PandasBasketball

PandasBasketball is a small module intended to scrape data from basketball-reference and convert it to useful pandas data structures, such as data frames, for future analytical purposes. The use of jupyter notebooks is encouraged.

Installation

pip install PandasBasketball

After installation you can then import it to your environment like this:

from PandasBasketball import pandasbasketball as pb

Requirements

Please make sure you meet the following rquirements:

• Python 3.6+
• requests
• pandas
• Beautiful Soup 4

All the requirements can easily be met with the installation of the Anaconda distribution.

Usage

:basketball: Players

Inside a player's page on the basketball-reference website you can find several tables, and most of these tables can be obtained as a pandas data frame by calling get_player(player, stat). The 'player' refers to the name of the html file used by basketball-reference inside the url, and the 'stat' means the type of table.

The currently supported tables are:

• Per Game (per_game)
• Totals (totals)
• Per 36 Minutes (per_minute)
• Per 100 Poss (per_poss)
• Advanced (advanced)
• Playoffs Per Game (playoffspergame)
• Playoffs Totals (playoffs_totals)
• Playoffs Per 36 Minutes (playoffsperminute)
• Playoffs Per 100 Poss (playoffsperposs)
• Playoffs Advanced (playoffs_advanced)

The rest of the tables will come in the future.

Example

To get the 'Per Game' table for LeBron James you would do something like this:

df = pb.getplayer("jamesle01", "pergame")

Optional Arguments

The get_player() method supports two optional arguments:

• numeric - returns the data frame with its columns alreay converted to numeric
• s_index - returns the data frame with its column 'Season' as the index

Both are set to False by deault.

Considerations

• The resulting data frame does not include the table's footer.
• The resulting data frame will have the same column names as the table's header but it will not have a set index. To set the 'Season' column as index set the argument s_index to True.
• The columns will be of type 'object', so in order to perform arithmetic functions on them you will need to convert them to numeric. You can do something like this:

lbjpg = pb.getplayer("jamesle01", "per_game")
lbjpg[lbjpg.columns] = lbjpg[lbjpg.columns].apply(pd.to_numeric, errors="ignore")

Or you cant set the optional argument numeric to True.

:basketball: Player Game Logs

You can get all of a player's games in a season by calling getplayergamelog(player, season). The season argument must be the last year in which the season took place.

Example

To get all of Kawhi Leonard's games during the 2017-2018 season:

df = pb.getplayergamelog("leonaka01", "2018")

Optional Arguments

The function getplayergamelog supports one optional argument:

• playoffs - returns only the playoffs games if set to True

Set to False by default.

Considerations

• The resulting data frame will use the 'Rk' column as its index
• The data frame does not include those rows which are just the header again
• If the player missed a game, the row will be filled with blanks ("")

:basketball: Last n days

Get a data frame with all the season's available players stats over the last n days by calling getndays(days).

Example

df = pb.getndays(10)

Optional arguments

getndays supports one optional argument:

• player - returns a pandas series with the stats of the specifed player

player is set to all by default.

Considerations

• The resulting data frame will have the column 'Players' as its index by default
• The data frame is in descending order by GmSc

:basketball: Teams

You can call a team's seasons table with get_team(name). The argument name is the team's three-letter abbreviation (e.g. OKC, MAV).

Example

To get OKC's table:

df = pb.get_team("OKC")

:basketball: Generate player code

Baskteball-reference uses a special code to build each player's unique html page. As of now, almost all functions in PandasBasketball expect that code to get the stats for the specified athlete. If you don't want to copy and paste the code from the URL into the function you can try calling pb.generate_code(player).

Note: this is not fully tested, so it is possible to get an incorrect code.

Example

To get the player code for LeBronJames:

pb.generate_code("LeBron James")

This will output 'jamesle01'

Using it with other functions:

df = pb.getplayer(pb.generatecode("Donovan Mitchell"), "per_game")

Future

• Make this project pip-installable
• Add support for the rest of tables on a player's page
• Implement function to obtain team stats per season
• Implement function to get the last n days stats
• Implement function to obtain game results by date

Contributions & Known Issues

If you notice an issue or want to contribute open an issue over at the issues section.