-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathMainModeller.m
228 lines (226 loc) · 10.6 KB
/
MainModeller.m
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
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
% Code for modelling the football bets
% Codes developed by Arman Hassanniakalager
% Last modified 10 Sep. 2018 13:28 BST.
clear;
clc;
% Add the codes for Conditional Logit
addpath([cd,'\clogit']);
% Suppress warning notifications within the loops
warning('off','all');
rng(0);
%% IS-OOS specification
% Insample period
ISper=3; % years
OOSper=1;
yrrng=2010:2017-(ISper-1)-OOSper;
ISrng=[yrrng',yrrng'+(ISper-1)];
% Out-of-sample period
OOSrng=[ISrng(:,2)+1,ISrng(:,2)+OOSper];
rectbl=table();
CapitalSize=100;
InvestSize=5; % In %
for i=1:size(ISrng,1)
disp(['Modelling year ',num2str(ISrng(i,2)),' ...']);
% IS data
tblIS=readtable(['England_',num2str(ISrng(i,1)),'_proc.xlsx']);
for k=ISrng(i,1)+1:ISrng(i,2)
addtbl=readtable(['England_',num2str(k),'_proc.xlsx']);
tblIS=[tblIS;addtbl];
end
% OOS data
tblOOS=readtable(['England_',num2str(OOSrng(i,1)),'_proc.xlsx']);
if OOSper>1
for m=OOSrng(i,1)+1:OOSrng(i,2)
addedtbl=readtable(['England_',num2str(m),'_proc.xlsx']);
tblOOS=[tblOOS;addedtbl];
end
end
tblIS{:,'IPres'}=ones(size(tblIS,1),1)*1e-4;
tblOOS{:,'IPres'}=ones(size(tblOOS,1),1)*1e-4;
AHresIS=(tblIS.FTHG-tblIS.FTAG)+tblIS.AHh;
AHresIS(AHresIS>=0)=1; % Home Win
AHresIS(AHresIS<0)=2; % Lose
tblIS.AH=AHresIS;
AHresOOS=(tblOOS.FTHG-tblOOS.FTAG)+tblOOS.AHh;
AHresOOS(AHresOOS>=0)=1;
AHresOOS(AHresOOS<0)=2;
tblOOS.AH=AHresOOS;
AHcte=abs(min(tblIS.AHh))+1;
%% Logit specification for 1X2
%% IS specification
% Y is an N x 1 vector of integers 1 through J indicating which
% alternative was chosen.
Y_IS1X2=tblIS.FTR;
% X is an N x K1 matrix of individual-specific covariates.
indpredictors1X2=[tblIS.PtH5,tblIS.PtA5,tblIS.GsH5,tblIS.GsA5,tblIS.GcH5,tblIS.GcA5,tblIS.IPAHH,tblIS.IPAHA,tblIS.AHh+AHcte];
X_IS1X2=[indpredictors1X2,tblIS.IP1,tblIS.IPX,tblIS.IP2];
%% OOS specification
Y_OOS1X2=tblOOS.FTR;
indpredictors_OOS1X2=[tblOOS.PtH5,tblOOS.PtA5,tblOOS.GsH5,tblOOS.GsA5,tblOOS.GcH5,tblOOS.GcA5,tblOOS.IPAHH,tblOOS.IPAHA,tblOOS.AHh+AHcte];
%% MATLAB built-in fcn
B1X2 = mnrfit(X_IS1X2,Y_IS1X2,'Model','nominal','Interactions','on');
pihat1X2 = mnrval(B1X2,X_IS1X2);
[~,Predict1X2]=max(pihat1X2,[],2);
%% CLogit performance
ISaccuracy1X2(i)=sum(Predict1X2==Y_IS1X2)/numel(Y_IS1X2);
X_OOS1X2=[indpredictors_OOS1X2,tblOOS.IP1,tblOOS.IPX,tblOOS.IP2];
pihat_OOS1X2 = mnrval(B1X2,X_OOS1X2);
