L2M

KO | EN

gitlite — search

#hacktoberfest #react #llm #golang #android #machine-learning #deep-learning #linux #docker #macos #kubernetes #awesome-list #pytorch #open-source

(★ 25)

Last two minute report data from the NBA

L2M Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# L2M

[this is the csv of the data that you most likely
want](1-tidy/L2M/L2M_stats_nba.csv)

## Articles

- [How To Extract The Last Two Minute
  Reports](2-eda/2-how-to-last-two-minutes)
- [Updating The Last Two Minute Reports: data through the 2019 NBA
  Finals](2-eda/2-through-2019-finals)

## Description

This repository maintains the [Last Two Minute (L2M)
reports](https://official.nba.com/2021-22-nba-officiating-last-two-minute-reports/)
that the NBA releases for some NBA games depending on how close the game
was in the final two minutes. The clean, processed version of the data
can be found in
[1-tidy/L2M/L2M_stats_nba.csv](1-tidy/L2M/L2M_stats_nba.csv) while the
corresponding [`R`](https://www.r-project.org/) code to create this data
is scattered across the [0-data](0-data) and [1-tidy](1-tidy) folders.
The [0-data/L2M/archived-pdf](0-data/L2M/archived-pdf) folder hosts the
raw pdf data from the games before the NBA utilized an API for storing
L2M information. The
[0-data/official_nba/raw/l2m-api](0-data/official_nba/raw/l2m-api)
folder hosts individual game files from the API.

![](https://raw.githubusercontent.com/atlhawksfanatic/L2M/master/README_files/figure-gfm/calls-1.png)

| Season | Games | Grades per period | Calls per period | IC per period | CC Percentage | INC per period | Bad Calls Percentage | CNC per period |
|:--:|:--:|:--:|:--:|:--:|:--:|:--:|:--:|:--:|
| 2015 | 139 | 11.27 | 4.82 | 0.15 | 97% | 1.37 | 25% | 5.08 |
| 2016 | 439 | 12.88 | 4.77 | 0.21 | 96% | 1.66 | 29% | 6.45 |
| 2017 | 428 | 15.00 | 4.15 | 0.08 | 98% | 2.49 | 39% | 8.36 |
| 2018 | 475 | 19.67 | 4.14 | 0.12 | 97% | 2.46 | 39% | 13.07 |
| 2019 | 455 | 21.28 | 3.59 | 0.13 | 96% | 2.55 | 44% | 15.14 |
| 2020 | 389 | 18.05 | 3.79 | 0.15 | 96% | 1.08 | 25% | 13.18 |
| 2021 | 405 | 17.18 | 3.85 | 0.19 | 95% | 0.87 | 22% | 12.47 |
| 2022 | 441 | 17.53 | 4.18 | 0.19 | 95% | 1.18 | 26% | 12.18 |
| 2023 | 474 | 20.80 | 4.41 | 0.16 | 96% | 1.15 | 24% | 15.24 |
| 2024 | 411 | 17.50 | 4.34 | 0.17 | 96% | 0.92 | 21% | 12.24 |
| 2025 | 423 | 14.96 | 4.38 | 0.14 | 97% | 0.55 | 14% | 10.02 |

