Soccerdata

New features

Add support for scraping World Cup data

The World Cup was added to the default available leagues for the WhoScored and FBref readers. Other tournaments can be added by modifying the `league_dict.json` config file.

python
from soccerdata import WhoScored, FBref

ws = WhoScored(leagues="INT-World Cup", seasons="2022")
fb = FBref(leagues="INT-World Cup", seasons="2022")


Changes
- The WhoScored reader now uses the non-headless mode by default. Scraping in headless mode typically results in getting blocked quickly. The old behaviour can be recovered by initializing the reader as `WhoScored(..., headless=True)`.

Fixes
- The WhoScored reader can now deal with an empty match schedule, which can occur before the start of a season or tournament round.

New features

Faster scraping of Big 5 leagues stats (by andrewRowlinson)

FBref has [pages for the big five European leagues](https://fbref.com/en/comps/Big5/2020-2021/2020-2021-Big-5-European-Leagues-Stats) that allow you to more efficiently get team and player data from multiple leagues. This commit adds a special "Big 5 European Leagues Combined" league option to get data from these pages.

python
import soccerdata as sd
fbref = sd.FBref(leagues="Big 5 European Leagues Combined", seasons="20-21")
team_season_stats = fbref.read_team_season_stats(stat_type="standard")
player_season_stats = fbref.read_player_season_stats(stat_type="standard")

New features

FBref

Faster scraping of player season stats (69)

Previously, the `fbref.read_team_season_stats` method visited the page of each individual team in a league to obtain stats for players in a league. FBRef now has a single page for each league/season where player stats can be obtained for each player in the league (e.g., https://fbref.com/en/comps/9/stats/Premier-League-Stats). Due to this change the `fbref.read_team_season_stats(...)` method now uses 15-20x less requests, leading to a large speed-up.

Support retrieving "Opponent Stats" (78)

A "opponent_stats" flag was added to the `fbref.read_season_stats(...)` function, which enables retrieving the "Opponent Stats" table of a team.

Always group "MP" under "Playing Time" (79)

FBRef is inconsistent in how it displays the "MP" (Matches Played) column. For some seasons, it is displayed as a separate category, while it is grouped under "Playing Time" for other seasons. This results in a column with NaN values when two seasons are merged. Therefore, the "MP" column is now always put under "Playing Time".

Docs

Add docs for specifying custom proxy (83)

Not all Tor distribution use the same default port of 9050. The docs now describe how to configure a custom port.

Breaking Changes

- Several columns were renamed, added and droped in the output dataframes to increase uniformity between
datasources.

New features

WhoScored

The WhoScored reader can now return event data in various output formats. The following formats are supported:

- A dataframe with all events.
- A dict with the original unformatted WhoScored JSON.
- A dataframe with the [SPADL representation](https://socceraction.readthedocs.io/en/latest/documentation/SPADL.html#spadl)
of the original events.
- A dataframe with the [Atomic-SPADL representation](https://socceraction.readthedocs.io/en/latest/documentation/SPADL.html#atomic-spadl)
of the original events.
- A ``socceration.data.opta.OptaLoader`` instance.
- No data. This is useful for caching data.

See https://soccerdata.readthedocs.io/en/latest/datasources/WhoScored.html for examples.

Breaking Changes

- The `use_tor` parameter was replaced by a `use_proxy='tor'` parameter in all readers

New features

- You can specify a custom proxy using the `use_proxy` parameter for all readers.

python

Bugfixes

WhoScored
- The summary tab is now used as a backup for retrieving the schedule when the fixtures tab is empty. This often occurs for multi-stage tournaments. (15)
- Fixed incorrect resolver rules for the Tor proxy. (23)

MatchHistory
- Football-data.co.uk switched from http to https only.

Docs

- Added example notebooks for reading data from each supported data source.