38

Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

  • Upload
    others

  • View
    10

  • Download
    1

Embed Size (px)

Citation preview

Page 1: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Football betting odds

Yll Mehmeti, Hugo Met and Kevin Turatsinze

ENAC, IENAC15 SITA-PREV

April 7, 2017

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 2: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

1 IntroductionContextDependant variablesExplanatory variables

2 First model : simultaneousHome Win oddAway win oddDraw oddTests

3 Second model : hybridHome Win odd using Draw and Away �tted odds

4 Third model : sequentialDraw odd using Away �tted oddsHome odd using the new Draw �tted odds

5 Conclusion

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 3: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

ContextDependant variablesExplanatory variables

How to bet ?

Table � A game on BWin website

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 4: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

ContextDependant variablesExplanatory variables

Our sample

Teams

p = 7 teams

Caen, Lyon, Marseille, Monaco, Paris, Saint-Etienne, Toulouse

Number of observations

n = 2 ∗ ×∑p−1

k=1 k = 2× p(p−1)2

= p(p − 1) = 7× 6 = 42

So we focused on 42 games of the 2015-2016 Ligue 1 season.

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 5: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

ContextDependant variablesExplanatory variables

BWin odds : BWH

Graph � Histogram and stats for home odds.

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 6: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

ContextDependant variablesExplanatory variables

BWin odds : BWA

Graph � Histogram and stats for away odds.

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 7: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

ContextDependant variablesExplanatory variables

BWin odds : BWD

Graph � Histogram and stats for draw odds.

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 8: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

ContextDependant variablesExplanatory variables

Delta variables

Formula used

∆ = ∆home −∆away

Delta variables

DELTAFIFA,DELTAFORM,DELTAPOINTS ,DELTARANK ,

DELTAAFTER,DELTABEFORE

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 9: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

ContextDependant variablesExplanatory variables

LAST10HOME / LAST10DRAW / LAST10AWAY

Formula used

Last10(Home/Draw/Away) =10∑k=1

k

10× δk,outcome

where outcome = win of the current home team / draw / win ofthe current away team.

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 10: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

ContextDependant variablesExplanatory variables

Other variables

Strictness of the referee

Strictness =NYellowCards + 4× NRedCards

Ngames

Other variables

FIELDGRADE ,TEMPERATURE ,PACKING ,DISTANCE

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 11: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Simultaneous model

B̂WH | B̂WD | B̂WA

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 12: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Home Win odd : estimation

Number of observations and variables

n= 42 and k=13

The linear equation of the model

B̂WH = β̂0 + β̂1DELTAFIFA + β̂2DELTAFORM + β̂3DELTAPOINTS

+ β̂4DELTARANK + β̂5DELTAAFTER + β̂6DELTABEFORE

+ β̂7LAST10HOME + β̂8STRICTNESS + β̂9FIELDGRADE

+ β̂10TEMPERATURE + β̂11PACKING + β̂12DISTANCE

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 13: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Home Win odd : marginal e�ects

Variable Expected e�ect Real e�ect Signi�cant

DeltaFifa - - Yes

DeltaForm - - Yes

DeltaPoints - - Yes

DeltaRank + - Yes

DeltaAfter - - Yes

DeltaBefore - - No

Last10Home - - Yes

Strictness - - No

FieldGrade ? + Yes

Temperature ? 0 No

Packing - 0 No

Distance - 0 No

Table � Marginal e�ects tableYll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 14: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Home Win odd : estimation results

Table � eqBWH_�rst_model

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 15: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Home Win odd : actual and �tted BWH

Graph � eqBWH_�rst_model

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 16: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Home Win odd : biggest errors

Toulouse - Paris Saint-Etienne - Paris Paris - Caen

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 17: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Away Win odd : estimation

Number of observations and variables

n= 42 and k=13

The linear equation of the model

B̂WA = β̂0 + β̂1DELTAFIFA + β̂2DELTAFORM + β̂3DELTAPOINTS

+ β̂4DELTARANK + β̂5DELTAAFTER + β̂6DELTABEFORE

+ β̂7LAST10AWAY + β̂8STRICTNESS + β̂9FIELDGRADE

+ β̂10TEMPERATURE + β̂11PACKING + β̂12DISTANCE

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 18: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Away Win odd : marginal e�ects

Variable Expected e�ect Real e�ect Signi�cant

DeltaFifa + + Yes

DeltaForm + + Yes

DeltaPoints + + No

DeltaRank - + No

DeltaAfter + + No

DeltaBefore + + No

Last10Away - - No

Strictness ? - No

FieldGrade ? + Yes

Temperature ? - No

Packing ? + Yes

Distance ? 0 No

Table � Marginal e�ects tableYll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 19: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Away Win odd : estimation results

