i
Does sacking a coach eally
help? E idence om a
Di e ence-in-Di e ences
app oach
Gus a o de Souza Machado Fab ício
Disse a ion p esen ed as pa ial
equi emen o ob aining he mas e ’s
deg ee in Ad anced Analy ics
iii
NOVA In o ma ion Managemen School
Ins i u o Supe io de Es a ís ica e Ges ão de In o mação
Uni e sidade No a de Lisboa
Does sacking a coach eally help? E idence om a
Di e ence-in-Di e ences app oach
By
Gus a o de Souza Machado Fab ício
This p ojec is p esen ed as pa ial equi emen o ob aining he mas e ’s
deg ee in Da a Science and Ad anced Analy ics
Supe iso : B uno Damásio
No embe 2021
i
Acknowledgemen s
I would like o exp ess my g a i ude o my supe iso , p o esso B uno Damásio who
ga e me all he suppo and a en ion du ing he esea ch and analysis, p o iding
ma e ials, his ime o discuss abou he me hods, his pa ience, his ision ha helped me
o make his hesis a eali y.
I am ex emely g a e ul o my pa en s M . Jose Ca los Fab ício and M s. Solange
Fab ício o hei us in me, o he ime and e o in es ed in my educa ion, wi hou
hem none o his would be possible. Also, a special hank you o my li e pa ne
Danielle New on o he pa ience and suppo du ing my mas e ’s.
Finally, my hanks o my mas e ’s iends ha made his jou ney mo e in e es ing han
i al eady was, wi h special men ion o Vic o Gonzalez my pa ne in mos o he
p ojec s du ing he mas e ’s. Thanks o you suppo , and pa ne ship.
To all o you, wi hou whom I would no be who I am and wha I ha e achie ed, Thank
You.
Abs ac
This p ojec looks o e alua e i oo ball clubs should o should no change hei coach
in o de o imp o e hei pe o mance in he na ional league. Fo his analysis I selec ed,
h ee o he mos impo an Eu opean oo ball leagues, La Liga (Spain), Se ie A (I aly)
and P emie League (England).
The da a used in his p ojec was aken om he ans e ma k websi e, a la ge oo ball
pla o m. The da a pe iod is om season 2005-06 o season 2019-20 and has in o ma ion
abou indi idual games esul s and squad alue by playe .
The s eps be o e he analysis we e a da a cleaning and consolida ion o he in o ma ion,
c ea ion o new ea u es as a pe o mance measu e and selec ion o cases o in e es o
his analysis based on club and coach p o ile. Nume ic a iables we e s anda dized o
be on he same scale and make di e en seasons compa able. A K-means was applied
o iden i y clubs acco ding o hei in es men s which has a p opo ional co ela ion
wi h pe o mance.
Finally, a di e ence in di e ences analysis was applied o e alua e i a club would
ob ain a pe o mance gain i hey decided o sack hei coach be ween game wel e and
wen y-six o he season a e a poo pe o mance in conside a ion o squad p ice.
As a gene al conclusion, i is possible o conside ha on a e age he clubs in he
ea men g oup and compa ison g oup eco e hei pe o mance a e a pe iod o
unde pe o ming, bu he eco e y o he clubs ha sack hei coach is lowe compa ed
wi h he clubs ha keep hem.
