这篇文章参考自 这个notebook ，无意中看到的，觉得不错，就用Julia重新写了一遍，我们用到的库有
- Plots
- DataFrames
- CSV
- StatsPlots
- StatsBase
using Plots, CSV, DataFrames
plotly()
Plots.default(show=true)histWorldCup = CSV.read("data/world-cup/WorldCupsSummary.csv", DataFrame)
该数据表包含了从1930到2018年间共21届世界杯赛事的汇总信息（1942、1946两年因二战停办），从HostContinent列中我们可以得知2002年韩日世界杯是世界杯首次落户亚洲，2010年南非世界杯是首次落户非洲。
从表格中我们还可以看到”Germany FR(联邦德国)”的信息，因此有必要对数据进行清洗，接下来我们进行数据预处理：
let columns = names(histWorldCup)
for column in columns
replace!(histWorldCup[!, column], "Germany FR" => "Germany")
end
end完成基础的数据预处理工作之后，我们来分析如下问题：
- 历年现场观众人数变化趋势
- 参赛队伍数变化趋势
- 历年进球数变化趋势
- 历史上夺冠次数最多的国家队是哪支?
- 夺冠队伍所在洲分析
- 哪些国家队能经常打入决赛/半决赛?
- 进入决赛的队伍夺冠概率是多少?
- 东道主（主办国）进入决赛/半决赛大吗?
let title = "Attendance Number"
x = histWorldCup.Attendance
y = histWorldCup.Year
xticks = [500000, 1000000, 1500000, 2000000, 250000000, 3000000, 35000000, 4000000]
yticks = histWorldCup.Year
scatter(x, y, xticks=(:all, xticks), yticks=yticks, title=title) |> display
end
可以看到，世界杯的现场观众总数整体呈上升趋势（1942、1946因二战停办两届），观众总数最多的一届是1994年的美国世界杯
let title = "QUalifiedTeams Number"
x = histWorldCup.QualifiedTeams
y = histWorldCup.Year
xticks = [0, 16, 24, 32, 48]
yticks = histWorldCup.Year
scatter(x, y, xticks=(:all, xticks), yticks=yticks, title=title) |> display
end
可以看到，世界杯参赛队伍从13支扩展到现在的32支。期间经历了两次队伍扩充，分别是1982年由16支队伍扩充到24支，以及1998年从24支扩充到32支。在下一届世界杯（2026美加墨世界杯），FIFA决定将参赛队伍扩充到48支。有了参赛队伍数的分析后，我们再来看看历届世界杯进球总数：
let title = "Goals Number"
x = histWorldCup.GoalsScored
y = histWorldCup.Year
xticks = [50, 75, 100, 125, 150, 175, 200]
yticks = histWorldCup.Year
scatter(x, y, xticks=(:all, xticks), yticks=yticks, title=title) |> display
end
可以看到，随着世界杯参赛队伍的增多，比赛总进球数也在增加。目前单届世界杯总进球数均没有超过175球，我们可以看看2022卡塔尔世界杯结束后能否创造进球数记录。
分析完总体趋势后，我们再来看看各支队伍夺冠情况/进入四强的情况
# 夺冠次数分析
import StatsBase: countmap
let title = "Champion Number Statistic"
x = histWorldCup.Winner
cmap = countmap(x)
xs = convert(Vector{String}, collect(keys(cmap)))
ys = collect(values(cmap))
bar(xs, ys, title=title) |> display
end
可以看到巴西是夺冠次数最多的国家，无愧足球王国的称号。德国、意大利两个足球紧随其后，分别是4次夺冠。我们再来看看各国家队进入半决赛（4强）和决赛的次数统计。
cmap1 = countmap(histWorldCup.Winner)
cmap2 = countmap(histWorldCup.Second)
cmap3 = countmap(histWorldCup.Third)
cmap4 = countmap(histWorldCup.Fourth)
countries = DataFrame()
_countries = cat(collect(keys(cmap1)),
collect(keys(cmap2)),
collect(keys(cmap3)),
collect(keys(cmap4)), dims=1) |> unique |> sort
countries[!, :Index] = _countries
fn(country::AbstractString, cmap::Dict{<:AbstractString,Int}) = begin
if !haskey(cmap, country)
return 0
else
return cmap[country]
end
end
countries[!, :Winner] = map(x -> fn(x, cmap1), countries[!, :Index])
countries[!, :Second] = map(x -> fn(x, cmap2), countries[!, :Index])
countries[!, :Third] = map(x -> fn(x, cmap3), countries[!, :Index])
countries[!, :Fourth] = map(x -> fn(x, cmap4), countries[!, :Index])
