nflverse/nflfastR
Functions to Efficiently Access NFL Play by Play Data

Package index
Search the nflverse/nflfastR package
Vignettes
Functions
127
Source code
51
Man pages
28
Home
/
GitHub
/
nflverse/nflfastR
/

In nflverse/nflfastR: Functions to Efficiently Access NFL Play by Play Data 

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)
future::plan("multisession")
options(dplyr.summarise.inform = FALSE)
options(nflreadr.verbose = FALSE)

If you are new to R or are having trouble understanding the code in the below sections we highly recommend the nflfastR beginner's guide in vignette("beginners_guide").

The Main Functions

nflfastR comes with a set of functions to access NFL play-by-play data and team rosters. This section provides a brief introduction to the essential functions.

nflfastR processes and cleans up play-by-play data and adds variables through it's models. Since some of these tasks are performed by separate functions, the easiest way to compute the complete nflfastR dataset is build_nflfastR_pbp(). The main input for that function is a set of game ids which can be accessed with fast_scraper_schedules(). The following code demonstrates how to build the nflfastR dataset for the Super Bowls of the 2017 - 2019 seasons.

library(nflfastR)
library(dplyr, warn.conflicts = FALSE)
ids <- nflfastR::fast_scraper_schedules(2017:2019) |>
  dplyr::filter(game_type == "SB") |>
  dplyr::pull(game_id)
pbp <- nflfastR::build_nflfastR_pbp(ids)

In most cases, however, it is not necessary to use this function for individual games, because nflfastR provides both a data release and two main play-by-play functions: load_pbp() and update_db(). We cover load_pbp() below, and please see [Example 8: Using the built-in database function] for how to work with the database function update_db().

The easiest way to access the data from the release is the function load_pbp(). It can load multiple seasons directly into memory and supports multiple data formats. Loading all play-by-play data of the 2022-2024 seasons is as easy as 

pbp <- nflfastR::load_pbp(2022:2024)

Joining roster data to the play-by-play data set is possible as well. The data can be accessed with the function fast_scraper_roster() and its application is demonstrated in [Example 10: Working with roster and position data].

Application Examples

All examples listed below assume that the following two libraries are installed and loaded.

``` {r load, warning = FALSE, message = FALSE} library(nflfastR) library(tidyverse)

## Example 1: Completion Percentage Over Expected (CPOE)

Let's look at CPOE leaders from the 2009 regular season.

As discussed above, `nflfastR` has a data release for all available seasons, so there's no need to actually scrape them. Let's use that here with the convenience function `load_pbp()` which fetches data from the release (for non-R users, .csv and .parquet are also available in the [data release](https://github.com/nflverse/nflverse-data/releases/tag/pbp)).

``` {r ex3-cpoe, warning = FALSE, message = FALSE}
games_2009 <- nflfastR::load_pbp(2009) |> dplyr::filter(season_type == "REG")
games_2009 |>
  dplyr::filter(!is.na(cpoe)) |>
  dplyr::summarize(
    passer = nflreadr::stat_mode(passer_player_name),
    cpoe = mean(cpoe), 
    Atts = n(), 
    .by = passer_player_id
  ) |>
  dplyr::filter(Atts > 200) |>
  dplyr::slice_max(cpoe, n = 5) |>
  knitr::kable(digits = 1)