Keywo ds
Di e ence in Di e ences; S a is ics; Foo ball; Da a Sc aping; Clus e ing
i
Index
1. In oduc ion ............................................................................................................. 1
1.1 Thesis Objec i e .............................................................................................. 2
2. Li e a u e Re iew .................................................................................................... 2
3. Theo e ical backg ound ........................................................................................... 4
3.1 K-Means ........................................................................................................... 4
3.2 Backwa d S epwise .......................................................................................... 5
3.3 Di e ences in Di e ences ............................................................................... 5
3.3.1 No a ion and modelling ............................................................................ 6
3.3.2 Unbiased es ima o assump ions .............................................................. 7
4. Me hodology ........................................................................................................... 8
4.1 Da a Sc aping ................................................................................................... 8
4.2 Da a p epa a ion ............................................................................................... 9
4.3 Clubs’ segmen a ion ...................................................................................... 12
4.4 Fil e ing cases o in e es and da a analysis ................................................... 15
4.5 Di e ence in di e ences model..................................................................... 17
5. Resul P esen a ion ................................................................................................ 18
5.1 Se ie A ........................................................................................................ 18
5.2 La Liga ....................................................................................................... 19
5.3 P emie League .......................................................................................... 20
5.4 Resul s O e iew ........................................................................................... 21
6. Conclusion and u u e wo k .................................................................................. 22
7. Limi a ions ............................................................................................................ 23
8. Re e ences ............................................................................................................. 24
9. Appendix ............................................................................................................... 26

ii
Lis o Figu es
Figu e 1 - Di e ence in di e ences es ima ion ............................................................................ 6
Figu e 2 - Sc apping and da a p epa a ion .................................................................................. 9
Figu e 3- A e age numbe o clubs by Fi s coach quan i y o games. .................................... 10
Figu e 4 - Pe cen age o sacked coaches in he inal da a se by league and clus e . ................ 16
Figu e 5 - Boxplo – Pe o mance dis ibu ion o Sé ie A by ime and Clus e ...................... 16
Figu e 6 - Boxplo – Pe o mance dis ibu ion o La Liga by ime and Clus e ...................... 17
Figu e 7 - Boxplo – Pe o mance dis ibu ion o P emie League by ime and Clus e . ........ 17
iii
Lis o Tables
Table 1- Va iables o K-means segmen a ion. ......................................................................... 13
Table 2 - K-means segmen a ion o Sé ie A ............................................................................ 13
Table 3 - K-means segmen a ion o La Liga............................................................................ 13
Table 4 - K-means segmen a ion o P emie League ............................................................... 13
Table 5 - Pe o mance o ea ed g oup in 𝑡0 by League and Clus e . ..................................... 14
Table 6 - Quan i y o obse a ions by league, clus e and ea ed ............................................ 15
Table 7- DiD Reg ession Resul s o Se ie A. .......................................................................... 18
Table 8 - DiD Reg ession Resul s o La Liga .......................................................................... 19
Table 9 - DiD Reg ession Resul s o P emie League ............................................................. 20
Table 10 - Ma ch da a se example ............................................................................................ 26
Table 11 - Playe s da a se example .......................................................................................... 26
Table 12 - Coach consolida ion da a se .................................................................................... 26
ix
Lis o abb e ia ions and ac onyms
DiD Di e ence in Di e ences
Pe o mance To al poin s won di ided by o al possible poin s
Compa ison G oup G oup o obse a ions wi hou in e en ion / ea men
T ea men G oup G oup o obse a ions ha su e ed in e en ion
Wa ning Pe o mance A pe o mance pe cen age ha clubs usually s a s o
sack hei coach.
1
1. In oduc ion
In ecen decades, in collec i e spo s, he impo ance o he coach has been highligh ed.
The coach is he pe son who o e sees aining and eam o ma ion, p epa ing i o a
good pe o mance o ob ain posi i e esul s. The coach is an expe in echnical and
ac ical di ec ion and in he playe ’s psychological and physical de elopmen . In
oo ball his is no di e en , pa o a success ul season o a oo ball club in ol es hi ing
he ideal coach o manage he eam, o ming he bes squad, and equen ly indica ing
playe s o hi e.
In oo ball his o y we ha e a ew examples o coaches ha s ayed o decades a he
same club, such as Guy Roux, who s ayed o o y- ou yea s (1961 - 2005) in cha ge
o Auxe e-FRA, leading he club om he hi d di ision o he i le o F ench League
in 1995. In England, he mos success ul manage in he oo ball his o y, Alex Fe guson,
led Manches e Uni ed o wen y-se en yea s (1986 - 2013) winning a o al o o y-
nine i les.
Bu he his o y has mo e cases o clubs sacking hei coaches du ing he season han
examples o longe i y. Fo example, in he I alian i s di ision du ing he las i e
seasons (2015-16 – 2019-20) an a e age o wel e clubs pe season kep hei coaches
o he du a ion o he season and in he same pe iod in ele en occasions clubs he e
we e h ee o mo e coaches du ing he same season.
