Vous êtes sur la page 1sur 6

Skill and Chance in Association Football

Author(s): C. Reep and B. Benjamin


Source: Journal of the Royal Statistical Society. Series A (General), Vol. 131, No. 4 (1968), pp.
581-585
Published by: Blackwell Publishing for the Royal Statistical Society
Stable URL: http://www.jstor.org/stable/2343726
Accessed: 05/10/2010 11:43

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=black.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Royal Statistical Society and Blackwell Publishing are collaborating with JSTOR to digitize, preserve and
extend access to Journal of the Royal Statistical Society. Series A (General).

http://www.jstor.org
1968] 581

Skill and Chance in Association Football

By C. REEP and B. BENJAMIN

IN associationfootballthe ball is passed fromplayerto playeramong the eleven


membersof a side untila particularplayerloses possession of the ball eitherby
interceptionor tackle on the part of a memberof the defendingteam or by an
infringement of the rulesof the game or by himselfshootingat the defendingside's
goal. We may definean "r-pass movement"as one in whicha playerof side A,
havingjust obtainedthe ball, sets offa seriesof r successfulpasses amongmembers
of his own team afterwhichthereis eithera shot at goal by the rthrecipientor an
infringement or thereis an attempted(r+ 1)th pass whichis intercepted.We note
that a 0-pass movementmeans that A's firstattemptedpass is interceptedor that
thereis a shot at goal withouta precedingpass, as, for example,fromthe penalty
spot. Thereare a numberof factorsaffecting thelikelihoodof a successfulrthpass:
(1) the positionsof the playersbetweenwhom the pass is attemptedand the
defendingplayerswho tryto intercept;
(2) the relativeskills of the playersand the effectiveness and confidencewith
whichthoseskillsare applied at thisparticularstage of the game.
In evenlymatchedteamsplayingundertheconditionsnormallyobtainingin good
class football(for example,in the firstthreeEnglishF.A. Divisions) the second of
thesefactorsdoes not varywidelyfromone attemptedpass to anotherbut as the
attackproceedstheopponentsprogressively disposethemselves to improvethechance
of interception or tackle; theyare in morecompactformationand closerto theball.
The probabilityof an r-passmovementP(r) = [Pl -P2 - P Pr(l Pr+1)] where
Pl > P2 > P3 ***Pr > Pr+l.
As to the formof the functionPr one would expectp1 to -befairlyhighthough
less than unityand Pr to fall rapidlyto some low value beyondwhichthereis little
further decrease;an exponentialformseemslikely.
One of us (C. R.), who has an acute interestin the implicationsfor strategic
training, has compiledcarefulrecordsof actual frequenciesagainstwhichto testthis
theory.Table 1 showsfourobserveddistributions togetherwithvalues Of Pr derived
fromthesedistributions.It will be seen that generallyPr declinesas r increases,if
somewhatirregularly.At highervalues of r whennumbersare smallerthe observed
ratios{P(r - 1)}/{P(r- 2)} becomeirregular and thisproducesevengreaterirregularity
in p..
Investigationof the data suggestedthat P(r) would be a more convenient
parameterto attemptto "smooth"in orderto producea standardmodel. In contem-
platingthis "smoothing"and the shape of the distribution P(r), it seemed from
inspectionthatone wouldexpectP(r) to fitcloselyto a negativebinomialdistribution
and thisexpectationwas justified.
Table 2 showstheresultofattempting to fitnegativebinomialsto thedistributions
of Table 1. Generallytheseare reasonablygood fits;the fitfor the miscellaneous
582 - Skill and Chancein Association
REEP AND BENJAMIN Football [Part4,

TABLE 1
Frequencyof r-passmovesin F.A. matchesand derivedvaluesof
Pr= 1-[[{P(r- 1)}/{P(r-2)}] {(l /pr,)- 1}i

42 First Division 12 First Division 36 Miscellaneous 1I World Cup


matches1957-58t matches1961-62+ matches1965-66 matches1966?
r
No. of Pr No. of Pr No. of Pr No. of Pr
moves moves moves moves

0 10,580 *445 2,331 *380 3,407 *387 1,862 *368


1 6,923 *555 1,629 *620 2,320 *613 1,220 *632
2 3,611 *475 1,035 *571 1,435 570 785 '619
3 1,592 *424 565 *523 767 *534 518 *604
4 608 400 304 *502 422 *533 316 *567
5 280 *427 154 *466 215 *518 153 *534
6 107 *381 60 *420 130 *526 78 *577
7 33 *380 31 *463 62 *456 65 *626
8 9 497 15 400 24 *431 27 503
9+ 11 6 20 39
Total 23,754 6,130 8,802 5,063
Mean 1.00 1.30 1-35 1-56
Variance 1150 2K11 2-52 3-29

t All involvingSheffield
Wednesday.
+ All involvingArsenal.
? One matchwentto extratime.