Example 2: Using Drive Information

When working with nflfastR, drive results are automatically included. We use fixed_drive and fixed_drive_result since the NFL-provided information is a bit wonky. Let's look at how much more likely teams were to score starting from 1st & 10 at their own 20 yard line in 2015 (the last year before touchbacks on kickoffs changed to the 25) than in 2000.

``` {r ex4, warning = FALSE, message = FALSE} pbp <- nflfastR::load_pbp(c(2003, 2015))

out <- pbp |> dplyr::filter(season_type == "REG" & down == 1 & ydstogo == 10 & yardline_100 == 80) |> dplyr::mutate(drive_score = dplyr::if_else(fixed_drive_result %in% c("Touchdown", "Field goal"), 1, 0)) |> dplyr::summarize(drive_score = mean(drive_score), .by = season)

out |> knitr::kable(digits = 3)

So `r scales::percent(out$drive_score[1], accuracy = 0.1)` of 1st & 10 plays from teams' own 20 would see the drive end up in a score in 2003, compared to `r scales::percent(out$drive_score[2], accuracy = 0.1)` in 2015. This has implications for Expected Points models (see [this article](https://www.opensourcefootball.com/posts/2020-09-28-nflfastr-ep-wp-and-cp-models/)).

## Example 3: Plot offensive and defensive EPA per play for a given season

Let's build the **[NFL team tiers](https://rbsdm.com/stats/stats/)** using offensive and defensive expected points added per play for the 2005 regular season. Creating data viz including NFL team logos (or wordmarks, or headshots), we recommend the nflverse R package [nflplotR](https://nflplotr.nflverse.com).

When using `load_pbp()`, the helper function `clean_pbp()` has already been run, which creates "rush" and "pass" columns that (a) properly count sacks and scrambles as pass plays and (b) properly include plays with penalties. Using this, we can keep only rush or pass plays.

```r
library(nflplotR)
pbp <- nflfastR::load_pbp(2005) |>
  dplyr::filter(season_type == "REG") |>
  dplyr::filter(!is.na(posteam) & (rush == 1 | pass == 1))
offense <- pbp |>
  dplyr::group_by(team = posteam) |>
  dplyr::summarise(off_epa = mean(epa, na.rm = TRUE))
defense <- pbp |>
  dplyr::group_by(team = defteam) |>
  dplyr::summarise(def_epa = mean(epa, na.rm = TRUE))
offense |>
  dplyr::inner_join(defense, by = "team") |>
  ggplot2::ggplot(aes(x = off_epa, y = def_epa)) +
  ggplot2::geom_abline(slope = -1.5, intercept = c(.4, .3, .2, .1, 0, -.1, -.2, -.3), alpha = .2) +
  nflplotR::geom_mean_lines(aes(y0 = off_epa, x0 = def_epa)) +
  nflplotR::geom_nfl_logos(aes(team_abbr = team), width = 0.07, alpha = 0.7) +
  ggplot2::labs(
    x = "Offense EPA/play",
    y = "Defense EPA/play",
    caption = "Data: @nflfastR",
    title = "2005 NFL Offensive and Defensive EPA per Play"
  ) +
  ggplot2::theme_bw() +
  ggplot2::theme(
    plot.title = ggplot2::element_text(size = 12, hjust = 0.5, face = "bold")
  ) +
  ggplot2::scale_y_reverse()

Example 4: Expected Points calculator

We have provided a calculator for working with the Expected Points model. Here is an example of how to use it, looking for how the Expected Points on a drive beginning following a touchback has changed over time.

While I have put in 'SEA' for home_team and posteam, this only matters for figuring out whether the team with the ball is the home team (there's no actual effect for given team; it would be the same no matter what team is supplied).

``` {r ex6a} data <- tibble::tibble( "season" = 1999:2019, "home_team" = "SEA", "posteam" = "SEA", "roof" = "outdoors", "half_seconds_remaining" = 1800, "yardline_100" = c(rep(80, 17), rep(75, 4)), "down" = 1, "ydstogo" = 10, "posteam_timeouts_remaining" = 3, "defteam_timeouts_remaining" = 3 ) nflfastR::calculate_expected_points(data) |> dplyr::select(season, yardline_100, td_prob, ep) |> knitr::kable(digits = 2)

Not surprisingly, offenses have become much more successful over time, with the kickoff touchback moving from the 20 to the 25 in 2016 providing an additional boost. Note that the `td_prob` in this example is the probability that the next score within the same half will be a touchdown scored by team with the ball, **not** the probability that the current drive will end in a touchdown (this is why the numbers are different from Example 4 above).

We could compare the most recent four years to the expectation for playing in a dome by inputting all the same things and changing the `roof` input:

