-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathanalysis.R
185 lines (156 loc) · 7.85 KB
/
analysis.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
rm(list=ls())
# Loading Packages
library(data.table);
library(dplyr);
library(reshape);
library(stringdist)
rename = dplyr::rename
################################################################################
DataFiles = fread('meta_data.txt')
################################################################################
# Download data if missing
apply(DataFiles,1,function(x){
file_location=paste('data/',x[1],'.csv',sep='')
if(!file.exists(file_location)){
download.file(x[2],file_location)
}
})
################################################################################
# Reading in the raw data
TourneySeeds <- fread("data/TourneySeeds.csv")
SampleSubmission <- fread("data/SampleSubmission.csv")
Seasons <- fread("data/Seasons.csv")
Teams <- fread("data/Teams.csv")
TourneySlots <- fread("data/TourneySlots.csv")
TourneyDetailedResults <- fread("data/TourneyDetailedResults.csv")
#TourneyCompactResults <- fread("data/TourneyCompactResults.csv")
RegularSeasonDetailedResults <- fread("data/RegularSeasonDetailedResults.csv")
#RegularSeasonCompactResults <- fread("data/RegularSeasonCompactResults.csv")
KenPom <- fread("data/KenPom.csv")
TeamSpelling <- fread("data/TeamSpellings.csv")
################################################################################
# Preproces data
# Rename for merge
TeamSpelling <- rbind(TeamSpelling,rename(Teams,name_spelling=Team_Name,team_id=Team_Id))
# Extracting seeds for each team
TourneySeeds <- TourneySeeds %>%
mutate(SeedNum = gsub("[A-Z+a-z]", "", Seed)) %>% select(Season, Team, SeedNum)
# fix external data
colnames(TeamSpelling)[1] = "Team"
KenPom$Team = tolower(KenPom$Team)
Teams$Team_Name = tolower(Teams$Team_Name)
# mamp spelling
KenPom = left_join(KenPom, TeamSpelling, by=c('Team'='Team'))
################################################################################
# Construct summary data
winner_Tdata = select(RegularSeasonDetailedResults, Season, Wteam, Wscore, Wfgm:Wpf)
loser_Tdata = select(RegularSeasonDetailedResults, Season, Lteam, Lscore, Lfgm:Lpf)
Team_Game_History = rbind(winner_Tdata, setNames(loser_Tdata, names(winner_Tdata)))
colnames(Team_Game_History) = c("season",
"team",
"score",
"fgm",
"fga",
"fgm3",
"fga3",
"ftm",
"fta",
"or",
"dr",
"ast",
"to",
"stl",
"blk",
"Wpf")
team_data_by_season = Team_Game_History %>% group_by(season, team) %>% summarise_each(funs(mean))
#team_data = Team_Game_History %>%group_by(team) %>% summarise_each(funs(mean))
# Prepare data for joining
#team1_data = data.frame(team_data)
#colnames(team1_data) <- paste("team1", colnames(team1_data), sep = "_")
#team2_data = data.frame(team_data)
#colnames(team2_data) <- paste("team2", colnames(team2_data), sep = "_")
team1_data_by_season = data.frame(team_data_by_season)
colnames(team1_data_by_season) <- paste("team1", colnames(team1_data_by_season), sep = "_")
team2_data_by_season = data.frame(team_data_by_season)
colnames(team2_data_by_season) <- paste("team2", colnames(team2_data_by_season), sep = "_")
team1_kenpom_data_by_season = data.frame(KenPom)
colnames(team1_kenpom_data_by_season) <- paste("team1", colnames(team1_kenpom_data_by_season), sep = "_")
team2_kenpom_data_by_season = data.frame(KenPom)
colnames(team2_kenpom_data_by_season) <- paste("team2", colnames(team2_kenpom_data_by_season), sep = "_")
################################################################################