countries[!, :SemiFinal] = countries[!, :Winner] .+ countries[!, :Second] .+ countries[!, :Third] .+ countries[!, :Fourth]
countries[!, :Final] = countries[!, :Winner] .+ countries[!, :Second]
let
x = countries[!, :Index]
y = countries[!, :SemiFinal]
title = "SemiFinal Statistic"
bar(x, y, title=title, xrotation=45, xticks=:all) |> display
end
可以看到，德国队是进入半决赛次数最多的队伍，紧随其后的是巴西队和意大利队，这和夺冠数量的分布基本一致。我们再来看看进入决赛的队伍统计，是否也是这个趋势
filterfn = [:Winner, :Second] => (winner, second) -> !((winner == 0) & (second == 0) != 0)
finalist = filter(filterfn, countries)
let x = finalist[!, :Index]
y = finalist[!, :Final]
title = "Final Statistic"
bar(x, y, title = title, xticks = :all, xrotation = 45) |> display
end
同样的结论，德国、巴西、意大利3个足球强国也是进入决赛次数最多的队伍。接下来我们来看看进入决赛后各支队伍夺冠的概率如何？
let
finalist[!, :ChampionProb] = finalist[!, :Winner] ./ finalist[:, :Final]
indexs = (finalist[!, :Second] .> 0) .| (finalist[!, :Winner] .> 0)
ratios = finalist[indexs, :]
x = ratios[!, :Index]
y = ratios[!, :ChampionProb]
title = "Percentage of winning reaching the final"
bar(x, y, title = title, xticks = :all, xrotation = 45) |> display
end
我们可以看到英格兰队、西班牙队和乌拉圭队进入决赛后均获得冠军，夺冠概率是100%。 不过从半决赛（4强）队伍次数统计中的结果我们可以看到，英格兰、西班牙、乌拉圭这3支队伍进入决赛的次数分别是1次、1次和2次，统计数据有较大的不确定性。我们还是拿德国、巴西、意大利3支世界杯强队来看，巴西进入决赛夺冠的概率略胜一筹。
巴西、德国、意大利3支强队代表了南美洲和欧洲足球的最高水平，这两个大洲也是现代足球的发源和兴起地，相信很多球迷朋友都会关注五大联赛（注：五大联赛是指西甲、英超、德甲、意甲和法甲），足以说明足球在欧洲的盛行。我们来看看世界杯夺冠队伍所在洲的分布，是不是以南美和欧洲为主
let cmap = countmap(histWorldCup.WinnerContinent)
xs = collect(keys(cmap))
ys = collect(values(cmap))
title = "Champion Continent Numbers"
bar(xs, ys, title = title, xticks = :all) |> display
summary = reduce(+, ys)
_values = map(x -> x / summary, ys)
title = "Champion Continent Ratios"
pie(xs, _values, title = title) |> display
end
let hostTop4 = map(row -> in(row.HostCountry, [row.Winner, row.Second, row.Third, row.Fourth]) ? 1 : 0, eachrow(histWorldCup))
cmap = countmap(hostTop4)
x = collect(keys(cmap))
y = collect(values(cmap))
title = "Host in Top4"
bar(x, y, title = title, xticks = :all) |> display
summary = reduce(+, y)
_values = map(x -> x / summary, y)
title = "Percentage"
pie(x, _values, title = title) |> display
end
# 东道主进入决赛概率
let hostTop2 = map(row -> in(row.HostCountry, [row.Winner, row.Second]) ? 1 : 0, eachrow(histWorldCup))
cmap = countmap(hostTop2)
x = collect(keys(cmap))
y = collect(values(cmap))
title = "Host in Top2"
bar(x, y, title = title, xticks = :all) |> display
summary = reduce(+, y)
_values = map(x -> x / summary, y)
title = "Percentage"
pie(x, _values, title = title) |> display
end
# 东道主夺冠概率
let hostWinner = map(row -> row.HostCountry == row.Winner ? 1 : 0, eachrow(histWorldCup))
cmap = countmap(hostWinner)
x = collect(keys(cmap))
y = collect(values(cmap))
title = "Host in Winner"