``` {r ex6b}
data <- tibble::tibble(
  "season" = 2016:2019,
  "week" = 5,
  "home_team" = "SEA",
  "posteam" = "SEA",
  "roof" = "dome",
  "half_seconds_remaining" = 1800,
  "yardline_100" = c(rep(75, 4)),
  "down" = 1,
  "ydstogo" = 10,
  "posteam_timeouts_remaining" = 3,
  "defteam_timeouts_remaining" = 3
)
nflfastR::calculate_expected_points(data) |>
  dplyr::select(season, yardline_100, td_prob, ep) |>
  knitr::kable(digits = 2)

So for 2018 and 2019, 1st & 10 from a home team's own 25 yard line had higher EP in domes than at home, which is to be expected.

Example 5: Win probability calculator

We have also provided a calculator for working with the win probability models. Here is an example of how to use it, looking for how the win probability to begin the game depends on the pre-game spread.

While I have put in 'SEA' for home_team and posteam, this only matters for figuring out whether the team with the ball is the home team (there's no actual effect for given team; it would be the same no matter what team is supplied).

``` {r ex7} data <- tibble::tibble( "receive_2h_ko" = 0, "home_team" = "SEA", "posteam" = "SEA", "score_differential" = 0, "half_seconds_remaining" = 1800, "game_seconds_remaining" = 3600, "spread_line" = c(1, 3, 4, 7, 14), "down" = 1, "ydstogo" = 10, "yardline_100" = 75, "posteam_timeouts_remaining" = 3, "defteam_timeouts_remaining" = 3 ) nflfastR::calculate_win_probability(data) |> dplyr::select(spread_line, wp, vegas_wp) |> knitr::kable(digits = 2)

Not surprisingly, `vegas_wp` increases with the amount a team was coming into the game favored by.

## Example 6: Using the built-in database function

If you're comfortable using `dplyr` functions to manipulate and tidy data, you're ready to use a database. Why should you use a database?

* The provided function in `nflfastR` makes it extremely easy to build a database and keep it updated
* Play-by-play data over 20+ seasons takes up a lot of memory: working with a database allows you to only bring into memory what you actually need
* R makes it *extremely* easy to work with databases.

### Start: install and load packages

To start, we need to install the two packages required for this that aren't installed automatically when `nflfastR` installs: `DBI` and `RSQLite` (advanced users can use other types of databases, but this example will use SQLite). The `if` statements make sure the packages won't be updated if they are already installed:

``` {r eval = FALSE}
if (!require("DBI")) install.packages("DBI")
if (!require("RSQLite")) install.packages("RSQLite")

As with always, you only need to install these once. They don't need to be loaded to build the database because nflfastR knows how to use them, but we do need them later on when working with the database.

library(DBI)
library(RSQLite)

Build database

There's exactly one function in nflfastR that works with databases: update_db(). Some notes:

Let's say I just want to dump a database into the current working directory. Here we go!

``` {r create-db} nflfastR::update_db()

This created a database in the current directory called `pbp_db`.

Wait, that's it? That's it! What if it's partway through the season and you want to make sure all the new games are added to the database? What do you run? `update_db()`! (just make sure you're in the directory the database is saved in or you supply the right file path)

``` {r update-db}
nflfastR::update_db()

If it's partway through a season and you want to re-build a season to allow for data corrections from the NFL to propagate into your database, you can specify one season to be rebuilt:

``` {r update-db-yr} nflfastR::update_db(force_rebuild = 2020)

### Connect to database

Now we can make a connection to the database. This is the only part that will look a little bit foreign, but all you need to know is where your database is located. If it's in your current working directory, this will work:

``` {r}
connection <- DBI::dbConnect(RSQLite::SQLite(), "./pbp_db")
connection

It looks like nothing happened, but we now have a connection to the database. Now we're ready to do stuff. If you aren't familiar with databases, they're organized around tables. Here's how to see which tables are present in our database: 

DBI::dbListTables(connection)

Since we went with the defaults, there's a table called nflfastR_pbp. Another useful function is to see the fields (i.e., columns) in a table:

DBI::dbListFields(connection, "nflfastR_pbp") |>
  utils::head(10)

This is the same list as the list of columns in nflfastR play-by-play. Notice we had to supply the name of the table above ("nflfastR_pbp").