When a club decides o sack i s coach be o e he end o hei con ac i causes some
incon eniences. One o hem is a e mina ion ine, ha depends on he con ac be ween
bo h pa ies bu when we alk abou he iches leagues in he wo ld his usually in ol es
a la ge amoun o money. As an example, when José Mou inho was sacked om
To enham in Ap il 2021, he had a con ac un il he end o he season 2022-23, and as
a consequence o his b each o con ac To enham had o pay a se e ance package o
a ound i een million pounds o Mou inho acco ding o he websi e oo ball.london
[13].
This p ojec will analyse da a om he op h ee leagues in Eu ope, La Liga – Spain,
Se ie A – I aly and P emie League - England be ween seasons 2005-06 and 2019-20.
All leagues ha e almos he same s uc u e. The en i e season uns om Augus h ough
o May and he league consis s o wen y clubs playing agains each o he bo h a home
8
4. Me hodology
This chap e ocuses on explaining in dep h he whole analysis p ocess o answe he
p oposed ques ions in his hesis, s a ing om da a sc aping, da a p epa a ion, club’s
segmen a ion, il e cases o in e es , desc ip i e analysis and DiD model applica ion.
4.1 Da a Sc aping
The da a used in his analysis was sc aped om he web si e ans e ma k .co.uk/
T ans e ma k o e s he wo ld's la ges oo ball da abase wi h all in o ma ion on
playe s, clubs, and compe i ions as well as one o he la ges oo ball communi ies and
playing a eas o anyone who wan s o sha e and exchange ideas abou oo ball. An
o e iew o playe ma ke alues in addi ion o news, s a is ics, consul an in o ma ion
and an insigh s. The pla o m was ounded in May 2000 by Ma hias Seidel.
The sc aping p ocess was made wi h R using he lib a y es , who e icien ly sc aped
he da a om seasons 2005-06 o 2019-2020 o he English, Spanish and I alian i s
di ision leagues. Fo each league he e was a wo-sc ap p ocess, one o he game esul s
and one o playe ’s in o ma ion’s.
The game esul s da a consis s o in o ma ion such as da e o he ma ch, ime, home
eam and away eam, hei posi ion on he league able, coach and o ma ion and
a endance and esul . Each line ep esen s one game, an example o his able is
a ailable in he appendix a able 10. This in o ma ion is undamen al o iden i y when
a club sacked a coach du ing he season and calcula e he pe o mance.
The playe ’s in o ma ion consis s o he playe s name, club, age, posi ion, ma ke alue,
bes oo , and heigh . Each line ep esen s one playe du ing a speci ic season. This
in o ma ion will be help ul o iden i y club pe o mance expec a ions based on he squad
alue and o c ea e independen a iables o explain pe o mance. An example o his
able is a ailable in he appendix a able 11.
The diag am bellow ep esen s he p ocess o sc apping he da a om ans e ma k
using R, da a cleaning and ans o ma ion o build he da a ables o each league in his
hesis.

9
Figu e 2 - Sc apping and da a p epa a ion
S ep 1: Using he lib a y es , access he ans e ma k web si e in a speci ic URL o
access he in o ma ion o in e es om wo main URL s uc u es, one o ma ches
in o ma ion ha was composed by ou a iables ha changes acco ding wi h club,
season, and league. The second was o ex ac he playe s in o ma ion and was
composed by ee a iables as season and wo ela ed o he club.
S ep 2: S o e ha in o ma ion in o lis s, whe e each lis ep esen s one speci ic
in o ma ion e.g. he sco e, name o he club playing home, name o he club playing
away.
S ep 3: Using he lib a y idy clean a iable ou o he o ma and ans o ming lis s
in o da a ables, as men ioned be o e, one o ma ches and o he o playe s.
S ep 4: Expo hose da a ables o x ile ha going o be used on Py hon o he analysis
4.2 Da a p epa a ion
This s ep is essen ial o he p ojec , has all consolida ions, alida ions and c i e ia
de ini ions we e made o be used in he analysis. The idea is o agg ega e he in o ma ion
by Season - Club – Manage o calcula e pe o mance, e alua e he cases o in e es and
de ine he h ee a iables o apply DiD eg ession, T ea ed, Time and T ea ed x Time.