TABLE 2
Frequencyof r-passmovesin F.A. matches-negativebinomialtests

42 First Division 12 First Division 36 Miscellaneous 11 World Cup


matches1957-58t matches 1961-62+ matches1965-66 matches1966?
r
Actual Ex- A-E Actual Ex- A-E Actual Ex- A-E Actual Ex- A-E
pected pected pected pected

0 10,580 10,542 + 38 2,331 2,224 +107 3,407 3,341 +66 1,862 1,777 +85
1 6,923 7,075 -152 1,629 1,788 -159 2,320 2,407 -87 1,220 1,313 -93
2 3,611 3,542 +69 1,035 1,061 -26 1,435 1,428 +7 785 829 -44
3 1,592 1,572 +20 565 556 +9 767 786 -19 518 503 +15
4 608 653 -45 304 271 +33 422 415 +7 316 293 +23
5 280 260 +20 154 126 +28 215 215 0 153 165 -12
6 107 101 +6 60 57 +3 130 109 +21 78 93 -15
7 33 38 -5 31 25 +6 62 55 +7 65 52 +13
8 9 14 -5 15 11 +4 24 27 -3 27 29 -2
9andover 11 8 +3 6 10 -4 20 22 -2 39 31 +8
Total 23,754 23,805 6,130 6,129 8,802 8,805 5,063 5,085
Mean 1P00 1a303 1-348 1P558
Variance 1-495 2-109 2-519 3-286
p(x 2) > 10 <*0001 > 30 < 01

t InvolvingSheffield
Wednesdayon each occasion.
I InvolvingArsenalon each occasion.
? One matchwentto extratime.
1968] - Skill and Chancein Association
REEP AND BENJAMIN Football 583

seriesis veryclose whilethatforthe Arsenalseriesis bad. But theseare onlyfour


exampleschosenarbitrarily fromthe extensiverecordsshownin Table 3.

TABLE 3
Passingmovedistributions
1953-67

No. of Frequency
Detail matches I
0 1 2 3 4 5 6 7+
1953-54Wolverhampton W. 12 *417 *299 *169 071 *029 009 004 *002
1955-56Miscellaneous 15 395 *295 *167 081 *036 *017 005 004
1955-56Sheffield
W. 42 *433 *297 *158 *068 027 *011 004 *002
1956-57Sheffield
W. 42 *443 297 *152 *064 *027 *010 005 *002
1956-57Miscellaneous 18 *446 *296 *145 *062 *029 *013 *006 003
1957-58Sheffield
W.t 42 445 *291 152 067 *025 013 005 *002
1957-58Miscellaneous 12 414 *284 *158 076 *036 018 007 007
1958 WorldCup 11 *384 *262 *161 090 *051 027 012 *013
1958-59Miscellaneous 33 *398 *290 *170 077 *036 *017 *008 004
1959-60Miscellaneous 41 395 *296 *159 *080 039 *017 007 007
1960-61Miscellaneous 38 *389 *266 *160 090 049 *025 *012 009
1960-61Tottenham H. 12 395 *259 *156 *092 049 *028 *012 009
1961-62TottenhamH. 12 *378 *260 *157 094 055 *031 *012 *013
1961-62Arsenalt 12 *380 *266 *169 *092 050 *025 *010 *008
1961-62Miscellaneous 42 *378 *262 *166 *092 *051 *028 *012 *011
1961-62Burnley 8 *384 *275 *175 *083 040 *022 *012 009
1962-63TottenhamH. 11 400 *251 *160 *092 050 *026 *012 009
1962-63Miscellaneous 42 *396 *268 *160 *086 045 *024 *012 009
1962-63Coventry 14 *434 *294 *151 *067 *032 *014 005 003
1962 WorldCup 18 330 *245 *170 *110 *066 *036 *020 *023
1963-64Miscellaneous 19 *391 *268 *153 *091 -047 *027 *013 *010
1964-65Miscellaneous 17 *366 *263 *167 093 *056 *025 *014 *016
1965-66Miscellaneoust 36 *387 *263 *163 *087 *048 *025 *015 *012
1966 WorldCupt 11 *368 *241 *155 *103 *062 030 *015 *026
1966-67Miscellaneous 18 *369 *231 *163 *102 057 033 *023 *022
All matches-Actual 578 394 *275 *162 *084 044 *022 *010 009
Expected+ *398 *275 *158 *084 043 *022 *011 009
A-E -004 - + 004 - +*001 - -*001

t Shownin Tables 1 and 2.


+ On thehypothesis
of a negativebinomial.

In thistablethefrequencydistributionshavebeenstandardizedto a totalfrequency
of unity. Theirgeneralsimilarityis striking.Taking all matchestogether, thefitto
the negativebinomialdistribution is almostexact,and thisis the patternadheredto
by mostof the seriesfrom1958 to 1967. Priorto 1958,thenegativebinomialis still
representativethoughthe parametersare those of the firstdistributionshown in
Table 2. Generaltimechangesin the styleof play would be expectedto affectthe
basic parametersof the distribution but not its mathematicalcharacter.
There are otherregularitiesto be observedin associationfootball,and some of
theseare shownin Table 4. If thepitchis dividedintofourquarterslaterallythenthe
584 REEP AND BENJAMIN- Skill and Chancein AssociationFootball [Part4,

*j, z X o.s.Sz *~ tn N 00 cs m- t- c s tS
00 r 00 "o 'I N

-.