With all that out of the way, there's only a couple more things to learn. The main driver here is tbl, which helps get output with a specific table in a database:

pbp_db <- dplyr::tbl(connection, "nflfastR_pbp")

And now, everything will magically just "work": you can forget you're even working with a database!

pbp_db |>
  dplyr::group_by(season) |>
  dplyr::summarize(n = dplyr::n())
pbp_db |>
  dplyr::filter(rush == 1 | pass == 1, down <= 2, !is.na(epa), !is.na(posteam)) |>
  dplyr::group_by(pass) |>
  dplyr::summarize(mean_epa = mean(epa, na.rm = TRUE))

So far, everything has stayed in the database. If you want to bring a query into memory, just use collect() at the end:

russ <- pbp_db |>
  dplyr::filter(name == "R.Wilson" & posteam == "SEA") |>
  dplyr::select(desc, epa) |>
  dplyr::collect()
russ

So we've searched through about 1 million rows of data across 300+ columns and only brought about r round(nrow(russ), -1) rows and two columns into memory. Pretty neat! This is how I supply the data to the shiny apps on rbsdm.com without running out of memory on the server. Now there's only one more thing to remember. When you're finished doing what you need with the database:

DBI::dbDisconnect(connection)

For more details on using a database with nflfastR, see Thomas Mock's life-changing post here. More detailed information on dbplyr (the dplyr database back-end) are given in the second edition of Hadley Wickham's R for Data Science (2e).

Example 7: working with the expected yards after catch model

The variables in xyac are as follows:

Some other notes:

Let's create measures for EPA and first downs over expected in 2015:

``` {r ex9-xyac, warning = FALSE, message = FALSE} nflfastR::load_pbp(2015) |> dplyr::group_by(receiver, receiver_id, posteam) |> dplyr::mutate(tgt = sum(complete_pass + incomplete_pass)) |> dplyr::filter(tgt >= 50) |> dplyr::filter(complete_pass == 1, air_yards < yardline_100, !is.na(xyac_epa)) |> dplyr::summarize( epa_oe = mean(yac_epa - xyac_epa), actual_fd = mean(first_down), expected_fd = mean(xyac_fd), fd_oe = mean(first_down - xyac_fd), rec = dplyr::n() ) |> dplyr::ungroup() |> dplyr::select(receiver, posteam, actual_fd, expected_fd, fd_oe, epa_oe, rec) |> dplyr::slice_max(epa_oe, n = 10) |> knitr::kable(digits = 3)

The presence of so many running backs on this list suggests that even though it takes into account target depth and pass direction, the model doesn't do a great job capturing space. Alternatively, running backs might be better at generating yards after the catch since running with the football is their primary role.

## Example 8: Working with roster and position data

At long last, there's a way to merge the new play-by-play data with roster information. Use the function to get the rosters:

``` {r roster}
roster <- nflfastR::fast_scraper_roster(2019)

Now let's load play-by-play data from 2019: ``` {r roster_pbp_load} games_2019 <- nflfastR::load_pbp(2019)

Here is what the player IDs look like because `nflfastR` now automatically decodes IDs to look like the old format with GSIS IDs:

``` {r roster_pbp}
games_2019 |>
  dplyr::filter(rush == 1 | pass == 1, posteam == "SEA") |>
  dplyr::select(name, id)

Now we're ready to join to the roster data using these IDs: ``` {r decode_join} joined <- games_2019 |> dplyr::filter(!is.na(receiver_id)) |> dplyr::select(posteam, season, desc, receiver, receiver_id, epa) |> dplyr::left_join(roster, by = c("receiver_id" = "gsis_id"))