As desc ibed be o e, in he able each line ep esen s a game o a speci ic club, ha
agg ega ion was made o c ea e a able wi h club, season, coach, numbe o games, da e
o i s game and las game o he coach. As an example, he able 12 in he appendix,
10
shows ha Ala es had h ee coaches du ing he season 2005-2006. Tha means each line
ep esen s a wo k pe iod o a coach in a club du ing a speci ic season.
F om he numbe o games in he agg ega e able, i is iden i ied which coach was
sacked o no du ing he season and o how long hey managed he club. The c i e ia
a e, i he coach had a wo k pe iod o hi y- ou games o mo e, i is conside ed ha he
s ayed o he whole season, be ween wen y-se en and hi y- h ee games, he coach
s ayed o mo e han hal o he season, be ween wel e and wen y-six games he s ayed
o hal o he season and less han wel e games he s ayed o less han hal o he
season.
Acco ding o his in o ma ion, each coach's wo k pe iod will be di ided in o wo
di e en g oups, he ones whose wo k pe iod las ed a leas hi y- ou games, hese
obse a ions will ep esen he compa ison g oup, which means obse a ions wi hou
ea men du ing he season, and he wo k pe iod ha was in e up ed du ing he season
o s a ed in some poin a e he s a o he season, will be ep esen ing he ea ed
g oup, obse a ions ha su e ed an in e en ion du ing he season.
Figu e 3- A e age numbe o clubs by Fi s coach quan i y o games.
No e: Figu e 3 ep esen s he a e age numbe o clubs o e he i een seasons ha kep
hei coach by ou di e en pe iods o ime and league.
As we can see in igu e 2, du ing a season in Se ie A, only ele en clubs on a e age will
main ain hei coach un il he end o he season, and se en clubs on a e age will sack
11
hei couch hal way h ough he season o be o e. Be ween he h ee leagues, I aly has
he highes numbe o changes.
E alua ing he g oup wi h ea men , he ideal homogeneous beha iou o all clubs wi h
ea men would be, club A du ing season one had coach X o he i s nine een games
and hen coach Y o he las nine een games, whe e coach X ep esen s 𝑡 and coach Y
ep esen s 𝑡. Bu in gene al, his does no happen, he da a base has se e al cases o
clubs wi h mo e han wo coaches du ing he same season by consequence coaches wi h
less han en games in 𝑡 which will be conside ed a low numbe o games o e alua e
pe o mance. To make a ai e alua ion, his hesis will only conside cases whe e he
coach ha s a ed he season (𝑡) had a leas be ween wel e and wen y-six games, o
he cases ha he club had wo o mo e coaches a e he i s coach, i will be conside ed
one single wo k pe iod, in o he wo ds, i will e alua e he pe o mance o he coach
ha s a ed he season and he pe o mance o he club a e he coach was sacked.
The compa ison g oup also need o be spli in 𝑡 and 𝑡, in o de o compa e pe o mance
in wo di e en momen s, in his hesis o he coaches who s ayed he en i e season his
b eak will be he game wi h lowes cumula i e pe o mance be ween i h and wen y-
hi d game. Wi h hese c i e ia he season will be di ided in he mos c i ical momen
ha is a simila condi ion o wha happens wi h he ea ed g oup.
The coach pe o mance is calcula ed di iding he o al poin s won by o al possible
poin s, ha means i he coach won wen y-one poin s in ou een games, his o al
possible poin s will be ou een imes h ee equals o o y- wo, so his pe o mance is
wen y-one di ided by o y- wo equals o i y pe cen . This me ic will be he ou come
o he model.
The o he da a se ha will be used in his p ojec is ela ed o squad, each line has
in o ma ion abou one playe in a speci ic season and club like age, es ima ed alue and
posi ion. This da a se will be consolida ed by season and club, wi h in o ma ion o o al,
a e age and highes alue by playe ’s posi ion, hose new ea u es will be used as
independen a iables o explain pe o mance and c ea e clus e o clubs/ season.