S? .Sk
,j .ON 00 el ON ON - oN - O CO
tN ?, ON o o x
Z: m Bseq m N m N en N m m N " N cq m 6 66

ci~ ~ - . c

QstBRots~~~~~~
.~~~~~~~Os
C30OC
e6
0
01-1a,\ONo
tn * q,
tn r- n
* t m DWO n?
'--
O o ?
o
"tw O

* ^ <> t wo n t tntn t0
m tn tn tn -t WI t mtn t
t~~~~~O X
. . . . . . . . . . . . . . . . . . .

z~~~~~~~~~~~~~~~~ 5 v w ON 'C1 c

- ,
41::

Ci) 0i
t2 i r ri
2 tc0X)riC WO m (i2
+4~~~~~~~~~~~~~~
?Z
C-t
. S
z~~~~~~~~~~~e
Xq
.? tn t
oq cq
t- m
> N N N N N N N N
t
cq cq eq
NON
6 N

z~~~~~~~~~~~~~~~~ . . . 0

m~~~~~~~~~~~~~~~~~~~~~e m WIWI O
XDO 3
4~~~~~~~~~~~~~~~~O
ON_>> < t to ^ E - CD0q\ 00 ON \D t-ON \ r-m
*? s C Z toC ~~~~~. . . . . . N. .q cq. .q cq. .q eq
. cq. cq....o cq N

WY2~~~~~~~~,
C), o000 oo
\000 WI0 0 -ON ,ON I 00 000C)c

k x O~~~~~~~0
?0 "I C) cr- 00 m W) r- 11 o 00 't \0 \o 0 o \, C) In C)
a~~~~~~~ ~~~~~~~W
0 ON --I O
"tW m c
nmm"t
)W IW
q m 1
)W n tnt
vI )m0 -Z 0
at t I n"

O~~~~ S:
. . . . . . . . . . . . . . . . . . . . .

?d 0~~~~~~~~~~~~~~~~~
z~~~~~
ko ~~~~
- u m 7:$ mI:$ I: 7: d u d m m
m u d d d u m

4 -.4 "O - 0 -. Q -.4 -. -. -. -4q Q

.)tn o
t .0 .tn .t W) n tn o ) .t \0 o \1..O ~.O\0 o \
I 1 I 4N
_*t~~~~~~I I. Q, Lt 0
~~~~~~t
W) n'n tn t tn tn t tn tn tn tn \.I \_."C
_

S ~~ ~ ~ ^--~ ~ ~~". 1-- 1- --I -I 1-- -. ".I ".I 1-- " It - I" V- 111 "- 11 -- o
1968] REEP AND BENJAMIN- Skill and Chancein AssociationFootball 585

twoquarterscontainingthegoals are referred to as the"shootingareas". As mightbe


expectedmoregoals come frompassingmovesbeginningin thatarea; theproportion
can be seen fromcolumn(4) to be consistently just about 50 per cent. Column(10)
shows that such shootingarea originattacks are in the ratio of 1: 3 to attacks
reachingthe shootingarea, and fromcolumn(11) we see thatof theseshootingarea
originattacksslightlymore than 50 per cent are regainedpossessions,that is, the
defencefailsto get the ball clearlyaway fromthe shootingarea. The size of these
proportionsare not surprising but theirnearconstancyis remarkable.Again column
(6) showsthat30 per centof regainedpossessionslead to shotsat goal.
It can be seenfromcolumn(8) that15 percentofall attacksreachingtheshooting
area lead to shots at goal, but fromcolumn (7) that of attacksoriginatingin the
shootingarea 22 per centor so lead to shotsat goal.
Column (9) refersto "own half" breakdown,that is, to passingmoves which
breakdown in the attackers'own halfof thepitch. Of all goals scoredagainstthem
50 per centcome fromsuchfailuresto movetheball intothedefenders'half.
Finally,column(5) showsthatwithrareexceptions(forexample,the 1966World
Cup series)it takes 10 shotsto score 1 goal.
The observationthatthereis a stochasticelementin the numberof goals arising
froma particularnumberof shotsin one match(as wellas a near-constant proportion
overa largerseriesof matches)is easyfora statisticianto accept; indeedhe wouldbe
surprisedif it wereotherwise.It indicates,of course,thatan excessof shotsby one
team does not mean that,by chance,the otherside will not getmoregoals and thus
win the match. All this is so far removedfromcurrentsoccer beliefsand tactics
thatgeneralacceptanceof therandomelementhas been inhibited(thoughone of us,
C. R., has shown that a successfulstyleof play can be built upon it). It seems,
however,thatchancedoes dominatethegame and probablymostsimilarball games.

Centres d'intérêt liés