Sie sind auf Seite 1von 3

Rendezvous of the Poisson and Exponential Distributions at the World Cup of Soccer

KEYWORDS: Teaching; Football; Fitting distributions.

Singfat Chu-Chun-Lin
National University of Singapore. e-mail: fbachucl@nus.edu.sg

Summary Data from the World Cup provide excellent illustrations of Poisson and exponential distributions.
that were tied after the regulation 90 minutes were either settled via the golden goal or via a penalty shoot-out if no goal was scored within the maximum extra 30 minutes allowed. The extra-time goals and successful penalties are ignored in this analysis.
~ ~ ~ ~~~~ ~

4 INTRODUCTION 4
HE latest World Cup soccer tournament held in France between 10June and 12 July 1998 attracted global interest, no doubt including that of many probability and statistics students. This article demonstrates that the event contributes interesting illustrations of the Poisson and exponential distributions.
France 98 involved 32 national teams initially drawn into 8 groups (see table below). In the first stage of the competition, teams played each other within their groups. Thereafter the top two teams within each group advanced to the knockout stage that culminated in the final game between France and Brazil.

ILLUSTRATION OF THE 4 POISSON A N D EXPONENTIAL 4 DISTRIBUTIONS


A total of 170 goals was scored during the regulation times of the 64 games. It is reasonable to model X,the total number of goals scored in a randomly chosen 90 minute regulation game, as a Poisson random variable with mean 170/64. Table 1 displays the probability distribution of X and the expected and actual numbers of games ending with various numbers of goals.

4 DATASET 4

The Poisson distribution indeed provides a good fit to The original dataset was obtained from the Sporting the distribution of the total number of goals scored. The Life news organisations website: http://www.sporting-life.com/soccer/worldcup/results/apparent lack of fit with regards to 2 and 3 goals is not a cause for concern as the mean goals rate is between 2 and 3 goals. In fact, the discrepancy between the The Appendix gives the scores and the times of goals expected and actual number of games disappears when (to the nearest minute) for the 90 minutes regulation games with 2 or 3 goals are combined (29.89 versus time of each of the 64 games. The games are listed in 30). Furthermore, the Poisson distribution suffers a the official chronological order in which they were practical shortcoming as the precise durations of 90 drawn for the tournament. In the knockout stage, games
GroupA Brazil Morocco Norway Scotland Group B Austria Cameroon Chile Italy Group C Denmark France s Africa Group D Bulgaria Nigeria Paraguay Group E Belgium Holland Mexico S Korea Group F Germany Iran USA Yugoslavia
Group G

Colombia England Romania Tunisia

Group H Argentina Croatia

I SArabia I Spain

I Japan

Jamaica

60

Teaching Statistics.

Volume 21, Number 2, Summer 1999

Total number of goals per game


0 1

Poisson probability given mean 170/64


0.0702 0.1865 0.2477 0.2193 0.1456 0.1307 1

Expected number of games


4.49 11.94 15.85 14.04 9.32 8.36 64

Actual number of games 5


11 12 18 11 7 64

2 3 4 5 or more

Total

minutes regulation" games vary due to the different amounts of injury time added on at the discretion of the referees. The exponential distribution represents the waiting times between Poisson events, such as goal occurrences. Based on the mean goals rate of 170/64 per 90 minutes regulation game, the mean and standard deviation of the time between goals are expected to be around (64/170) x 90 = 33.88 minutes assuming the appropriateness of the exponential distribution. In the inaugural Brazil versus Scotland game, goals were scored at minutes 4, 38 and 73. In the following game between Morocco and Norway, goals occurred at minutes 38,45,59 and 61. Thus the times between goals are as follows: 34, 35, 55, 7, 14, 2 where 55 minutes represents the lag between the last goal in the BrazilScotland game and the first in the next. This procedure for computing the time between goals was repeated up to the final goal occurring at the 90th minute of the France-Brazil game. I need to mention that not all games were played in the pre-specified chronological order. Between 23 and 26 June inclusive, the two final games pitching teams from the same group were played simultaneously to avoid any potential collusion among the teams. On 23 June for instance, the Group B games Italy-Austria and Chile-

Cameroon both started at 4 p.m. while the Group A games Brazil-Norway and Scotland-Morocco started at 9 p.m. For the 8 groups, there are 28 (= 256) possible permutations of the chronological order of the 16games played. It is interesting to note that the mean time between goals is invariant to the permutations of the 16 games. To see this, observe that the time span between the last goal in the Romania-England game (22 June) and the first goal of the Italy-Norway game (27 June) equals the sum of the times between the goals scored irrespective of the permutations of the 16 games. The standard deviation of the times between goals however varies (slightly) with the permutations. Figure 1 contrasts the actual distribution of times between goals with the theoretical exponential fit for the games as listed in the Appendix. The actual time elapsing between goals has mean and standard deviation 34.04 and 33.46 minutes. These compare favourably with the theoretical figure of 33.88 minutes for both these statistics. More objectively, the nonparametric Kolmogorov-Smirnov test for lack of fit fails to reject the null hypothesis of exponential fit to the time between goals.