Analysing he e olu ion o he es ima ed ma ke alue o he playe season o e season
we no ice ha his alue has been ising o e ime, compa ing he a e age es ima ed
alue o a playe in season 2005-2006 agains season 2019-2020 we see Se ie A wi h a
jump om 2.36M o 6.44M, wi h mo e han 170% g ow h, La Liga om 3.02M o
12
7.72M wi h 153% g ow h and P emie e League om 3.7M o 10.73M wi h 189%
g ow h. In his scena io he p ice a iables by posi ion would no be e ec i e o p edic
pe o mance once pe o mance is a ixed ange be ween ze o and one and he p ice ha e
been g owing season o e season, he co ela ion be ween p ice and pe o mance would
depend on he yea .
A solu ion o his issue was, once he da a was consolida ed, s anda dise he alues o
each season indi idually, in his way i is possible o ha e he same scale o e he
seasons and exp ess he di e ence be ween clubs inside he season. The s anda dise
echnique was he s anda d sco e o Z - Sco e scale, his echnique is be e explained in
he Theo e ical Backg ound. Wi h his ans o ma ion he co ela ion be ween
pe o mance and o al alue o he squad s anda dized is 0.69 o Se ie A, 0.74 o
P emie League and 0.78 o La Liga, as expec ed a mo e aluable squad ends o ha e
be e esul s.
4.3 Clubs’ segmen a ion
As was commen ed be o e, a mo e aluable squad ends o ha e be e esul s, and
consequen ly i implies g ea e p essu e o coaches o ha e be e esul s. Nobody
expec s ha SPAL wi h an es ima ed squad alue o nine y-se en million and ou
hund ed housand eu os would ha e he same pe o mance o Ju en us wi h a squad
alue nine imes highe in season 2018-19. In o he wo ds, i Massimiliano Alleg i
(Ju en us coach) had he same pe o mance o Leona do Semplici (SPAL coach) o
37% p obably he would be sacked du ing he season, whe eas SPAL kep Semplici un il
he end o he season.
This segmen a ion is o he pu pose o di iding clubs in o clus e s acco ding o hei
squad alue and a e , iden i ying he c i ical pe cen age o pe o mance o he clubs-
season by clus e . The ou mos co ela ed a iables wi h pe o mance will be used o
pe o m a K-means segmen a ion. The K-means was applied o h ee hund ed
obse a ions ( wen y clubs pe season, imes i een seasons.) The Elbow g aph was
used o de e mine he numbe o clus e s, minimizing he a iance inside he g oups and
maximizing he a iance be ween g oup in o de o ha e he mos e icien numbe o
clus e .
13
Table 1- Va iables o K-means segmen a ion.
The segmen a ion gene a ed he ollowing esul s:
Table 2 - K-means segmen a ion o Sé ie A
As showed in he p e ious able, in Sé ie A Clus e 1 has clubs wi h low in es men s
e.g. B escia, Pesca a and T e iso. Clus e 3 has clubs wi h medium in es men s e.g.
Udinese, Fio en ina, and To ino. Clus e s 0 and 2 Clubs wi h high in es men s e.g
Ju en us, AC Milan, and In e . Those wo clus e s ha e almos he same clubs in
di e en seasons, o ha eason Clus e 2 will be inco po a ed o Clus e 0. Tha
happens because in some seasons hose clubs had a la ge di e ence o in es men s
compa e o he o he s.
Table 3 - K-means segmen a ion o La Liga.
Clus e 0 has clubs wi h low in es men e.g. Las Palmas, Le an e, Maio ca and
Numância. Clus e 2 has clubs wi h medium in es men e.g. Se ilha, Valencia and
Villa eal. Clus e 1 has clubs wi h high in es men and is composed jus wi h
Ba celona, Real Mad id, and A le ico Mad id in some seasons.