# Get going with actual machine learning:
# 1. set train,test,predict
# 2. add available data
# 3. build models
# 4. Make Predictions
################################################################################
# Prepare prediction target
games.to.train <- RegularSeasonDetailedResults %>%
mutate(season=Season, team1=Wteam, team2=Lteam, Score_diff=Wscore-Lscore, team1win=1) %>%
select(season, team1, team2, Score_diff, team1win)
games.to.test <- TourneyDetailedResults %>%
mutate(season=Season, team1=Wteam, team2=Lteam, Score_diff=Wscore-Lscore, team1win=1) %>%
select(season, team1, team2, Score_diff, team1win)
games.to.predict <- cbind(SampleSubmission$Id, colsplit(SampleSubmission$Id, split = "_", names = c('season', 'team1', 'team2')))
flippedGames = function(game){
flipped <- game %>%
rename(team1=team2,team2=team1) %>%
mutate(Score_diff=-Score_diff, team1win=0)
}
games.to.train = rbind(games.to.train,flippedGames(games.to.train))
# Add available data to each target game
addDataToGames = function(games) {
games <- data.frame(games) %>%
# add March madness seed to teams in games
left_join(TourneySeeds, by=c("season"="Season", "team1"="Team")) %>%
rename(team1seed = SeedNum) %>%
left_join(TourneySeeds, by=c("season"="Season", "team2"="Team")) %>%
rename(team2seed = SeedNum) %>%
mutate(team1seed = as.numeric(team1seed), team2seed = as.numeric(team2seed)) %>%
# add seasonal data
left_join(team1_data_by_season,by=c("season" = "team1_season", "team1"="team1_team")) %>%
left_join(team2_data_by_season,by=c("season" = "team2_season", "team2"="team2_team")) %>%
# add external data
left_join(team1_kenpom_data_by_season,by=c("season" = "team1_Year", "team1"="team1_team_id")) %>%
left_join(team2_kenpom_data_by_season,by=c("season" = "team2_Year", "team2"="team2_team_id"))
return(games)
}
games.to.train = addDataToGames(games.to.train)
games.to.test = addDataToGames(games.to.test)
games.to.predict = addDataToGames(games.to.predict)
games.to.train = games.to.train %>% na.omit()
games.to.test = games.to.test %>% na.omit()
getMissing = function(){
a=Teams$Team_Name[which(Teams$Team_Id %in% unique(games.to.test$team2[which(is.na(games.to.test$team2_Team))]))]
b=Teams$Team_Name[which(Teams$Team_Id %in% unique(games.to.test$team1[which(is.na(games.to.test$team1_Team))]))]
c=Teams$Team_Name[which(Teams$Team_Id %in% unique(games.to.train$team2[which(is.na(games.to.train$team2_Team))]))]
d=Teams$Team_Name[which(Teams$Team_Id %in% unique(games.to.train$team1[which(is.na(games.to.train$team1_Team))]))]
e=Teams$Team_Name[which(Teams$Team_Id %in% unique(games.to.predict$team2[which(is.na(games.to.predict$team2_Team))]))]
f=Teams$Team_Name[which(Teams$Team_Id %in% unique(games.to.predict$team1[which(is.na(games.to.predict$team1_Team))]))]
missing = unique(c(a,b,c,d,e,f))
print(missing)
return(missing)
}
m.score_diff <- lm(Score_diff~ .,
data=select(games.to.train,
-c(team1win,
team1_Team,
team2_Team,
team1,
team2)))
games.to.train$Predicted_Score_diff = predict(m.score_diff)
games.to.test$Predicted_Score_diff = predict(m.score_diff,games.to.test)
games.to.predict$Predicted_Score_diff = predict(m.score_diff,games.to.predict)
win_chance = ecdf(games.to.train$Score_diff)
games.to.train$Pred = win_chance(games.to.train$Predicted_Score_diff)
games.to.test$Pred = win_chance(games.to.test$Predicted_Score_diff)
games.to.predict$Pred = win_chance(games.to.predict$Predicted_Score_diff)
getLogLoss = function(games){
y = games$team1win
pred = games$Pred
logLoss = -mean(y*log(pred) + (1-y)*log(1-pred))
return(logLoss)
}
getLogLoss(games.to.train)
getLogLoss(games.to.test)
write.csv(games.to.predict %>% select(Id=SampleSubmission.Id, Pred), 'seed_submission.csv', row.names=FALSE)