The geography of male tennis

1
2
3
4
5
6
7
import pandas as pd
import glob
import matplotlib.pyplot as plt
import datetime, sys
import numpy as np

%matplotlib inline

1
2
3
4
5
6
def parse(t):
    string_ = str(t)
    try:
        return datetime.date(int(string_[:4]), int(string_[4:6]), int(string_[6:]))
    except:
        return datetime.date(1900,1,1)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
def readAllFiles():
    allFiles = glob.iglob("data/atp_rankings_" + "*.csv")
    ranks = pd.DataFrame()
    list_ = list()
    for filen in allFiles:
        df = pd.read_csv(filen,
                         index_col=None,
                         header=None,
                         parse_dates=[0],
                         date_parser=lambda t:parse(t),
                         names = ["ranking_date", "ranking", "player_id", "ranking_points"])
        list_.append(df)
    ranks = pd.concat(list_)
    return ranks

1
2
3
4
5
6
7
def readPlayers():
    return pd.read_csv("data/atp_players.csv",
                       index_col=None,
                       header=None,
                       parse_dates=[4],
                       names = ["player_id", "first_name", "last_name", "hand", "birth_date", "country_code"],
                       date_parser=lambda t:parse(t))

1
2
3
4
5
6
7
8
# merge rankings with players but keep only top 100

ranks = readAllFiles()
ranks = ranks[(ranks['ranking']<100)]
ranks['player_id'] = ranks['player_id'].apply(lambda row: int(row))
players = readPlayers()
players = ranks.merge(players,right_on="player_id",left_on="player_id")
players.head()

1
2
3
4
5
6
7
# we are interested in ranking date, country code and counting the number of players. 
# To keep the data cleaner get only countries with at least 4 players in top 100

aggregate = players[["ranking_date", "country_code"]].groupby(["ranking_date","country_code"]).size().to_frame()
aggregate = pd.DataFrame(aggregate.to_records()) # transform the series to df
aggregate_least_4 = aggregate[(aggregate["0"] > 3)]
aggregate_least_10 = aggregate[(aggregate["0"] > 9)]

1
2
3
4
5
6
# countries with at least 10 players in top 10 at a given time.
# we can see that the US has dominated until 1997, when Spain started to take the lead.
# Since 2000, Spain and France had the most players in top 100. 

pivoted = pd.pivot_table(aggregate_least_10, values='0', columns='country_code', index="ranking_date")
pivoted.plot(figsize=(20,10))

1
2
3
4
5
# Let's see that figure again without the outlier USA

aggregate_no_usa = aggregate[(aggregate["country_code"] != "USA") & (aggregate["0"] > 9)]
pivoted = pd.pivot_table(aggregate_no_usa, values='0', columns='country_code', index="ranking_date")
pivoted.plot(figsize=(20,10))

1
2
3
4
5
# after 2000 Spain and France dominate with an average of 15 players in top 100

aggregate_after_2000 = aggregate[(aggregate["ranking_date"] > "2000-01-01" ) & (aggregate["0"] > 9)]
pivoted = pd.pivot_table(aggregate_after_2000, values='0', columns='country_code', index="ranking_date")
pivoted.plot(figsize=(20,10))

1
2
3
4
5
6
7
8
# let's see how it fairs by continent

import json

with open('iso3.json') as data_file:    
    countries = json.load(data_file)
with open('continent.json') as data_file:    
    continents = json.load(data_file)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
ioc_countries = {}
for country in countries:
    if country["ioc"]:
        ioc_countries[country["ioc"]] = country["alpha2"]
ioc_countries["YUG"] = "MK"
aggregate_continents = aggregate.copy()
ioc_countries
aggregate_continents
aggregate_continents["country_code"] = aggregate_continents["country_code"].apply(lambda row: continents[ioc_countries[row]])

aggregate_continents = aggregate_continents[["ranking_date", "country_code", "0"]].groupby(["ranking_date", "country_code"]).sum()
aggregate_continents = pd.DataFrame(aggregate_continents.to_records()) # transform the series to df
pivoted = pd.pivot_table(aggregate_continents, values='0', columns='country_code', index="ranking_date")
pivoted.plot(figsize=(20,10))