[~,Predict_OOS1X2]=max(pihat_OOS1X2,[],2);
OOSaccuracy1X2(i)=sum(Predict_OOS1X2==Y_OOS1X2)/numel(Y_OOS1X2);
tblnew1=payoffestimator3(Predict_OOS1X2,tblOOS,1);
tblOOS.Properties.VariableNames(12:14)={'BbAvH','BbAvD','BbAvA'};
%% Logit specification for Over/Under 2.5 goals
%% IS specification
Y_ISOvUn=tblIS.OvUn;
indpredictorsOvUn=[tblIS.PtH5,tblIS.PtA5,tblIS.GsH5,tblIS.GsA5,tblIS.GcH5,tblIS.GcA5,tblIS.IPAHH,tblIS.IPAHA,tblIS.AHh+AHcte];
X_ISOvUn=[indpredictorsOvUn,tblIS.IPOv,tblIS.IPUn];
%% OOS specification
Y_OOSOvUn=tblOOS.OvUn;
indpredictors_OOSOvUn=[tblOOS.PtH5,tblOOS.PtA5,tblOOS.GsH5,tblOOS.GsA5,tblOOS.GcH5,tblOOS.GcA5,tblOOS.IPAHH,tblOOS.IPAHA,tblOOS.AHh+AHcte];
B_OvUn = mnrfit(X_ISOvUn,Y_ISOvUn,'Model','nominal','Interactions','on');
pihatOvUn = mnrval(B_OvUn,X_ISOvUn);
% B_OvUn = glmfit(X_ISOvUn,Y_ISOvUn-1,'binomial');
% yhat = glmval(B_OvUn,X_ISOvUn,'logit');
[~,PredictOvUn]=max(pihatOvUn,[],2);
ISaccuracyOvUn(i)=sum(PredictOvUn==Y_ISOvUn)/numel(Y_ISOvUn);
X_OOSOvUn=[indpredictors_OOSOvUn,tblOOS.IPOv,tblOOS.IPUn];
pihat_OOSOvUn = mnrval(B_OvUn,X_OOSOvUn);
[~,Predict_OOSOvUn]=max(pihat_OOSOvUn,[],2);
OOSaccuracyOvUn(i)=sum(Predict_OOSOvUn==Y_OOSOvUn)/numel(Y_OOSOvUn);
tblnew2=payoffestimator3(Predict_OOSOvUn,tblOOS,2);
tblOOS.Properties.VariableNames(15:16)={'BbAvOver','BbAvUnder'};
%% Logit specification for Asian Handicap
%% IS specification
Y_ISAH=tblIS.AH;
indpredictorsAH=[tblIS.PtH5,tblIS.PtA5,tblIS.GsH5,tblIS.GsA5,tblIS.GcH5,tblIS.GcA5,tblIS.IPAHH,tblIS.IPAHA,tblIS.AHh+AHcte];
X_ISAH=indpredictorsAH;
%% OOS specification
Y_OOSAH=tblOOS.AH;
indpredictors_OOSAH=[tblOOS.PtH5,tblOOS.PtA5,tblOOS.GsH5,tblOOS.GsA5,tblOOS.GcH5,tblOOS.GcA5,tblOOS.IPAHH,tblOOS.IPAHA,tblOOS.AHh+AHcte];
B_AH = mnrfit(X_ISAH,Y_ISAH,'Model','nominal','Interactions','on');
pihatAH = mnrval(B_AH,X_ISAH);
% B_AH = glmfit(X_ISAH,Y_ISAH-1,'binomial');
% yhat = glmval(B_AH,X_ISAH,'logit');
[~,PredictAH]=max(pihatAH,[],2);
ISaccuracyAH(i)=sum(PredictAH==Y_ISAH)/numel(Y_ISAH);
X_OOSAH=indpredictors_OOSAH;
pihat_OOSAH = mnrval(B_AH,X_OOSAH);
[~,Predict_OOSAH]=max(pihat_OOSAH,[],2);
OOSaccuracyAH(i)=sum(Predict_OOSAH==Y_OOSAH)/numel(Y_OOSAH);
tblOOS.Properties.VariableNames(17)={'BbAHh'};
tblOOS.Properties.VariableNames(18:19)={'BbAvAHH','BbAvAHA'};
tblnew4=payoffestimator4(Predict_OOSAH,tblOOS,4);
%% Logit specification for Correct Scores
% Generating categorical outputs for IS
outcomes={'0:0','0:1','0:2','0:3','0:4','1:0','1:1','1:2','1:3',...
'1:4','2:0','2:1','2:2','2:3','2:4','3:0','3:1','3:2',...
