MULTIVARIATE STATISTICAL MODELS TO FORECAST THE RESULTS OF EURO2016 QUALIFIERS Bence Jámbor – Dávid Szabó – Máté Jámbor Budapest Business School - Consultant: Dr. Tamás Kovács
MULTIVARIATE STATISTICAL MODELS TO FORECAST THE RESULTS OF EURO2016 QUALIFIERS
Bence Jámbor – Dávid Szabó – Máté JámborBudapest Business School - Consultant: Dr. Tamás Kovács university docent
SUBSTANCE
I. IntroductionII. Overview of the literature III. Development of the teams
performance forecasting model
IV. Results
INTRODUCTION Hungarian football: nadir (WC1986; EC: 1972)
Present and the recent past: failures(2006: Malta 1-2; 2013: Netherlands 8-1)
Successes in the past: 1938, 1954 (Aranycsapat/The Gold team) Puskás; WC: II.place 1964: EC: III. place
(Mészöly-Bene-Albert) 1966: WC-quarterfinal Olympic Champions:
1952,1964,1968
LITERATUREHungarian and international scientific articles forecast of a football match outcome FIFA World Ranking (MAREK KAMINSKI –
INCONSISTENCIES IN THE FIFA RANKINGS) InStat (PÉTER KAKAS- JUST CAN NOT
PLAY FOOTBALL INSTEAD OF THEM ALL-SEEING SOFTWARE)
Home field advantage(HENRIK HEGEDŰS: AWAY THE DRAW IS GOOD RESULT TOO)
BASIC CONCEPT OF THE FORECASTING MODEL
Final outcome of a national match
1. Performance form of the players in their club teams
2. Basic qualities of the players
3. Other factors
Defending(%):- Succesful passes- Challenges won- Aerial challenges won - Succesful tackles
Attacking(%):- Succesful passes- Challenges won- Aerial challenges won - Succesful keypasses - Shots on goal- Succesful dribblings
- Fifa world ranking
- Last 5 comptetitivegames
- Home field
10.000 data
Balázs’s performance on the club match before the examined national match
Keypass% (here it’s missing)
We used this page to examine all the players of the 58 matches.
An example of the 58 examined matches
Defense skills
Attack skills
Create one defense- and attack value in each teamsCreating defense factors SPSS, factor analysis, 3 factors from 5 variablesVariance-proportion
method the variables with the closest relation one factor
1.factor: save-, tackles %
2. factor: challenges won-, aerial challenges won %
3. factor: passes %
3 factors 1 defense value
Creating attack factors
3 factors from 6 variables
1. factor: challenges won-, aerial challanges won-, dribbling %
2. factor: keypasses-, shots on goal %
3. factor: aerial challenges won-, dribblings-, passes %
3 factors 1 attack value
MULTIVARIATE STATISTICAL MODELSMultiple linear regression
difference between current form of attacker players and current form of opponent defender players (both team)
difference between current InStat index of attacker players and current InStat index of opponent defender players (both team)
difference between FIFA index of attacker players and opponent defender players
number of scored goals of attacker players in their previous club game
number of assists of attacker players in their previous club game
difference between the trends of their previous five matches of the two national teams
difference between the Fifa world rank score of the the two national teams
dummy variable of home field
HOME FIELD ADVANTAGE
In 2012 English Premier League decided in the last fixture Manchester City is the champion
REGRESSION MODEL #1Model
Coefficientst p
B Std. Error
Own attacker– opponent defender ,030 ,007 4,097 ,000
Own attacker– opponent defender InStat ,005 ,003 2,028 ,045
Own attacker– opponent defender Fifa ,092 ,032 2,836 ,005
Own FIFA World Rank– Opponent FIFA World Rank -,001 ,000 -1,680 ,096
Own Homefield Advantage ,675 ,221 3,052 ,003
REGRESSION MODEL #2
ModelCoefficients
t pB Std.
Error
Attacker– Defender (based on Fifa
form)
,033 ,010 3,337 ,001
Own attacker– Own defender (InStat),011 ,002 4,653 ,000
Own homefield advantage,850 ,227 3,754 ,000
THE EXPLANATORY POWER OF THE REGRESSION MODELSR - multiple correlation coefficientR2 - multiple determination coefficient
Model #1
R R Square Adjusted R
Square
,717 ,515 ,493
Model 2#
R R Square Adjusted R
Square
,678 ,459 ,440
TESTING THE RESULTS OF THE MODELSn=58 matchesPrediction of the outcomes
Model #1 56,9% (33/58) Model #2 62,1% (36/58)
Preditction of the number of scored goals Model #1 33,6% (39/116)Model #2 31,0% (36/116)
HITSMatch Result Estimated Result
Greece– Bosnia 0-0 0-0
Bosnia – Slovakia 0-1 0-1
Ukraine– England 0-0 0-0
Greece– Slovakia 1-0 1-0
France– Spain 0-1 0-1
Portugal– Sweden 1-0 1-0
Croatia– Island 2-0 2-0
Switzerland– Slovenia 1-0 1-0
SOME OUTCOME HITTED MATCHES
Match Result Estimated ResultBosnia-Greece 3-1 1-0
Romania-Greece 1-1 0-0
Austria-Germany 1-2 0-1
Sweden-Austria 2-1 1-0
Island-Slovenia 2-4 0-1
Italy-Denmark 3-1 2-0
Denmark-Italy 2-2 0-0
COULD HUNGARY… QUALIFIY TO EC2016 ?
sequence match w d l score
goals receiv
edpoin
ts
1. ROM 10 7 2 1 10 2 23
2. NIR 10 5 3 2 9 7 18
3. HUN 10 4 4 2 8 6 16
4. GRE 10 3 3 4 4 6 12
5. FIN 10 3 2 5 7 9 11
Already played matchesEstimated result
Qualify to EC2016Remount-qualifier
HUN ROM GRE NIR FIN
HUN - 1-0 0-0 1-2 1-0
ROM 1-1 - 1-0 2-0 1-0
GRE 1-0 0-1 - 0-2 1-0
NIR 1-1 0-0 0-0 - 2-1
FIN 1-1 0-2 1-1 2-0 -
THANK YOU FOR ATTENTION!
Jámbor Bence – Szabó Dávid – Jámbor Máté