Table � eqBWA_�rst_model

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 20: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Away Win odd : actual and �tted BWA

Graph � eqBWA_�rst_model

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 21: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Draw odd : estimation

Number of observations and variables

n= 42 and k=13

The linear equation of the model

B̂WD = β̂0 + β̂1DELTAFIFA + β̂2DELTAFORM + β̂3DELTAPOINTS

+ β̂4DELTARANK + β̂5DELTAAFTER + β̂6DELTABEFORE

+ β̂7LAST10DRAW + β̂8STRICTNESS + β̂9FIELDGRADE

+ β̂10TEMPERATURE + β̂11PACKING + β̂12DISTANCE

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 22: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Draw odd : estimation results

Table � eqBWD01_�rst_model

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 23: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Draw odd : actual and �tted BWD

Graph � eqBWD01_�rst_model

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 24: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Draw odd : new estimation

Number of observations and variables

n= 42 and k=13

The new linear equation of the model

B̂WD = β̂0 + β̂1DELTAFIFA2 + β̂2DELTAFORM

2 + β̂3DELTAPOINTS2

+ β̂4DELTARANK2 + β̂5DELTAAFTER

2 + β̂6DELTABEFORE2

+ β̂7LAST10DRAW + β̂8STRICTNESS + β̂9FIELDGRADE

+ β̂10TEMPERATURE + β̂11PACKING + β̂12DISTANCE

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 25: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Draw odd : new estimation results

Table � eqBWD02_�rst_model

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 26: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Draw odd : actual and new �tted BWD

Graph � eqBWD02_�rst_model

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 27: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win oddAway win oddDraw oddTests

Tests : normality of the errors

BWH BWD BWA

R2 0.76 0.82 0.81

Adjusted R2 0.66 0.69 0.73

JB = n−kn

JBE 9.44 1.19 0.37

χ20.95(2) 5.99 5.99 5.99

Reject normality Yes No No

Table � normality tests results

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 28: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win odd using Draw and Away �tted odds

Hybrid model

B̂WA B̂WD↘ ↙B̂WH

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 29: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win odd using Draw and Away �tted odds

Home Win odd using Draw and Away �tted odds

Table � eqBWH_third_model Table � Wald_�tted_coe�s

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 30: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win odd using Draw and Away �tted odds

Wald tests on Away win and Draw �tted coe�cients

Table � Wald_BWAF_coe� Table � Wald_BWDF_coe�

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 31: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Home Win odd using Draw and Away �tted odds

Home Win odd using Draw and Away �tted odds

Graph � eqBWH_second_model

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 32: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Draw odd using Away �tted oddsHome odd using the new Draw �tted odds

Sequential model

B̂WA −→ B̂WD −→ B̂WH

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 33: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Draw odd using Away �tted oddsHome odd using the new Draw �tted odds

Draw odd using Away �tted odds

Table � eqBWD_third_model Table � Wald_BWAF_coe�

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 34: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Draw odd using Away �tted oddsHome odd using the new Draw �tted odds

Draw odd using the Away �tted odds

Graph � eqBWD_third_modelYll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 35: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Draw odd using Away �tted oddsHome odd using the new Draw �tted odds

Home Win odd using new Draw �tted odds

Table � eqBWH_third_model Table � Wald_BWDF_coe�

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 36: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Draw odd using Away �tted oddsHome odd using the new Draw �tted odds

Home odd using the new Draw �tted odds

Graph � eqBWH_third_model

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 37: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Tests : normality of the errors

simultaneous hybrid sequential

BWH BWD BWA BWH BWD BWA BWH BWD BWA

R2 0.76 0.67 0.81 0.85 0.67 0.81 0.87 0.79 0.81Adjusted R2 0.66 0.53 0.73 0.77 0.53 0.73 0.81 0.70 0.73

JB 9.44 1.19 0.37 0.26 1.19 0.37 0.19 0.16 0.37χ20.95(2) 5.99 5.99 5.99 5.99 5.99 5.99 5.99 5.99 5.99

Rejectnormality

Yes No No No No No No No No

Table � normality tests results

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds

Page 38: Football betting odds - ENACrecherche.enac.fr/.../example_projects/betting.pdfableT of contents Introduction First model : simultaneous Second model : hybrid Third model : sequential

Table of contentsIntroduction

First model : simultaneousSecond model : hybrid

Third model : sequentialConclusion

Football-data. www.football-data.co.uk.

LFP. www.lfp.fr.

SoFIFA. so�fa.com.

Fussbal WettPoint. fussball.wettpoint.com.

Soccer Punter. www.soccerpunter.com.

Distance between two cities. www.distance2villes.com.

Weather records. www.historique-meteo.net.

BWin. sports.bwin.fr.

Yll Mehmeti, Hugo Met and Kevin Turatsinze Football betting odds