'3:3','3:4','4:0','4:1','4:2','4:3'};
delimiterIS=[];
delimiterIS(1:size(tblIS,1),1)=':';
scorelineIS=cellstr([num2str(tblIS.FTHG),delimiterIS,num2str(tblIS.FTAG)]);
scorelinecatIS=nan(size(tblIS,1),1);
for f=1:size(tblIS,1)
if ~isempty(find(strcmp(scorelineIS(f),outcomes), 1))
scorelinecatIS(f)=find(strcmp(scorelineIS(f),outcomes));
else
scorelinecatIS(f)=size(outcomes,2)+1;
end
end
tblIS.SC=scorelinecatIS;
Y_ISCS=scorelinecatIS;
indpredictorsCS=[tblIS.PtH5,tblIS.PtA5,tblIS.GsH5,tblIS.GsA5,tblIS.GcH5,tblIS.GcA5,tblIS.IP1,tblIS.IPX,tblIS.IP2,tblIS.IPOv,tblIS.IPUn,tblIS.IPAHH,tblIS.IPAHA,tblIS.AHh+AHcte];
%% OOS categorical
delimiterOOS=[];
delimiterOOS(1:size(tblOOS,1),1)=':';
scorelineOOS=cellstr([num2str(tblOOS.FTHG),delimiterOOS,num2str(tblOOS.FTAG)]);
scorelinecatOOS=nan(size(tblOOS,1),1);
for f=1:size(tblOOS,1)
if ~isempty(find(strcmp(scorelineOOS(f),outcomes), 1))
scorelinecatOOS(f)=find(strcmp(scorelineOOS(f),outcomes));
else
scorelinecatOOS(f)=size(outcomes,2)+1;
end
end
Y_OOSCS=scorelinecatOOS;
tblOOS.SC=scorelinecatOOS;
% %% Matlab built-in
% B_CS = mnrfit(X_ISCS,Y_ISCS,'Model','nominal','Interactions','on');
% pihatCS = mnrval(B_CS,X_ISCS);
% Clogit code
% IS specification
columns=unique(Y_ISCS);
J=numel(columns);
cte=ones(size(tblIS,1),1);
X_IS=[cte,indpredictorsCS];
Z_IS=nan(size(tblIS,1),1,J);
firstSCoddcolumn=find(contains(tblOOS.Properties.VariableNames,'IP0_0'));
Z_IS(:,1,:)=tblIS{:,firstSCoddcolumn+columns-1};
options=optimset('Algorithm','trust-region','Disp','off','LargeScale','on','MaxFunEvals',2000000,'MaxIter',15000,'TolX',1e-8,'Tolfun',1e-8,'GradObj','on','DerivativeCheck','off','FinDiffType','central','UseParallel',true);
%options = optimoptions('fminunc','Algorithm','trust-region');
rng(0);
b=rand(size(X_IS,2),J);
bz=2;
bAns = b(:)-repmat(b(:,J),J,1);
bAns = cat(1,bAns(1:(J-1)*size(X_IS,2)),bz);
startval = bAns;
bEst = fminunc('clogit',startval,options,[],Y_ISCS,X_IS,Z_IS);
Probs = pclogit(bEst,Y_ISCS,X_IS,Z_IS);
[~,PredictCS]=max(Probs,[],2);
ISaccuracyCS(i)=sum(PredictCS==Y_ISCS)/numel(Y_ISCS);
% OOS specification
indpredictorsCSOOS=[tblOOS.PtH5,tblOOS.PtA5,tblOOS.GsH5,tblOOS.GsA5,tblOOS.GcH5,tblOOS.GcA5,tblOOS.IP1,tblOOS.IPX,tblOOS.IP2,tblOOS.IPOv,tblOOS.IPUn,tblOOS.IPAHH,tblOOS.IPAHA,tblOOS.BbAHh+AHcte];
cte_OOS=ones(size(tblOOS,1),1);
X_OOSCS=[cte_OOS,indpredictorsCSOOS];
Z_OOSCS=nan(size(tblOOS,1),1,J);
Z_OOSCS(:,1,:)=tblOOS{:,firstSCoddcolumn+columns-1};
randYOOS=(1:J)';
