jacoby BrissettARZ | +1.71 pp | 227 |
| 15 | Jared GoffDET | +1.31 pp | 188 |
| 16 | Dak PrescottDAL | +0.81 pp | 240 |
| 17 | Matthew StaffordLA | +0.80 pp | 239 |
| 18 | Baker MayfieldTB | +0.32 pp | 93 |
| 19 | Sam DarnoldSEA | -0.12 pp | 139 |
| 20 | Tyler ShoughNO | -0.68 pp | 225 |
| 21 | Kirk CousinsLV | -0.75 pp | 203 |
| 22 | Aaron RodgersPIT | -0.85 pp | 217 |
| 23 | Cam WardTEN | -1.30 pp | 165 |
| 24 | Michael Penix Jr.ATL | -1.49 pp | 107 |
| 25 | Deshaun WatsonCLV | -2.17 pp | 171 |
| 26 | Jalen HurtsPHI | -2.19 pp | 142 |
| 27 | Bo NixDEN | -2.40 pp | 171 |
| 28 | Daniel JonesIND | -3.41 pp | 204 |
| 29 | Jordan LoveGB | -3.55 pp | 221 |
| 30 | Justin HerbertLAC | -5.65 pp | 173 |
| 31 | Jalon DanielsTB | -6.57 pp | 94 |
| 32 | Malik WillisMIA | -7.34 pp | 128 |
| 33 | Jameis WinstonNYG | -7.52 pp | 155 |
Expected completion describes the attempts a QB chose. CPOE measures how often those attempts were completed above the model’s expectation. Both depend on receivers and context as well as the quarterback. Low-probability throws are defined here as 50% or less.
The nflfastR model uses play-by-play features such as air yards, field position and QB hits. It does not use tracking-based receiver separation and is different from NFL Next Gen Stats’ completion model.
There is no route-created openness score in this dataset. FTN contested flags describe contact at the catch point. They do not measure yards of separation. Play-action splits show an association; they cannot separate coaching, route running, defensive coverage, QB timing and target selection.
Purdy’s FTN sample covers five games. Week 6 against Washington contained placeholder charting in the October 20 snapshot and is excluded. His PFF and nflverse totals include all six games. Missing charting is never counted as an uncontested target.
Frozen video companion: 2026 regular season, Weeks 1–6, through October 19; retrieved October 20, 2026. This is not a live-season leaderboard. Base cohort: 33 quarterbacks with at least 100 nflverse analytical dropbacks, each also qualifying for the PFF 100-dropback cohort.
nflverse play-by-play: QB EPA on passes, sacks and scrambles, excluding spikes, kneels and missing EPA. QB EPA adjusts receiver fumbles but still includes yards after catch. Success means positive play EPA. Completion models apply only to attempts with an available probability.
PFF: passing summary, pressure, concept and time-in-pocket reports, filtered to Weeks 1–6. PFF counts 192 Purdy dropbacks; the nflverse filter counts 191. Provider models and denominators are kept separate.
Minimums: 50 modeled attempts for CPOE; 15 low-probability attempts; 50 FTN charted targets; 15 contested targets; 20 charted targets in each play-action split; 30 PFF split dropbacks. A dash means the sample did not qualify. Ranks use unrounded values and competition ties (1, 1, 3). Context measures have no “best” rank.
The full chart and table use the same observations. Axes are linear; no points are repositioned to create an outlier. These are descriptive results, without opponent adjustment or a causal separation of QB and supporting cast.