Football Predictions
Football Predictions
BETTING MARKETS
(EMH) is a cornerstone of
financial theory and, in its
simplest form, states that an
investor should not be able to
consistently obtain returns
above the average. Finding a
forecasting model of football
that can generate better-thanaverage or even positive
returns usually results in a
publication for the academic as
an example of a violation of the
EMH, but the proprietary nature
of the models means that the
published ones rarely (if ever)
represent the very best models,
and even less often generate
positive returns consistently.
Features_24_C2201R2308.BK.indd 24
28/5/10 11:41:58
25
online
http://kn.theiet.org/magazine/videos
http://engtechmag.
wordpress.com
The proprietary
nature of prediction
models means that
the published ones
rarely (if ever)
represent the very
best models
popularity of football
fans never really know the
outcome of a match before it has
happened making watching the
game an exciting proposition.
Pundits make careers based
on this uncertainty of outcome
and spend endless hours
contemplating each and every
possible event in a match. There
are countless upsets in domestic
football every weekend, and the
World Cup throws up its fair
share of giant-killings. Given
this inherent unpredictability,
how can we make predictions?
There are two broad
approaches to modelling football
match outcomes. First, one can
predict the result directly
whether the result will be a
win, a draw or a loss. In this case,
the outcome (win, loss, draw) is
an ordinal variable (a win is
better than a draw which is better
than a loss). The model of choice
here is an ordinal regression one,
such as ordered probit. The
output from such a model is the
probability of each outcome,
so that for an upcoming match
one can use the model to calculate
the probability of a win, draw
or loss.
The second approach for
modelling a match outcome is
an indirect method where the
analyst models the exact score
of the game. In this case, the
analyst estimates the probability of each possible number
of goals scored by each team.
One can then infer the probability of a team winning,
drawing or losing the match by
summing the relevant exact
score probabilities.
AN ORDERED PROBIT
Features_25_C2201R2308.BK.indd 25
28/5/10 11:42:16
26 engineering football
Below: A formula to create a past-performance metric,
which takes into account the fact that past results
should be weighted relative to the result and relative
to the strength of the opposition
Right: The past performance metric plotted on a graph.
1
0.8
0.6
0.4
0.2
rankdiff 204
1+
408
rankdiff 204
ppm (result ) = 0.5 +
408
rankdiff 204
0+
408
if result = win
-200
-150
-100
-50
50
100
150
200
-0.2
if result = draw
-0.4
-0.6
if result = loss
-0.8
-1
Number
of wins
My
Estimated
Probability
FIFA
ranking
Bet365
odds
Bet365
scaled
probabilities
SPA
11,633
0.1163
4-1
0.1751
BRA
10,256
0.1026
5-1
0.1459
FRA
9,108
0.0911
10
18-1
0.0461
HOL
8,480
0.0848
12-1
0.0673
ARG
7,787
0.0779
7-1
0.1094
ENG
5,044
0.0504
6-1
0.1251
GER
4,772
0.0477
12-1
0.0673
NIG
3,939
0.0394
20
150-1
0.0058
POR
3,779
0.0378
28-1
0.0302
GRE
3,723
0.0372
12
150-1
0.0058
ITA
3,689
0.0369
12-1
0.0673
MEX
3,432
0.0343
17
80-1
0.0108
CAM
3,330
0.0333
19
100-1
0.0087
SLN
2,751
0.0275
23
250-1
0.0035
AUS
2,457
0.0246
20
150-1
0.0058
SER
2,291
0.0229
16
66-1
0.0131
IVC
2,271
0.0227
27
33-1
0.0257
URU
1,613
0.0207
18
100-1
0.0087
PAR
1,613
0.0161
30
80-1
0.0108
GHA
1,551
0.0155
32
66-1
0.0131
SWI
1,506
0.0151
26
200-1
0.0044
DEN
1,250
0.0125
35
150-1
0.0058
ALG
1,132
0.0113
31
350-1
0.0025
USA
786
0.0079
14
80-1
0.0108
SLK
396
0.0040
38
250-1
0.0035
HON
334
0.0033
40
1000-1
0.0009
SA
245
0.0025
90
150-1
0.0058
SK
170
0.0017
47
200-1
0.0044
JAP
142
0.0014
45
350-1
0.0025
CHI
31
0.0003
85
66-1
0.0131
NZ
26
0.0003
78
2000-1
0.0004
NK
0.0003
106
1500-1
0.0006
Graph.indd 1
Features_26_C2201R2308.BK.indd 26
28/5/10 11:42:32
27
HOT TIPS: THE E&T PANEL GIVES ITS WORLD CUP PREDICTIONS
E&T has gathered a panel of
four experts people with a
professional or personal interest
in the science of predictions to
kindly stick their neck out and
state who they judge will be
involved at the latter stages of
the tournament and who, above
all, will claim the cup.
Dr Ian McHale
Dr Ian McHale is a
Senior Lecturer in
Statistics at the
University of
Salford. His
research interests
include statistics in
sports and the study of gambling
markets. If he had known
footballers and golfers earned so
much money, he would have
ignored his parents and
concentrated less on maths at
school and more on his sporting
activities. Still, mathematical
models no doubt came in handy for
the predictions here.
www.salford.ac.uk
http://sports.ladbrokes.com/en-gb
www.stevenageborofc.com
Prediction
Winner: Spain
Finalists: Spain and Brazil
Semi Finalists: Spain, Brazil,
Argentina and France.
Prediction
Winner: Spain
Finalists: Spain and Brazil
Semi Finalists: Spain, Brazil,
England and Argentina
Prediction
Winner: Brazil
Finalists: Brazil and Argentina
Semi Finalists: Brazil, Argentina,
England and Holland
David Williams
David Williams is
head of PR and
official spokesman
at Ladbrokes plc, a
leading retail
bookmaker in the
UK, Ireland,
Belgium and Spain with over 2,700
Chris Day
Goalkeeper Chris
Day started his
career at Tottenham
Hotspur before
moving to Crystal
Palace and then to
Watford under onetime England manager Graham
Laura Daligan
Psychic, artist and
pin-up Laura
Daligan specialises
in tarot card
readings and dream
interpretations.
She is also a colour
healer, using her artistic skills to
www.lauradaligan-psychic.com
Prediction
Winner: Spain
Finalists: Brazil and Spain
Semi Finalists: England, Brazil,
Argentina and Spain
How did they do?
When the World
Cup kicks off on
11 June, so will
the E&T World
blog. Visit the
Cup blog
blog to see all four experts on our
panel pit their skills against one
game a day, from the opening
match through to the final.
Follow the action online and
see how they do. Why not join
in and make your own
prediction?
http://engtechmag.wordpress.com/
Disclaimer
All predictions are for
entertainment only. Any
mortgages lost or other sums
squandered based on the
information supplied are the
readers own responsibility
Features_27_C2201R2308.BK.indd 27
28/5/10 11:42:58