Griffis-Soccer-Analysis

Overview

This package includes several of the functions I use to analyze football data. More functions will be added as I get everything situated.

Currently, I just have functions to generated style/profile similarities between players, teams, and leagues. As this is the first time I'm publishing a package, I needed to start small.

Similarity Scores

Similarity scores are purely based on the style, not quality of a player, league, or team. They are generated from over 35 variales (all z-scores so they are the same scale) and use Cosine Similarity. This is a pretty standard method, and StatsBomb uses Cosine Similarity as one method for their similar player search.

Installation

To install the package, you can type or copy/paste this into you command line:

pip install git+https://github.com/griffisben/griffis_soccer_analysis.git

If you want to update the package, you can first uninstall the old package by entering this line into your command line:

pip uninstall griffis_soccer_analysis

Here is an example of how to load the package.

from griffis_soccer_analysis.fbref_code import *
from griffis_soccer_analysis.similarity import *
from griffis_soccer_analysis.whoscored_match_report import *

Whoscored Match Report Example

Download a csv of events for a given match on the Detailed Tournaments section of Whoscored, and make a post-match report image

whoscored_match_report(
    # The URL of the Whoscored match (detailed tournamnets only)
    url = 'https://www.whoscored.com/Matches/1735525/Live/Belgium-Jupiler-Pro-League-2023-2024-Union-St-Gilloise-Eupen',
    # The match name
    match = 'Union St. Gilloise 4-1 Eupen',
    
    # Home team name
    team_h = 'Union St. Gilloise',
    # Whoscored's ID for the home team (can find in the html slug of the team's page)
    teamId_h = 2647,
    
    # Away team name
    team_a = 'Eupen',
    # Away team ID
    teamId_a = 2166,
    
    # League name
    lg = 'Belgian Pro League',
    # Game date
    date = 'Oct 20, 2023',
    # Your signature
    sig = '@BeGriffis',
    # Language for the image (English, Spanish, Portuguese, French)
    language = 'English',
    
    # Location of your Chromedriver
    chrome_driver_loc = r"C:\Users\Ben\chromedriver.exe",
    # Where you want the data to download to
    data_download_loc = "C:/users/ben/downloads",

    # Where you want the final image to export to
    img_save_loc = "C:/users/ben/downloads",
    # Path to home team's image
    home_img = "C:/users/ben/club images/Union St. Gilloise.png",
    # Path to away team's image
    away_img = "C:/users/ben/club images/Eupen.png"
)

FBRef Code Examples

Download the latest FBRef data (via Opta) for UEFA top 5 players (note, will only work for the current, ongoing season)

scrape_fbref_t5_leagues_players(season = '2023-2024')

Download the latest FBRef data (via Opta) for all other leagues FBRef has advanced data for (note, will only work for the current, ongoing season)

comps = ['2. Bundesliga',
         'Ligue 2',
         'Serie B',
         'La Liga 2',
         'Belgian Pro League',
         'Argentine Primera División',
         'Argentina Copa de la Liga'
         'Liga MX',
         'MLS',
         'Brasileirão',
         'Eredivisie',
         'Primeira Liga',
         'Championship',
]
ssns = ['2023-2024',
        '2023-2024',
        '2023-2024',
        '2023-2024',
        '2023-2024',
        '2023',
        '2023',
        '2023-2024',
        '2023',
        '2023',
        '2023-2024',
        '2023-2024',
        '2023-2024',
       ]

scrape_fbref_next12_leagues_players(comps = comps, seasons = ssns)

Combine all FBRef scraped leagues into one file

combine_t5_next12_fbref(t5_season = '2023-2024')

Make a quick scouting card of 12 key stats for a player

fbref_scout_report(season = '2023-2024',
            program = 'mf',   # gk, df, mf, fw (each has 12 selected metrics for the position)
            player_pos = 'MF',  # GK, DF, MF, FW (FBRef positions)
            playerPrompt = 'Lucas Besozzi',
            SquadPrompt = '',
            minutesPlayed = 700,
            compP = 'Argentine Primera División',
            saveinput = 'n',
            signature = '@BeGriffis',
            data_date = 'Data as of 9/29/23',
            fbref_file_path = 'C:/Users/Ben/From Mac/Python/'
           )

Similarity Examples

For a Juypter Notebook file with example code & output and more info on each variable, please see this file

See available leagues

available_leagues()

See available teams in a league

teams_in_league(
    league = "A-League Men 22-23"
)

See available players in a team

available_players(
    league = "A-League Men 22-23",
    team = 'Western Sydney Wanderers'
)

League Similarities

# This is how to grab all outputs
df, info, dist_fig = league_similarity(
    league = "A-League Men",
    season = "22-23",
    nlgs = 15
)
# This is one way to print the information and then the similarity dataframe
for i in range(len(info)):
    print(info[i])
df

Team similarities

# This is how to grab all outputs
df, info, dist_fig = team_similarity(
    team = "Western Sydney Wanderers",
    league = "A-League Men",
    season = "22-23",
    nlgs = 15
)
# This is one way to print the information and then the similarity dataframe
for i in range(len(info)):
    print(info[i])
df

Player similarities

"""
CHOICES FOR compare_league VARIABLE

All
UEFA T5
UEFA Next 10
UEFA Next 20
UEFA T5 2nd Tiers
8 AFC
CONMEBOL Top 4
Argentina & Brazil
Scandinavia (note: this only includes top tiers from Denmark, Norway, Sweden, and Finland)

Or, make your own by writing each league-season name (can find with available_leagues()) separated by a space & comma:
    'MLS 2023, Liga MX 23-24'
    'Premier League 23-24, Championship 23-24, League One 23-24, League Two 23-24'

"""

df, info, dist_fig = player_similarity(
    player = "E. Spertsyan (23, Krasnodar, Russian Premier League 23-24)",
    position = "CM",
    nplayers = 20,
    compare_leagues = 'All',
    min_age = 16,   # Minimum age of similar players (always still inside the top 5%)
    max_age = 40,   # Maximum age of similar players (always still inside the top 5%)
    similar_lg_team = False,
    mean_sim = False,
)

# This is one way to print the information and then the similarity dataframe
for i in range(len(info)):
    print(info[i])
df

typewind / griffis_soccer_analysis Goto Github PK