Table 4 - K-means segmen a ion o P emie League
League
Se ie A To a Value Mean Value Mean Value De ence To al Value A ack
LA Liga Mean Value To a Value Mean Value A ack Max Value De ence
P emie League To a Value Mean Value To al Value De ence Mean Value A ack
K-Means Va iables - S anda dized
Clus e To a Value Mean Value To al Value De ence Mean Value A ack N
03,05 2,85 2,63 2,83 43
10,38 0,39 0,36 0,42 156
24,40 4,42 4,40 4,26 35
31,42 1,30 1,22 1,37 66
Clus e To a Value Mean Value Max Value De ence Mean Value A ack N
0 0,29 0,32 0,43 0,26 202
1 4,71 4,65 4,63 4,45 32
2 1,56 1,58 1,80 1,43 66
Clus e To a Value Mean Value To al Value De ence Mean Value A ack N
01,24 1,33 1,17 1,17 64
14,38 4,24 4,43 4,25 44
20,41 0,42 0,46 0,41 152
33,03 2,69 2,71 2,73 40

14
The P emie League is he league wi h mo e in e sec ion o clubs be ween clus e s, as
an example Manches e Ci y appea s in all clus e s, he eason is he g ow h o inancial
po en ial wi h in es men s o e he yea s. Clus e 2, clubs wi h low in es men s e.g.
Bi mingham, She ield Uni ed and Wes B om. Clus e 0, clubs wi h medium
in es men s e.g. Wes Ham, Sou hamp on and Wol e hamp on; and Clus e 1 and 3
wi h clubs wi h high in es men s e.g. Manches e Uni ed, Li e pool and Chelsea. Those
wo clus e s we e g ouped oge he due o he simila i y o he clubs.
The nex s ep is o me ge he clus e in o ma ion wi h he ea ed da ase , in doing ha ,
i is possible o e alua e he wa ning pe o mance pe cen age o each clus e by
analysing he pe o mance o he obse a ions in 𝑡. Once he wa ning pe o mance o
each clus e is known hese alues will be used o selec he obse a ions o be used in
he DiD eg ession. Fo hose coaches whose pe o mance is equal o o below he
wa ning pe o mance in 𝑡 means ha hey we e wi h hei posi ion in isk, so i is
easonable o compa e hem wi h hose who, in he same le el o in es men s we e
sacked.
Table 5 - Pe o mance o ea ed g oup in 𝑡 by League and Clus e .
As was expec ed, he clubs wi h high in es men s, had a highe s anda d o
pe o mance. On able 5 i is possible o no ice ha La Liga has a highe a e age
pe o mance o he ea ed obse a ions when compa ed o he o he leagues, which
explains he di e ence o in es men s be ween he clubs. Jus ou o cu iosi y, be ween
he seasons 2005-06 and 2019-20 only i e di e en clubs inished he league in he op
h ee, and he league had only h ee di e en champions.
Max P 90 Mean Min
High 60% 57% 50% 39% 9
Medium 46% 44% 38% 21% 16
Low 37% 33% 25% 13% 44
High 70% 70% 65% 60% 4
Medium 57% 51% 42% 26% 11
Low 38% 33% 28% 14% 38
High 67% 66% 50% 31% 8
Medium 56% 43% 33% 21% 10
Low 41% 35% 28% 11% 32
Se ie A
P emie
League
La Liga
Pe o mance
Clus e League Q
15
4.4 Fil e ing cases o in e es and da a analysis
Once he ea ed club/season cases a e al eady p epa ed and he pe o mance by segmen
is known, he nex s ep is il e ing he non- ea ed obse a ions in 𝑡 wi h a simila
pe o mance by clus e in he ea ed g oup in 𝑡 in o de o ha e a ai compa ison. Fo
his il e he pe cen ile 90% will be used. Tha means i a non- ea ed club A in season
X o he P emie League ha belongs o clus e medium in es men wi h a pe o mance
equal o lowe o 43% in 𝑡, his obse a ion will be selec ed o he analysis.
In he same example abo e, i would no be ai o compa e a non- ea ed obse a ion
ha ha e 70% pe o mance in 𝑡, because in no mal condi ions his pe o mance le el
is ou side o he pe o mance ange ha clubs use o sack hei coach.
The inal da abase o he analysis is composed by non- ea ed cases, compa ison g oup,
ha pe o mance in 𝑡 by clus e is lowe han pe cen ile 90% o he same clus e in
he ea ed cases plus he ea ed cases, ea men g oup, ha he manage a 𝑡 had
be ween wel e and wen y-six games.