bar(x, y, title = title, xticks = :all) |> display
summary = reduce(+, y)
_values = map(x -> x / summary, y)
title = "Percentage"
pie(x, _values, title = title) |> display
end
matches = CSV.read("data/world-cup/WorldCupMatches.csv", DataFrame)
中国队参加的比赛
let indexs = (matches[!, Symbol("Away Team Name")] .== "China PR") .| (matches[!, Symbol("Home Team Name")] .== "China PR")
matches[indexs, :]
end
类似分析世界杯汇总信息，在数据预处理阶段，我们来完成数据清洗和特殊字段的添加工作：
# 统一联邦德国和德国
let columns = names(matches)
for column in columns
replace!(matches[!, column], "Germany FR" => "Germany")
end
end
# 类型转化
matches[!, "Home Team Goals"] = convert(Vector{Int}, matches[!, "Home Team Goals"])
matches[!, "Away Team Goals"] = convert(Vector{Int}, matches[!, "Away Team Goals"])
matches[!, "Result"] = map((x, y) -> "$(x) - $(y)", matches[!, "Home Team Goals"], matches[!, "Away Team Goals"])
let top5Attendance = first(sort(matches, [order(:Attendance, rev=true)]), 5)
top5Attendance[!, :VS] = map((x, y) -> "$(x) VS $(y)", top5Attendance[!, "Home Team Name"], top5Attendance[!, "Away Team Name"])
x = top5Attendance[!, :VS]
y = top5Attendance[!, :Attendance]
bar(x, y) |> display
end这里数据有些问题，在 Attendance 字段有缺失值，这里我省略掉
比赛最令球迷兴奋的当然是进球了，我们来找出历史上单场比赛进球数最多的比赛
matches[!, :TotalGoals] = matches[!, "Home Team Goals"] .+ matches[!, "Away Team Goals"]
matches[!, :VS] = map((x, y) -> "$(x) VS $(y)", matches[!, "Home Team Name"], matches[!, "Away Team Name"])
let top10Goals = first(sort(matches, [order(:TotalGoals, rev=true)]), 10)
top10Goals[!, :VS] = map((x, y) -> "$(x) VS $(y)", top10Goals[!, "Home Team Name"], top10Goals[!, "Away Team Name"])
top10Goals[!, :TotalGoalsStr] = map(x -> "$(x) goals scored", top10Goals[!, "TotalGoals"])
top10Goals[!, "Home Team Goals"] = convert(Vector{Int}, top10Goals[!, "Home Team Goals"])
top10Goals[!, "Away Team Goals"] = convert(Vector{Int}, top10Goals[!, "Away Team Goals"])
top10Goals[!, "Result"] = map((x, y) -> "$(x) - $(y)", top10Goals[!, "Home Team Goals"], top10Goals[!, "Away Team Goals"])
x = top10Goals[!, "VS"]
y = top10Goals[!, "TotalGoals"]
title = "Top10 Goals Match"
bar(x, y, title = title, xrotation = 90, xticks = :all) |> display
end
我们再来分析比赛分差最大的比赛
let
matches[!, "DifferenceGoals"] = abs.(matches[!, "Home Team Goals"] .- matches[!, "Away Team Goals"])
top10Difference = first(sort(matches, [order("DifferenceGoals", rev = true)]), 10)
top10Difference[!, "DifferenceGoals"] = convert(Vector{Int}, top10Difference[!, "DifferenceGoals"])
top10Difference[!, "DifferenceGoalsStr"] = map(x -> "$(x) goals difference", top10Difference[!, "DifferenceGoals"])
top10Difference[!, "Result"] = map((x, y) -> "$(x) - $(y)", top10Difference[!, "Home Team Goals"], top10Difference[!, "Away Team Goals"])
x = top10Difference[!, "VS"]
y = top10Difference[!, "DifferenceGoals"]