``` {r decode_table}
# the real work is done, this just makes a table and has it look nice
joined |>
  dplyr::filter(position %in% c("WR", "TE", "RB")) |>
  dplyr::group_by(receiver_id, receiver, position) |>
  dplyr::summarize(tot_epa = sum(epa), n = n()) |>
  dplyr::arrange(-tot_epa) |>
  dplyr::ungroup() |>
  dplyr::group_by(position) |>
  dplyr::mutate(position_rank = 1:n()) |>
  dplyr::filter(position_rank <= 5) |>
  dplyr::rename(Pos_Rank = position_rank, Player = receiver, Pos = position, Tgt = n, EPA = tot_epa) |>
  dplyr::select(Player, Pos, Pos_Rank, Tgt, EPA) |>
  knitr::kable(digits = 0)

Not surprisingly, all 5 of the top 5 WRs in terms of EPA added come in ahead of the top RB. Note that the number of targets won't match official stats because we're including plays with penalties.

Example 9: Replicating official stats

The columns like name, passer, fantasy etc are nflfastR-created columns that mimic "real" football: i.e., excluding plays with spikes, counting scrambles and sacks as pass plays, etc. But if you're trying to replicate official statistics -- perhaps for fantasy purposes -- use the *_player_name and *_player_id columns.

Let's try to replicate this page of passing leaders.

``` {r stats1} nflfastR::load_pbp(2020) |> dplyr::filter(season_type == "REG", complete_pass == 1 | incomplete_pass == 1 | interception == 1, !is.na(down)) |> dplyr::group_by(passer_player_name, posteam) |> dplyr::summarize( yards = sum(passing_yards, na.rm = T), tds = sum(touchdown == 1 & td_team == posteam), ints = sum(interception), att = dplyr::n() ) |> dplyr::arrange(-yards) |> utils::head(10) |> knitr::kable(digits = 0)

These match the official stats on NFL.com (note the filter for `season_type == "REG"` since official stats only count regular season games). Note that we're using `passing_yards` here because `yards_gained` is not equal to passing yards on plays with laterals.

While the above code works in this case, there are several special cases where it is nearly impossible to get official player stats from nflfastR play-by-play data. The reason for this is that the idea of nflfastR play-by-play data is a “tidy” data structure. In other words, the aim is to have one row per play in the data. This can lead to problems if, for example, there are several changes of possession per play (i.e. several fumbles) or if the ball is lateraled in a play. These are just two examples of “abnormal” plays that are not fully captured in a tidy data structure.
We have solved this problem with the function `calculate_stats()`. This function uses playstats of the raw play-by-play data before it is parsed into a tidy structure by nflfastR. 

This function has the following features:

- It determines stats in offense, defense, and special teams,
- either on player level or on team level,
- and can summarize them on season level (separately for regular season and post season) or on week level.

For more information see the function documentation of `calculate_stats()`. Again, **don't try to get an exact match with official stats based on nflfastR play-by-play data**. It usually works, but fails because of details that are unsolvable.

Now let's replicate the above table using `calculate_stats()`:

``` {r stats2}
s <- nflfastR::calculate_stats(
  seasons = 2020, 
  summary_level = "season",
  stat_type = "player",
  season_type = "REG"
)
s |>
  dplyr::slice_max(passing_yards, n = 10) |> 
  dplyr::select(player_name, recent_team, completions, attempts, passing_yards, passing_tds, passing_interceptions, attempts) |>
  knitr::kable(digits = 0)

Frequent issues

The drive column looks wacky

Use fixed_drive and fixed_drive_result instead. See [Example 2: Using Drive Information].

Why are there so many win probability columns?

vegas_wp and vegas_home_wp incorporate the pregame spread and are much better models.

Need more help?

Please ask in the nflverse Discord server.



nflverse/nflfastR documentation built on April 17, 2025, 9:34 p.m.

R Package Documentation

rdrr.io home R language documentation Run R code online

Browse R Packages

CRAN packages Bioconductor packages R-Forge packages GitHub packages

We want your feedback!

Note that we can't provide technical support on individual packages. You should contact the package authors for that.
GitHub issue tracker
ian@mutexlabs.com
Personal blog

 
Embedding an R snippet on your website

Add the following code to your website.

For more information on customizing the embed code, read Embedding Snippets.

Close