Table 6 - Quan i y o obse a ions by league, clus e and ea ed
The able abo e shows ha in he clus e high he numbe o obse a ions in non-
ea ed is highe han ea ed, indica ing ha clubs wi h high in es men s and
consequen ly high pe o mance, end o keep hei coach when he pe o mance hi s
he wa ning pe cen age. On he o he side, clubs wi h low in es men s a e mo e likely
o change hei coach. A hypo hesis o ha beha iou can be ela ed o he elega ion,
clubs wi h low in es men s and pe o mance bellow he expec a ions ends o igh
agains elega ion and sack a coach expec ing a be e pe o mance is one o he mos
common e ec s. The plo below ep esen s he pe cen age o sacked coaches wi h
pe o mance below he wa ning pe cen age o he clus e .
Se ie A La Liga P emie
League
High Yes 9 4 8
High No 17 10 38
Medium Yes 16 11 10
Medium No 5 16 24
Low Yes 44 42 33
Low No 26 28 49
117 111 162
Q Obse a ions in
To al
Clus e T ea ed
𝑡

16
Figu e 4 - Pe cen age o sacked coaches in he inal da a se by league and clus e .
No e: Figu e 4 ep esen s he pe cen age o sacked coaches in he inal da ase by
League and clus e .
Compa ing leagues Se ie A a e mo e likely o sack hei coach in ela ion o he o he
leagues in e e y clus e . P emie League has he lowes pe cen age wi h 17% o he
clubs in he clus e High alue. Also, can be no ice ha he clubs wi h lowe in es men s
a e mo e likely o sack hei coaches. A possible explana ion is clubs wi h highe
in es men s ends o hi hei pe o mance goals easie in compa ison wi h clubs wi h
lowe in es men s.
The Boxplo below shows he dis ibu ion o pe o mance by ea ed/non- ea ed case
and ime di ided by clus e .
Figu e 5 - Boxplo – Pe o mance dis ibu ion o Sé ie A by ime and Clus e
17
Figu e 6 - Boxplo – Pe o mance dis ibu ion o La Liga by ime and Clus e
Figu e 7 - Boxplo – Pe o mance dis ibu ion o P emie League by ime and
Clus e .
No e: Figu es 5, 6 and 7 ep esen s he dis ibu ion o pe o mance o club/season by
ea ed and compa ison g oup in 𝑡 and 𝑡 by clus e .
By he dis ibu ion abo e, i is easy o no ice ha ime has a big in luence o bo h cases
ea ed and non- ea ed. When a club has a bad s a o he season acco ding o hei
expec a ions and capabili ies, hey end o eco e a e a ce ain poin . Fo e e y league
and clus e , he median is highe in 𝑡, compa ing wi h 𝑡. DiD will help o unde s and
i he e is pe o mance gain by changing he coach.
4.5 Di e ence in di e ences model
A e p epa ing he da a in he p e ious s eps a di e ence in di e ences model was
applied o each league and he ea u e selec ion p ocess was Backwa d S epwise
24
8. Re e ences
[1] D. Bloyce e al.(Decembe 2008), Playing he Game (Plan): A Figu a ional
Analysis o O ganiza ional Change in Spo s De elopmen in EnglandEu opean Spo
Managemen Qua e ly, Vol. 8, No. 4, 359 - 378
[2] By RT, Die enbach T, Kla ne P (2008). Ge ing O ganiza ional Change Righ in
Public Se ices: The Case o Eu opean Highe Educa ion. J Change Manag.
2008;8(1):2 1-35.
[3] Bes e s, Lucas M., Jan C. an Ou s, and Ma in A. an Tuijl. (2016), E ec i eness
o in‐season manage changes in English P emie League oo ball. De Economis 164:
335–56.
[4] González‐Gómez, F., Picazo‐Tadeo, A.J. and Ga cía‐Rubio, M.Á. (2011), The
impac o a mid‐season change o manage on spo ing pe o mance, Spo , Business
and Managemen , Vol. 1 No. 1, pp. 28-42.
[5] Ma ínez, J. A. (2012), En enado nue o, ¿ ic o ia segu a? E idencia en
balonces o /Chaning a coach, gua an ee he win? Re is a In e nacional de Medicina y
Ciencias de la Ac i idad Física y el Depo e 12 (48), 663-679.