All games with L2M Call Accuracy updated through 2025-05-13

| Season | Playoffs | Games | Grades per period | Calls per period | IC per period | CC Percentage | INC per period | Bad Calls Percentage | CNC per period |
|:--:|:--:|:--:|:--:|:--:|:--:|:--:|:--:|:--:|:--:|
| 2015 | FALSE | 113 | 10.74 | 4.84 | 0.15 | 97% | 1.36 | 24% | 4.54 |
| 2015 | TRUE | 26 | 13.40 | 4.74 | 0.14 | 97% | 1.43 | 25% | 7.23 |
| 2016 | FALSE | 410 | 12.58 | 4.81 | 0.21 | 96% | 1.58 | 28% | 6.18 |
| 2016 | TRUE | 29 | 17.26 | 4.12 | 0.18 | 96% | 2.82 | 43% | 10.32 |
| 2017 | FALSE | 403 | 14.81 | 4.14 | 0.08 | 98% | 2.46 | 38% | 8.21 |
| 2017 | TRUE | 25 | 18.29 | 4.18 | 0.04 | 99% | 3.07 | 43% | 11.04 |
| 2018 | FALSE | 452 | 19.79 | 4.14 | 0.11 | 97% | 2.47 | 39% | 13.18 |
| 2018 | TRUE | 23 | 17.35 | 4.12 | 0.23 | 94% | 2.42 | 41% | 10.81 |
| 2019 | FALSE | 424 | 21.18 | 3.61 | 0.13 | 96% | 2.57 | 44% | 14.99 |
| 2019 | TRUE | 31 | 22.67 | 3.33 | 0.15 | 95% | 2.21 | 43% | 17.13 |
| 2020 | FALSE | 364 | 17.88 | 3.80 | 0.15 | 96% | 1.08 | 25% | 13.00 |
| 2020 | TRUE | 25 | 20.50 | 3.67 | 0.13 | 96% | 1.03 | 25% | 15.80 |
| 2021 | FALSE | 379 | 17.13 | 3.82 | 0.18 | 95% | 0.84 | 22% | 12.48 |
| 2021 | TRUE | 26 | 17.87 | 4.26 | 0.29 | 93% | 1.26 | 28% | 12.35 |
| 2022 | FALSE | 412 | 17.31 | 4.16 | 0.19 | 95% | 1.17 | 25% | 11.98 |
| 2022 | TRUE | 29 | 21.17 | 4.47 | 0.23 | 95% | 1.33 | 27% | 15.37 |
| 2023 | FALSE | 442 | 20.63 | 4.40 | 0.16 | 96% | 1.15 | 24% | 15.09 |
| 2023 | TRUE | 32 | 23.17 | 4.58 | 0.14 | 97% | 1.19 | 23% | 17.39 |
| 2024 | FALSE | 385 | 17.28 | 4.30 | 0.16 | 96% | 0.90 | 20% | 12.09 |
| 2024 | TRUE | 26 | 20.83 | 4.97 | 0.24 | 95% | 1.21 | 23% | 14.66 |
| 2025 | FALSE | 400 | 14.89 | 4.42 | 0.14 | 97% | 0.55 | 14% | 9.92 |
| 2025 | TRUE | 23 | 16.07 | 3.74 | 0.11 | 97% | 0.63 | 17% | 11.70 |

L2M Call Accuracy updated through 2025-05-13

### Organization

The process for compiling the L2M dataset is to:

1.  Download the raw data, either from the older PDF system or through
    their API:
    - [Game IDs](0-data/stats_nba) from when the L2M began in 2014.
      Downloaded with
      [0-stats-nba-game-ids.R](0-data/0-stats-nba-game-ids.R) and
      [0-stats-nba-game-ids-pre-2015.R](0-data/0-stats-nba-game-ids-pre-2015.R).
      These provide the list of games that potentially could contain an
      L2M.
    - [Archived PDFs](0-data/L2M/archived-pdf) where the
      [0-L2M-download-archive-updated.R](0-data/0-L2M-download-archive-updated.R)
      file downloads all the PDFs then
      [0-L2M-pdftools-archived-pdf.R](0-data/0-L2M-pdftools-archived-pdf.R)
      will parse through the information with the
      [pdftools](https://github.com/ropensci/pdftools) package. Output
      can be found
      [pdftools_L2M_archive_all.csv](0-data/L2M/archived-pdf/pdftools_L2M_archive_all.csv).
    - [Official API](0-data/official_nba) contains data extracted from
      the API of official.nba.com. The
      [0-official-nba-l2m.R](0-data/0-official-nba-l2m.R) file accesses
      the API based on the previous day’s date and checking the games
      that have been played based on Game IDs. Output can be found
      [official_nba_l2m_api.csv](0-data/official_nba/official_nba_l2m_api.csv).
2.  Tidy up the raw L2M data
    - [L2M Folder](1-tidy/L2M/) contains tidied data. First pass is
      through [1-L2M-raw-stats-nba.R](1-tidy/1-L2M-raw-stats-nba.R)
      which combines the PDF L2M reports with the L2M reports accessed
      through the API. Resulting output is
      [L2M_raw_api.csv](1-tidy/L2M/L2M_raw_api.csv).
3.  Download box scores for games from
    [stats.nba.com](https://stats.nba.com/) game information to
    supplement reports and to match up committing/disadvantaged players.
    - [NBA Stats](0-data/stats_nba/) for both [box
      scores](0-data/0-stats-nba-box-data.R) and [play by
      play](0-data/0-stats-nba-pbp-data.R) data. Both utilize the NBA’s
      API for Stats, although there is no current use for the play by
      play data.
4.  Combine L2M reports with box score information
    - [Raw version](1-tid1-L2M-raw-stats-nba.R) which does not include
      box score info, the [csv file](1-tidy/L2M_raw_api.csv)
    - [Final version](1-tidy/1-L2M-stats-nba.R) which includes box score
      info, the [csv file](1-tidy/L2M/L2M_stats_nba.csv)

