Parametric Model on Soccer Scoring time

1. ECO 5120 Econometric theory and application Parametric Model on Soccer Scoring time Presented by: Ko Chiu Yu 05168070 Ng Shou Zhong 05168560

2. Instruments Literature Review Game Theorists View: Skill, Strategy and Passion: an empirical Analysis of Soccer (Frederic, Luca, Aldo, Working Paper, CentER, April 2000) Literature Review Model Data Estimation Conclusion

3. Instruments Literature Review Statisticians views: Maher(1982), Baxter and Steveson(1988), Ridder et al. (1994), Jackson(1994), Fahrmeir and Tutz (1994), Clarke and Norman (1995), Dixon and Coles(1997), Lee(1997), Pollard and Reep (1997), Rue and Salvesen (1997), and Kuonen (1997a,b). Literature Review Model Data Estimation Conclusion

4. Our Approach Combine views of game theorists and statisticians…… Literature Review Game Theorist Statistican Model Number of goal per game Probability of Scoring Data Team skills Current Score Home Field Advantage Simple Possion Mixed Possion Negative Binomial Estimation Conclusion Parametric Survival Model

5. Data Sample Literature Review Model Data Estimation Conclusion

6. Data Set Problem • Compared with count data in previous studies, our studies are having the following particular problems: • Panel data with attrition across club and across players • Censored Scoring Time • Lack of proxy for: • Players ability • Players strategies • Newly promoted clubs effect • Overwhelming strong clubs effect Literature Review Model Data Estimation Conclusion

7. Variables Collected • There are six categories of instruments: • Match specific • Field specific • Strategy specific • Last matches results • Own team specific • Opponent team specific Literature Review Model Data Estimation Conclusion

8. Variables Collected Literature Review Model Data Estimation Conclusion

9. Variables Collected Literature Review Model Data Estimation Conclusion

10. Estimation Model Literature Review Parametric (Weilbull) Survival Model: Hazard function: Model Taking Logarithm of both sides yields: Data Estimation where and x is a vector of independent variables Conclusion

11. Non-Parametric Estimation Empirical Kaplan-Meier Survival Estimate Literature Review Model Data Estimation Conclusion

12. Non-Parametric Estimation Empirical Kaplan-Meier Smoothed Hazard Estimate Literature Review Model Data Estimation Conclusion

13. Estimation Result Literature Review Model Data Estimation Conclusion

14. Parametric Estimation Literature Review Model Data Estimation Conclusion

15. Parametric Estimation Literature Review Model Data Estimation Conclusion

16. Parametric Estimation Literature Review Model Data Estimation Conclusion

17. Parametric Estimation Home effect on Hazard Function Literature Review Model Data Estimation Conclusion

18. Parametric Estimation Home effect on Survival Function Literature Review Model Data Estimation Conclusion

19. Parametric Estimation Current Score effect on Survival Function Literature Review Model Data Estimation Conclusion

20. Parametric Estimation Current Score effect on Survival Function Literature Review Model Data Estimation Conclusion

21. Side-story: Beckham’s effect We would like to test whether the leave of Beckham would fundamentally change the scoring time of Man United. Literature Review From 2003 to 2004 From 2001 to 2002 Model Data Estimation Conclusion

22. Side-story: Beckham’s effect Assuming the presence of Beckham is time-invariant fixed effect to the scoring probability of Manchester United, we tried to use dummy to proxy his contribution. Beckham’s effect is measured by: Literature Review Model Data Estimation in season where Conclusion

23. Side-story: Beckham’s effect Beckham effect is positive to the scoring probability Yet it is not so significant (p ~ 0.5) Literature Review Model Data Estimation Conclusion

24. Side-story: Beckham’s effect Beckham effect is positive to the scoring probability Yet it is not so significant (p ~ 0.5) Literature Review Model Data Estimation Conclusion

25. Possible Extension • The model derived can be extended by: • Including more mid-stream team data • Add strategic variables of formation arrangement • Observe the effect of pre/post half-time • Observe the difference between time to 1st goal and time to 2nd goal • Run panel data across different league in different countries • Evaluation of each player’s marginal contribution Literature Review Model Data Estimation Conclusion

26. References References 1. AD Fitt, CJ Howls and M Kabelka, [2005]: “ Valuation of soccer spread bets”, Journal of the Operational Research Society. C. S. Lam, [2005]: “Survival Analysis of the timing of goals in soccer games”, Hong Kong Economic Journal Monthly, pp.68-pp.71. 2. Isabelle Brocas, Juan D Carrillo, [2002]: “Do the ‘Three-point victory’ and ‘Golden goal’ rules make soccer game more exciting? A theoretical analysis of a simple game”, Centre for Economic Policy Research, Discussion paper series, no. 3266. 3. Myoung-jae Lee, [1996]: “Methods of moments and semiparametric econometrics for limited dependent and variable models”, New York: Springer. 4. Palomino, F., Rigotti, L. and Rustichini, A. [2000]: “Skill, Strategy and Passion: An Empirical Analysis of Soccer”, CentER. William H. Greene, [2003]: “Econometric Analysis”, Prentice Hall, Fifth edition. 5. Wooldridge, Jeffrey M., [2002]: “Econometric analysis of cross section and panel data”, Cambridge, Mass.: MIT Press. Literature Review Model Data Estimation Conclusion

27. THANK YOU