Playing the Odds | How NFL Coaches Learned to Take Risks

Plays analyzed

2016–2025 seasons

4th-down go rate

in 2025 (12.8% in 2016)

Go rate, 4th & short

4th & 2 or less, 2025 (34.0% in 2016)

4th downs converted

of go-for-it attempts, 2016–2025

Kickoffs returned

in 2025 (25.2% in 2023)

Home-team win rate

2023–2025 (58.2% in 2016–2018)

The Scoreboard · 2016–2025

Finding 1 · Fourth downs

Coaches go for it on fourth down nearly twice as often as they did in 2016

In 2016, teams went for it on 12.8% of fourth downs and punted or kicked a field goal on the rest. By 2025 that share had climbed to 23.3%. That is 882 fourth-down attempts in a season, up from 477.

The change is even bigger when only a yard or two is needed. On 4th and 2 or less, the go-for-it rate rose from 34.0% in 2016 to 63.9% in 2025. Going for it used to be the exception on a short fourth down; now it is the norm. The climb was not perfectly smooth (the overall rate dipped to 18.8% in 2022), but each of the last six seasons is above every season before 2020.

Share of 4th downs where the offense went for it

Regular season, by year

Finding 2 · Does the gamble pay?

Short fourth downs are a good bet: 4th and 1 is converted two times out of three

Across all ten seasons, teams converted 53.4% of their fourth-down attempts. Distance is everything. On 4th and 1 the success rate is 67.1% (2,648 attempts); on 4th and 2 it is 59.3%. It falls to 50.1% at 3–5 yards, 35.8% at 6–10 yards and 20.8% at 11 or more.

In investing terms, look at the expected return. Measured by expected points added (EPA, the average change in points a play is worth), a go-for-it attempt averaged +0.058 points, while a punt averaged −0.107. Coaches mostly go for it when the odds are good, so this does not prove every punt is a mistake, but the short-yardage numbers explain why the trend in Finding 1 happened.

4th-down conversion rate by yards needed

All go-for-it attempts, regular season 2016–2025

Finding 3 · Who takes the most risk

Dan Campbell's Detroit Lions own four of the ten most aggressive seasons

Ranking every team-season by fourth-down go rate, the 2023 Lions top the list at 33.9%. Detroit under head coach Dan Campbell appears four times in the top ten (2021, 2022, 2023 and 2024), and Carolina appears twice. Every one of the ten most aggressive seasons came in 2021 or later.

Over all ten years, Philadelphia went for it most often (22.5% of fourth downs), just ahead of Detroit (21.9%). The most cautious teams were Seattle (12.2%), Pittsburgh (13.9%) and San Francisco (14.8%).

Top 10 team-seasons by 4th-down go rate

Regular season; hover for coach and number of attempts

Finding 4 · Risk and winning

Aggressive teams seem to lose more, until you remove the desperate situations

A naive look says aggression loses. Splitting the 320 team-seasons into four equal groups by overall go rate, the least aggressive quarter won 55.0% of their games and the most aggressive won just 44.4%. But that is backwards cause and effect: teams that are losing late in games have to go for it.

So look only at "choice" fourth downs, where nobody is forced: 4th and 2 or less, in the first three quarters, with the score within 7 points. Now the pattern flips. Teams that went for it least in these spots (20.0% of the time) won 47.1% of games; teams that went for it most (73.0%) won 53.0%. This is a relationship, not proof, since good teams may simply trust their offense more.

Win % by how aggressive a team was

320 team-seasons split into four equal groups

Finding 5 · Run or pass

Passing still pays more, but running has nearly caught up

The analytics message of the 2010s was simple: pass more. In 2016 a pass play succeeded (gained positive EPA) 45.6% of the time, versus 39.8% for a run. By 2025 the run success rate had risen to 43.0% while passing slipped to 44.7%, shrinking a 5.8-point gap to 1.7 points.

Expected points tell the same story: the gap between a pass (+0.045) and a run (−0.056) in 2016 narrowed to +0.018 versus −0.006 in 2025. Teams have responded. The share of plays that were passes fell from 59.9% in 2016 to 56.8% in 2025.

Success rate: pass plays vs. run plays

Share of plays with positive expected points added, regular season

Finding 6 · After the touchdown

A two-point try is a coin flip that pays exactly what an extra point does

From 2016 to 2025, teams converted 47.2% of two-point tries. Worth two points each, that is an expected 0.94 points per try. Extra points were good 94.3% of the time, also 0.94 expected points.