title = "Top10 Biggest Difference Matches"
bar(x, y, title = title, xrotation = 90) |> display
end
可以看到，top10分差大的比赛都是聚集在小组赛阶段（stage：GroupX），只有一场是发生在16进8阶段。一般来说进入淘汰赛阶段，两队都会打得比较谨慎，发生大开大合比分的概率比较小
我们再来看看世界杯历史上进球最多的国家，大家可以先猜一下会不会分布在巴西队、德国队和意大利队这3个足球强国中：
using StatsPlots
let columns = names(matches)
for column in columns
replace!(matches[!, column], "Germany FR" => "Germany")
end
listCountries = unique(matches[!, "Home Team Name"])
listHome = Int[]
listAway = Int[]
for country in listCountries
indexs = matches[!, "Home Team Name"] .== country
goalsHome = reduce(+, matches[!, "Home Team Goals"][indexs])
push!(listHome, goalsHome)
indexs = matches[!, "Away Team Name"] .== country
goalsAway = reduce(+, matches[!, "Away Team Goals"][indexs])
push!(listAway, goalsAway)
end
df = DataFrame(Country = listCountries, TotalHomeGoals = listHome, TotalAwayGoals = listAway)
df[!, "TotalGoals"] = df[!, "TotalHomeGoals"] .+ df[!, "TotalAwayGoals"]
mostGoals = first(sort(df, [order(:TotalGoals, rev=true)]), 10)
x = mostGoals[!, "Country"]
y = select(mostGoals, ["TotalHomeGoals", "TotalAwayGoals", "TotalGoals"]) |> Matrix
groupedbar(x, y, xrotation = 90, labels = ["TotalHomeGoals" "TotalAwayGoals" "TotalGoals"]) |> display
end
和我们猜想的差不多，历史进球最多的队伍分别是德国队、巴西队、阿根廷队和意大利队；主场进球最队的国家队分别是巴西队、德国队、阿根廷队和意大利队；客场进球排名是德国、巴西、西班牙和法国队。
大家可能会有个疑问世界杯比赛为什么要分主客场？ 此处给大家做个科普，其实世界杯比赛的“主客场”并非真实意义的主、客场，主要是用来区分主客场球衣，方便区分参赛队伍双方的球衣颜色
看完进球数，我们再来分析失球数
let finalista = finalist[!, :Index]
goalsConcededHome = Int[]
goalsConcededAway = Int[]
match1 = Int[]
match2 = Int[]
for country in finalista
indexs = matches[!, "Home Team Name"] .== country
goalsConcHome = reduce(+, matches[!, "Away Team Goals"][indexs])
push!(goalsConcededHome, goalsConcHome)
counted1 = reduce(+, indexs)
indexs = matches[!, "Away Team Name"] .== country
goalsConcAway = reduce(+, matches[!, "Home Team Goals"][indexs])
push!(goalsConcededAway, goalsConcAway)
counted2 = reduce(+, indexs)
push!(match1, counted1)
push!(match2, counted2)
end
df = DataFrame(Country = finalista, GoalsConcededHome = goalsConcededHome, GoalsConcededAway = goalsConcededAway, MatchesHome = match1, MatchesAway = match2)
df[!, "TotalMatches"] = df[!, "MatchesHome"] .+ df[!, "MatchesAway"]
df[!, "TotalGoalsConceded"] = df[!, "GoalsConcededHome"] .+ df[!, "GoalsConcededAway"]
df[!, "GoalMatchRate"] = round.(df[!, "TotalGoalsConceded"] ./ df[!, "TotalMatches"], digits = 2)
goalsConceded = first(sort(df, [order("GoalMatchRate", rev=true)]), 10)
x = goalsConceded[!, "Country"]
y = goalsConceded[!, "TotalGoalsConceded"]
bp1 = bar(x, y, xrotation = 90)
x = goalsConceded[!, "Country"]
y = goalsConceded[!, "GoalMatchRate"]
bp2 = bar(x, y)
plot(bp1, bp2) |> display
end
可以看出，总失球数最多的进入决赛圈的国家分别是德国、巴西、阿根廷和意大利，这也和这四支强队进入到决赛次数多是正相关的。
从场均失球率来看，英格兰队、荷兰队和意大利队的场均失球率均低于1，说明这三支球队比较擅长防守