### Variables

The final [output](1-tidy/L2M/L2M_stats_nba.csv) includes the following
variables:

- `period`: period at which point the play occurred
- `time`: time remaining in the period when play occurred
- `call_type`: raw call type variable in L2M
- `committing`: committing player or team of the graded infraction in
  L2M, may be blank
- `disadvantaged`: disadvantaged player or team of the graded infraction
  in L2M, may be blank
- `decision`: judgment of L2M for the call, could be CC, CNC, IC, INC,
  or blank where CC = Correct Call, CNC = Correct Non-Call, IC =
  Incorrect Call, INC = Incorrect Non-Call and blank = not detectable
  without technology
- `comments`: L2M comments on the play
- `game_details`: game details of L2M, only relevant for PDFs
- `page`: page of L2M for pdf, not applicable for games accessed through
  API
- `file`: name of L2M file, will be the saved .csv file for API data
- `game_date`: game date according to L2M report header
- `away_team`: away team nickname
- `home_team`: home team nickname
- `date`: date of game in YYYY-MM-DD format
- `call`: first part of call_type, this is before the colon in call_type
- `type`: second part of call_type, this is after the colon in call_type
- `home`: home team abbreviation
- `away`: away team abbreviation
- `gid`: NBA game ID
- `gcode`: NBA game code
- `home_score`: home final score from the L2M reports, incomplete
  variable
- `away_score`: away final score from the L2M reports, incomplete
  variable
- `national_tv`: national TV network for the game, “no” indicates none
- `networks`: shorthand TV network for the game, NBALP indicates league
  pass
- `game_id`: nba.com url for L2M game, last part of
  “<https://official.nba.com/l2m/L2MReport.html/>”, not applicable for
  PDFs
- `PCTime`: from official API
- `ImposibleIndicator`: from official API
- `Difficulty`: from official API
- `VideolLink`: from official API
- `Qualifier`: from official API
- `posID`: from official API
- `posStart`: from official API
- `posEnd`: from official API
- `posTeamId`: from official API
- `teamIdInFavor`: from official API
- `errorInFavor`: from official API
- `imgChart`: from official API
- `GameId`: from official API
- `GameDate`: from official API
- `HomeTeamId`: from official API
- `AwayTeamId`: from official API
- `L2M_Comments`: from official API
- `season`: NBA season for which the graded play was a part of,
  convention is to use the last year of the NBA season so 2015 refers to
  the 2014-15 Season
- `playoff`: dummy variable equal to `TRUE` if the game occurred in the
  playoffs
- `GAME_ID`: NBA API game ID, same as `gid`
- `OFFICIAL_1`: name of first referee for game
- `OFFICIAL_2`: name of second referee for game
- `OFFICIAL_3`: name of third referee for game
- `OFFICIAL_ID_1`: ID of first referee for game
- `OFFICIAL_ID_2`: ID of second referee for game
- `OFFICIAL_ID_3`: ID of third referee for game
- `OFFICIAL_4`: name of fourth referee for game, typically only for
  playoff games
- `OFFICIAL_ID_4`: ID of fourth referee for game, typically only for
  playoff games
- `ATTENDANCE`: attendance for the game
- `committing_min`: total minutes played by player committing action
  (note, may be NA because the player did not play and likely an input
  error from NBA on L2M)
- `committing_team`: team for committing player
- `committing_side`: home/away for committing player
- `disadvantaged_min`: total minutes played by player disadvantaged by
  action
- `disadvantaged_team`: team for disadvantaged player
- `disadvantaged_side`: home/away for disadvantaged player
- `type2`: consistent format for type of infraction
- `time_min`: minutes remaining in period
- `time_sec`: seconds remaining in period
- `time2`: fractional minutes left (ie 1.9 would be one minute and 54
  seconds)

And an overview of the changes in L2M reporting is provided in
[2-eda/2-through-2019-finals](2-eda/2-through-2019-finals). Further,
there is a bit of a how-to for downloading and extracting data on the
L2M that is provided in
[2-eda/2-how-to-last-two-minutes](2-eda/2-how-to-last-two-minutes)

L2M

L2M Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

L2M

L2M Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README