[6] Da id Ca d & Alan B. K uege (1994), Minimum Wages and Employmen : A
Case S udy o he Fas -Food Indus y in New Je sey and Pennsyl ania; The Ame ican
Economic Re iew, Volume 84, Issue 4 (Sep. 1994), 772-793
[7] Callaway, B., & San ’Anna, P. H. C. (2020), Di e ence-in-di e ences wi h
mul iple ime pe iods. Jou nal o Econome ics.
[8] Book. Ad ances in K-means Clus e ing - A Da a Mining Thinking - Ha dback –
2012.
[9] Desboule s Loann (No embe 2018), A Re iew on Va iable Selec ion in
Reg ession Analysis Econome ics 6(4):45
[10] Book. Impac E alua ion in P ac ice (Second Edi ion) Paul J. Ge le , Sebas ian
Ma inez, Pa ick P emand, Lau a B. Rawlings, and Ch is el M. J. Ve mee sch - 2012
[11] Ande s F ed iksson, Gus a o de Oli ei a (Oc obe 2019), Impac e alua ion using
Di e ence-in-Di e ences RAUSP Managemen Jou nal 54(4):519-532

25
[12] Goodman-Bacon, A. (2021). Di e ence-in-di e ences wi h a ia ion in ea men
iming. Jou nal o Econome ics
[13] h ps://www. oo ball.london/ o enham-ho spu - c/news/cos -o -sacking-jose-
mou inho-20416106
[14] h ps://www. ans e ma k .co.uk/
26
9. Appendix
Table 10 - Ma ch da a se example
Table 11 - Playe ’s da a se example
Table 12 - Coach consolida ion da a se
Da e HomeTeam HomeTeamPos AwayTeam AwayTeamPos Fo ma ion Manage A endence Resul TeamMa ches Season
01/10/2000 AC Milan 6.0 Vicenza 16.0 3-4-3 Albe o
Zacche oni 46.836 02:00 AC Milan 2000
15/10/2000 Bolonha 17.0 AC Milan 2.0 3-4-3 Albe o
Zacche oni 34.631 02:01 AC Milan 2000
21/10/2000 AC Milan 8.0 Ju en us 2.0 3-4-1-2 Albe o
Zacche oni 81.954 02:02 AC Milan 2000
01/11/2000 AC Pa ma 15.0 AC Milan 8.0 3-4-3 Albe o
Zacche oni 21.572 02:00 AC Milan 2000
05/11/2000 AC Milan 12.0 A alan a 2.0 3-4-1-2 Albe o
Zacche oni 54.641 03:03 AC Milan 2000
Playe Names Posi ion Numbe Bi h Heigh Foo Joined Value Season Team
Diego
Ca alie iDiego
Ca alie i
Gua da-
Redes 28 01/12/1982 (27) 1.89 esque do 01/08/2010 600 mil 2010 Cesana
Alex Teodo aniA.
Teodo ani
Gua da-
Redes
91 21/09/1991 (18) 1.91 di ei o 01/07/2009 200 mil 2010 Cesana
F ancesco
An onioliF.
An onioli
Gua da-
Redes 1 14/09/1969 (40) 1.87 di ei o 07/07/2009 100 mil 2010 Cesana
Alex Calde oniA.
Calde oni
Gua da-
Redes
33 31/05/1976 (34) 1.82 esque do 07/01/2011 100 mil 2010 Cesana
Aldo SimonciniA.
Simoncini
Gua da-
Redes
86 30/08/1986 (23) 1.84 esque do 01/01/2011 100 mil 2010 Cesana
TeamMa ches Season Manage Ma ch_q Fi s Ma ch Las Ma ch
Ala és 2005 Chuchi Cos 18 27/08/2005 08/01/2006
Ala és
2005
Juan Ca los Oli a
5
15/01/2006
12/02/2006
Ala és
2005
Ma io Luna
15
18/02/2006
13/05/2006
Ala és
2016
Mau icio Pelleg ino
38
21/08/2016
20/05/2017
Ala és
2017
Luis Zubeldía
4
18/08/2017
17/09/2017
Page | i