Which NFL play-caller runs the best offense?
Andy Reid. Over eleven seasons calling Kansas City’s plays, 2016 to 2026, his offenses have averaged 0.113 expected points added per play, a 47.1% success rate and 26.6 points a game, with the most pass-heavy approach in the league at 6.5 points above the expected pass rate. Ken Dorsey’s two Buffalo seasons match him at 0.112, Ben Johnson is third at 0.098 and goes for it on 30% of fourth downs. Mike McCoy (-0.188 over two seasons) and Adam Gase (-0.105 over five) are the worst.
- 0.113EPA per play for Andy Reid’s offenses over 10,500 plays since 2016, the best of any play-caller with two seasons
- +6.5points above the expected pass rate for Reid; Todd Downing was 7.1 points below
- 30%of fourth downs Ben Johnson has gone for, against 20% for the league in the same seasons
- 0.305EPA per dropback for Matt Ryan under Kyle Shanahan; -0.051 under Frank Reich
Based on 352 team-seasons and 106 play-callers from 2016 through week 3 of 2026, regular season only, about 140,000 offensive plays; the 2026 season is partial. Explore the live data.
Which NFL play-caller has the most efficient offense?
Andy Reid, at 0.113 expected points added per play across eleven Kansas City seasons, 10,500 plays and a 121-47 record. Ken Dorsey’s two seasons in Buffalo (0.112, 49.0% success, 27.5 points a game) are a hair behind on a fifth of the sample. Ben Johnson is third at 0.098 over five seasons in Detroit and Chicago, then Liam Coen (0.088), Todd Monken (0.084), Brian Schottenheimer (0.079), Matt LaFleur (0.075) and Kyle Shanahan (0.071 over eleven seasons, with the league’s highest explosive-play rate at 11.2%). Mike McCoy’s two seasons in Denver and Arizona are the worst at -0.188.
| Play-caller | Seasons | Teams | EPA/play | Success | Points a game |
|---|---|---|---|---|---|
| Andy Reid | 11 | KC | 0.113 | 47.1% | 26.6 |
| Ken Dorsey | 2 | BUF | 0.112 | 49.0% | 27.5 |
| Ben Johnson | 5 | DET, CHI | 0.098 | 46.8% | 28.3 |
| Liam Coen | 3 | TB, JAX | 0.088 | 47.3% | 28.6 |
| Todd Monken | 5 | TB, CLE, BAL | 0.084 | 46.6% | 26.8 |
| Brian Schottenheimer | 5 | SEA, DAL | 0.079 | 46.3% | 27.6 |
| Matt LaFleur | 9 | TEN, GB | 0.075 | 45.9% | 24.5 |
| Kyle Shanahan | 11 | ATL, SF | 0.071 | 46.3% | 26.0 |
| Jason Garrett | 2 | NYG | -0.111 | 40.4% | 16.3 |
| Hue Jackson | 2 | CLE | -0.121 | 37.7% | 15.6 |
| Brian Callahan | 2 | TEN | -0.138 | 38.7% | 17.5 |
| Mike McCoy | 2 | DEN, ARI | -0.188 | 36.5% | 16.1 |
The best single season of 2025 belonged to Josh McDaniels in New England, 0.159 EPA per play and 14-3, ahead of Sean McVay (0.145) and Joe Brady (0.136). The report lets you sort the board on any of a dozen measures and set the minimum sample yourself.
Which NFL play-callers pass the most?
Measured against what the down, distance, score and clock would predict, Andy Reid passes 6.5 points more often than expected, the most of any play-caller with two seasons, with Byron Leftwich (+5.2 in Tampa Bay), Randy Fichtner (+4.6), Joe Lombardi (+4.1) and Mike McCarthy (+3.5) next. The most run-heavy are Todd Downing (-7.1), Arthur Smith (-6.8), Zac Robinson (-6.7) and Luke Getsy (-6.5). Ben Johnson (-3.2) and Kyle Shanahan (-2.7) run more than expected and still rank near the top for efficiency.
| Play-caller | Pass rate | Over expected | Early downs | EPA per play |
|---|---|---|---|---|
| Andy Reid | 65.2% | +6.5 | 58.3% | 0.113 |
| Byron Leftwich | 67.1% | +5.2 | 59.6% | 0.051 |
| Randy Fichtner | 65.7% | +4.6 | 57.8% | -0.014 |
| Joe Lombardi | 67.3% | +4.1 | 58.9% | 0.051 |
| Mike McCarthy | 66.1% | +3.5 | 58.2% | 0.017 |
| Luke Getsy | 58.6% | -6.5 | 49.9% | -0.080 |
| Zac Robinson | 57.0% | -6.7 | 47.8% | -0.009 |
| Arthur Smith | 56.0% | -6.8 | 46.9% | 0.003 |
| Todd Downing | 56.7% | -7.1 | 46.5% | -0.027 |
Passing more is not the same as passing well: Fichtner’s Steelers threw more than almost anyone and lost ground on every play. Greg Roman, with Lamar Jackson in Baltimore, ran 4.5 points more than expected and still averaged 0.052 EPA per play.
What does Ben Johnson’s offense do differently?