+ CONCLUSION +

Probability and statistics concepts become more digestible when illustrations arise from the habitual domains of the Time between goals Actual: Mean = 34.04 mins ; Std Dev=33.46 mins audience. While teaching the Exponential Model Fit: Mean = Std Dev = 33.88 mins Poisson distribution, I have 50 , , discussed the distribution of 40 chocolate chips per cookie, the 0 30 number of pepperoni per pizza 20 slice, the n u m b e r of Aedes e 10 m o s q u i t o larvae per litre of stagnant water, and so on. The 0 present illustrations of the numbers of goals scored and the times between goals during France'98 Time between goals (mins) have wide appeal on account of the global popularity of football. Fig 1. Actual times between goals compared to exponential fit.
Teaching Statistics. Volume 21, Number 2, Summer 1999

61

Appendix
Date Game Score Goal occurrences (minutes)
~

Date (minutes)
23 Junt

10 June Brazil - Scotland

I2
I

-1

4, 38, 73 38,45,59, 61

Italy -Austria Chile-Cameroon Brazil - Norway Scotland - Morocco

11June Italy -Chile

1 Morocco -Norway

2-2 2-2

I
24 Junc

1 Cameroon -Austria
12 June Paraguay -Bulgaria
S Arabia - Denmark France - S Africa
13 June Spain -Nigeria

1- 1

I 1
1
1- 1

21,56 78,83,89

22, 47, 85
13,42,56

0-0
0- 1
3-0 2-3 35,78,90 21, 24,47, 73, 78

France - Denmark SAfrica-SArabia Spain - Bulgaria Nigeria - Paraguay

2-1

1 2 - 2 1 18,45,73,90
6- 1
1- 3

6,18,53,56,81,88,90 1,11,59, 86 4, 19,75,90 7, 70

1 South Korea

- Mexico

1- 3

Holland - Belgium

----I

28,51, 74, 84

25 Junc Holland -Mexico

2-2

Yugoslavia - Iran Jamaica - Croatia

1-0
1-3

-1

Belgium - South Korea Germany-Iran


~~

1- 1

USA-Yugoslavia
27,45,53,69
-

I 2 - 0 I 50,58 I 0 - 1 -1 4

26 Junc Argentina - Croatia


I

15 June England -Tunisia


Romania -Colombia

2-0
1- 0

42,90

Japan -Jamaica

4 5
Romania -Tunisia
46,66 27 June Italy -Norway
1- 1

10,72
18

1-0 4- 1

Brazil - Chile
17 June Chile -Austria
70,90
8, 75,89

11,27,45,68,70

28 June France - Paraguay


Nigeria - Denmark

0-0

Italy - Cameroon

1- 4

18 June S Africa - Denmark


France - Saudi Arabia

13,52 36, 68, 77, 85

19 June Nigeria- Bulgaria


Spain - Paraguay

Belgium -Mexico

21 June Germany - Yugoslavia


Argentina - Jamaica USA - Iran
22 June Colombia -Tunisia

Holland - South Korea

1 I 3I
3 July
43,48,56, 63 37, 41, 71, 79,83 13,54,73, 80

1
I

29 Junc Germany-Mexico

2-1 2- 1

I 3, 12, 59, 76, 77 I 47,75,86


38,49, 90
45 I6,10,16,45

Holland - Yugoslavia
30 June Romania - Croatia

1- 0

27

Argentina - England Italy - France Brazil - Denmark

1 !!!
0-1

2, 11, 26, 50, 60

4 July

Holland -Argentina Germany - Croatia Brazil - Holland France - Croatia

2- 1
0-3
1- 1 2- 1

12, 18, 90

45,80,85
46,87 46,47,70 13,21,36 27,45,90

7 July 40, 84, 87 83 47,83,90

8 July
11July

Holland -Croatia

1-2 0-3

Romania - England

12 July Brazil - France

~~

62

Teaching Statistics.

Volume 21, Number 2, Summer 1999

Das könnte Ihnen auch gefallen