With the payoff equal on average, the choice only matters in specific game situations, which is why attempts stayed rare (between 81 and 153 per season). Year to year the two-point payoff swings more, from 1.07 points in 2023 to 0.81 in 2024: the same average return, with more risk.

Expected points per try

Two-point success rate × 2 vs. extra-point success rate × 1

Finding 7 · Kickers

Kickers made the 50-yard field goal routine, and coaches noticed

In 2016 kickers made 56.7% of field goals from 50 yards or more. In 2025 they made 68.8%. Coaches responded by trying far more of them: 266 attempts from 50+ in 2025 (and a peak of 279 in 2024), up from 150 in 2016.

Kicks under 40 yards were already near-automatic and stayed that way (94.1% in 2016, 95.1% in 2025). The long kick is where the improvement happened, and it gives coaches another option on fourth down in the 30–40 yard range.

Field goal success rate by distance

Regular season, by year

Finding 8 · Rule changes

The kickoff came back from the dead after two rule changes

For years about four in ten kickoffs were returned (39.4% in 2016, 37.5% in 2022). In 2023 a fair catch inside the 25 began counting as a touchback, and the return rate fell to 25.2%.

The league then redesigned the play. The new kickoff format in 2024 lifted returns to 32.8%, and after touchbacks were moved out to the 35-yard line in 2025, kicking into the end zone became a bad deal: 74.4% of kickoffs were returned and the touchback rate dropped from 64.2% to 20.7%. Here the rule makers changed the odds, and teams adjusted immediately.

What happened on kickoffs

Share of all kickoffs, regular season

Finding 9 · Home field

Home-field advantage is fading, and the drop started before COVID emptied the stadiums

Home teams won 58.2% of games from 2016 to 2018 but only 54.2% from 2023 to 2025. The low point was 2020, when most games were played with few or no fans: home teams won just 49.8%, less than half.

The surprise is 2019. With full stadiums, home teams won only 51.8% and were outscored on average by 0.14 points per game, a drop from 60.2% in 2018. Empty stadiums did not create the decline; they arrived in the middle of it.

Home-team win %

Regular-season games, ties excluded; orange bars = 2019 and 2020

About the data

Where it comes from

All data comes from nflverse, a free, public project that publishes official NFL play-by-play data with added measures such as expected points added (EPA). I downloaded the ten seasons from 2016 through 2025, one file per season (484,254 rows and 372 columns in total). The scripts that download, clean and analyze the data are in the GitHub repository, so every number on this page can be reproduced.

What one row is

One row is one play: one team's offense on one snap in one game, with the season, week, teams, down, distance, field position, score, play type and result. The cleaned file (data/plays.csv) has 419,148 rows and 28 columns. A second file (data/games.csv) has one row per game (2,761 games) with the final score.

Rows I dropped and why

No-plays: a penalty wiped out the play, so no result counted.
Non-plays: timeouts, end-of-quarter markers and the two-minute warning.
Kneel-downs and spikes: plays that only run or stop the clock, not real attempts to gain yards.
Broken rows: a handful with no offense or no EPA value.

That removes 65,106 rows. The report uses regular-season games only (400,398 plays); playoff games are kept in the data file and can be added on the dashboard.

How each number is calculated

4th-down go rate	Go-for-it attempts (runs and passes) ÷ all 4th-down plays that were a run, pass, punt or field goal try.
4th-down conversion	A go-for-it attempt that gained at least the yards needed or scored a touchdown ÷ all go-for-it attempts.
"Choice" 4th down	4th and 2 or less, in quarters 1–3, with the score within 7 points.
EPA per play	Average expected points added (from nflverse's model): how much a play changed the number of points the offense can expect to score.
Success rate	Plays with EPA greater than zero ÷ all plays in the group.
Pass rate	Pass plays ÷ (pass plays + run plays). Sacks count as passes; quarterback scrambles count as runs (nflverse's labels).
Team win %	(Wins + ½ × ties) ÷ games played, regular season.
Aggressiveness groups	Team-seasons sorted by go rate and split into four equal groups (quartiles); the win % shown is each group's average.
Expected points per try	Two-point: success rate × 2. Extra point: success rate × 1.
Field goal rate	Made field goals ÷ attempts (blocked kicks count as misses), grouped by kick distance.
Kickoff return rate	Kickoffs where a player fielded the ball to return it ÷ all kickoffs. Touchback rate is touchbacks ÷ all kickoffs.
Home win %	Games won by the home team ÷ games that did not end in a tie.

Explore the data yourself →