randYOOS=repmat(randYOOS,floor(size(tblOOS,1)/J),1);
randYOOS=[randYOOS;(1:(size(tblOOS,1)-size(randYOOS,1)))'];
Probs_OOSCS = pclogit(bEst,randYOOS,X_OOSCS,Z_OOSCS);
[~,Predict_OOSCS]=max(Probs_OOSCS,[],2);
OOSaccuracyCS(i)=sum(Predict_OOSCS==Y_OOSCS)/numel(Y_OOSCS);
tblnew3=payoffestimator3(Predict_OOSCS,tblOOS,3);
%tblnew=[tblOOS,tblnew1(:,end-1:end),tblnew1CF(:,end-1:end),tblnew2(:,end-1:end),tblnew2CF(:,end-1:end),tblnew3(:,end-1:end)];
tblnew=[tblOOS,tblnew1(:,end-1:end),... 1X2
tblnew2(:,end-1:end),... OVER/UNDER
tblnew4(:,end-1:end),... ASIAN HANDICAP
tblnew3(:,end-1:end)... CS
];
%% Gambling game
[betprofit1X2,hl_1X2,capser1X2]=gamblerfun(tblnew.Result_Bet_Reward,InvestSize/100,CapitalSize);
[betprofitOvUn,hl_OvUn,capserOvUn]=gamblerfun(tblnew.OvUn_Bet_Reward,InvestSize/100,CapitalSize);
[betprofitAH,hl_AH,capserAH]=gamblerfun(tblnew.AH_Bet_Reward,InvestSize/100,CapitalSize);
[betprofitCS,hl_CS,capserCS]=gamblerfun(tblnew.CS_Bet_Reward,InvestSize/100,CapitalSize);
%% record results
rectbl{i,'IS_Start'}=ISrng(i,1);
rectbl{i,'IS_Finish'}=ISrng(i,2);
rectbl{i,'IS_Period'}=ISrng(i,2)-ISrng(i,1)+1;
rectbl{i,'OOS_Start'}=OOSrng(i,1);
rectbl{i,'OOS_Finish'}=OOSrng(i,2);
rectbl{i,'OOS_Period'}=OOSrng(i,2)-OOSrng(i,1)+1;
rectbl{i,'GamesCount'}=size(tblOOS,1);
rectbl{i,'IS_Accuracy1X2'}=ISaccuracy1X2(i);
rectbl{i,'OOS_Accuracy1X2'}=OOSaccuracy1X2(i);
rectbl{i,'OOS_Profit1X2'}=sum(tblnew.Result_Bet_Reward)/size(tblOOS,1);
rectbl{i,'OOS_HalfLife1X2'}=hl_1X2;
rectbl{i,'OOS_Gamble1X2'}=betprofit1X2;
rectbl{i,'IS_AccuracyOvUn'}=ISaccuracyOvUn(i);
rectbl{i,'OOS_AccuracyOvUn'}=OOSaccuracyOvUn(i);
rectbl{i,'OOS_ProfitOvUn'}=sum(tblnew.OvUn_Bet_Reward)/size(tblOOS,1);
rectbl{i,'OOS_HalfLifeOvUn'}=hl_OvUn;
rectbl{i,'OOS_GambleOvUn'}=betprofitOvUn;
rectbl{i,'IS_AccuracyAH'}=ISaccuracyAH(i);
rectbl{i,'OOS_AccuracyAH'}=OOSaccuracyAH(i);
rectbl{i,'OOS_ProfitAH'}=sum(tblnew.AH_Bet_Reward)/size(tblOOS,1);
rectbl{i,'OOS_HalfLifeAH'}=hl_AH;
rectbl{i,'OOS_GambleAH'}=betprofitAH;
rectbl{i,'IS_AccuracyCS'}=ISaccuracyCS(i);
rectbl{i,'OOS_AccuracyCS'}=OOSaccuracyCS(i);
rectbl{i,'OOS_ProfitCS'}=sum(tblnew.CS_Bet_Reward)/size(tblOOS,1);
rectbl{i,'OOS_HalfLifeCS'}=hl_CS;
rectbl{i,'OOS_GambleCS'}=betprofitCS;
save(['E0Modelling_',num2str(ISrng(i,1)),'_',num2str(ISrng(i,2))])
writetable(tblnew,['Predicted_Season_',num2str(ISrng(i,1)),'_',num2str(ISrng(i,2)),'.xlsx'])
end
close;
writetable(rectbl,'Logit_Results.xlsx');