Lack of compatibility with Wyscout's Soccer-logs open dataset about socceraction

[probberechts]

@agdhruv I've added a notebook which downloads and converts the Wyscout dataset to the SPADL format in the wyscout_support branch. If you use it, you should be aware that there are still some bugs in the wyscout converter (see other issues).

[TomDecroos]

Hi kkoripl,

Thanks for flagging this. The core problem seems indeed to be that the soccerlog event data made available by Papalardo et al. is in a different format (i.e., matches.json, teams.json, event.json, players.json, ...) than how Wyscout usually provides their data through one self-contained file per match (match-1.json, match-2.json, match-3.json, ...)

Seeing how the soccerlog data set is probably going to be pretty important for many future soccer analytics researchers, I will try to add a converter to socceraction as soon as possible.

[kkoripl]

I came up with something like this and it seems working for me. Of course, there should be possibility to get seperate paths for matches, events, (you name it) files as arguments of function, but as I use only particular set in particular place that's how I did it.

def jsonfiles_to_h5(country, h5_file):
  with pd.HDFStore(h5_file) as store:
    events = load_events_as_df(country)
    match_ids = events[MATCH_ID].unique()
    for match_id in match_ids:
      store[f"events/match_{match_id}"] = events[events[MATCH_ID]==match_id]

    store["matches"] = load_matches_as_df(country).drop_duplicates(WY_ID)
    store["players"] = load_players_as_df().drop_duplicates(WY_ID)

    teams = load_teams_as_df()
    teams = teams[teams['type'] == 'club']
    teams = teams[teams['area'].apply(lambda area: area['name'] == country)]
    store["teams"] = teams

[agdhruv]

Seeing how the soccerlog data set is probably going to be pretty important for many future soccer analytics researchers, I will try to add a converter to socceraction as soon as possible.

Hi @TomDecroos is there any update on this? I'd like to use the Wyscout open dataset to convert data to SPADL.

@kkoripl Where can I find the code that worked for you?