It goes for it, runs from under center and fakes the handoff. Over his five seasons as a play-caller Johnson has gone for it on 30.2% of fourth downs against 20.3% for the rest of the league in the same seasons, used play action on 28.7% of plays against 21.9%, and sent a man in motion on 53.0% against 47.2%, while taking just 50.8% of snaps from the shotgun against a league 68.5%. His offenses pass less than the league (57.1% against 60.5%) and average 0.098 EPA per play against -0.006.
| Measure | Ben Johnson | League |
|---|---|---|
| Fourth-down go rate | 30.2% | 20.3% |
| Play action | 28.7% | 21.9% |
| Motion | 53.0% | 47.2% |
| Shotgun | 50.8% | 68.5% |
| No-huddle | 4.6% | 10.2% |
| Pass rate | 57.1% | 60.5% |
| EPA per play | 0.098 | -0.006 |
His passes to the deep middle have averaged 1.24 EPA against 0.50 for the league, and the intermediate middle 0.71 against 0.54. Joe Lombardi (26.3%), Joe Brady (25.9%) and Kliff Kingsbury (23.6%) are the next most aggressive on fourth down. The report’s Tendencies tab runs this comparison for any caller and any situation you pick: down, distance, field position, clock and score.
Which quarterbacks played best under which play-caller?
Ryan Tannehill under Arthur Smith in Tennessee, 0.259 EPA per dropback over 857 dropbacks in 2019 and 2020, is the best pairing with 800 dropbacks, with Josh Allen under Joe Brady (0.254) and Patrick Mahomes under Andy Reid (0.251, over 5,291 dropbacks and nine seasons) next, then Brock Purdy under Kyle Shanahan (0.237) and Allen under Ken Dorsey (0.216). The same quarterback can look like two players: Matt Ryan was worth 0.305 EPA per dropback under Shanahan in 2016 and -0.051 under Frank Reich in 2022.
| Quarterback | Play-caller | Dropbacks | EPA per dropback | CPOE |
|---|---|---|---|---|
| Ryan Tannehill | Arthur Smith | 857 | 0.259 | +4.6 |
| Josh Allen | Joe Brady | 1,196 | 0.254 | +2.4 |
| Patrick Mahomes | Andy Reid | 5,291 | 0.251 | +2.8 |
| Brock Purdy | Kyle Shanahan | 1,515 | 0.237 | +3.9 |
| Josh Allen | Ken Dorsey | 1,303 | 0.216 | +3.0 |
| Drew Brees | Sean Payton | 2,558 | 0.211 | +4.6 |
| Lamar Jackson | Todd Monken | 1,468 | 0.204 | +3.4 |
| Jared Goff | Ben Johnson | 1,850 | 0.192 | +1.7 |
| Aaron Rodgers | Matt LaFleur | 2,363 | 0.182 | +3.4 |
| Tom Brady | Josh McDaniels | 2,323 | 0.184 | +1.4 |
The biggest swings for one quarterback between two callers (500 dropbacks under each): Ryan from Shanahan to Reich, Deshaun Watson from Bill O’Brien (0.181) to Kevin Stefanski (-0.123), Tannehill from Smith to Adam Gase (-0.042), Jared Goff from Johnson to Anthony Lynn (-0.043) and Russell Wilson from Brian Schottenheimer (0.163) to Nathaniel Hackett (-0.075). The report shows every quarterback a caller has had, weighted by dropbacks, and whether each was better or worse under him than elsewhere.
Do NFL play-callers change when they are ahead?
All of them, but not by the same amount. The league passes on 48.6% of plays when leading by nine or more, 58.3% when tied and 72.5% when trailing by nine or more. Andy Reid keeps throwing with a lead, 55.6%, the most of any top-ten caller, and Ken Dorsey 52.3%; Todd Monken drops to 42.6% and Kyle Shanahan to 44.6%. Behind by nine, every one of them passes at least 68% of the time, Reid 77.1%.
| Play-caller | Leading by 9+ | Tied | Trailing by 9+ |
|---|---|---|---|
| Andy Reid | 55.6% | 65.8% | 77.1% |
| Ken Dorsey | 52.3% | 65.0% | 72.9% |
| Ben Johnson | 47.9% | 53.4% | 71.5% |
| Liam Coen | 51.2% | 59.3% | 74.5% |
| Todd Monken | 42.6% | 56.2% | 71.2% |
| Brian Schottenheimer | 49.1% | 55.3% | 71.7% |
| Matt LaFleur | 48.8% | 57.4% | 70.8% |
| Kyle Shanahan | 44.6% | 56.6% | 72.7% |
| League | 48.6% | 58.3% | 72.5% |
By down, the league passes 50.8% of the time on first down and 81.5% on third; Reid is at 55.5% and 83.2%, Greg Roman at 45.8% and 71.0%. Ken Dorsey’s offenses succeeded on 52.8% of third downs, the best of the leaders. The report draws both splits for any caller against the league.
How was this measured?
Play-by-play and coaching staffs come from nflverse and Wikipedia’s team-season articles, 2016 to the current season, refreshed weekly. The play-caller is a curated label: where a head coach called the plays it overrides the titled offensive coordinator, and a mid-season change is credited to the caller who finished the season. Expected points added, success and pass rate over expected are nflfastR model outputs; motion, play action and RPO come from FTN charting and exist from 2022.
Career figures weight each team-season by its plays; leaderboards need two seasons and 1,500 plays, quarterback pairings 800 dropbacks and quarterback seasons 100. The report’s Tendencies tab re-runs one governed query for the caller and situation chosen; a cell needs 60 plays before it is compared and a key for the defense needs a gap of five points against the league. The current season is partial.
The data lives in a Power BI semantic model on Microsoft Fabric. Every number on this page was produced by an AI assistant running the report’s own queries against that model, and the same model powers the live public report.
Want answers like this from your own data?
This started as public data. We added location and conditions, built a model worth trusting, and pointed an AI at it. We do the same thing with sales, operations and finance data that nobody outside your company will ever see.