Using Goal Difference to predict Premier League team points for 2014/15

Using Goal Difference to predict Premier League team points for 2014/15

By Mark Taylor Aug 7, 2014

Tweet

Soccer bettors persistently look for new ways to forecast the Premier League winner. This article focuses on using goal difference to predict 2014/15 Premier League team points as a balanced betting strategy.

With Pinnacle Sports’ pre-season 2014/15 Premier League team points now available to bet on, each market is vacant for counter predictions, based on pure statistical indicators, a more subjective approach or a combination of the two.

As mentioned here there are a number of limitations when using predictive soccer models that bettors should be aware of. The example of Liverpool in 2013/14 or the current exodus of star players at Southampton – which would be absent from most statistical models – should invite caution.

Nevertheless, bettors can aim to make any prediction as good as possible, within the natural limitations of what may occur over a 38 game season.

League points are closely linked to a team’s ability to score and concede goals, conveniently represented by a side’s goal difference.

Occasionally, as so dramatically demonstrated by Newcastle in 2011/12, league points out-perform the more usual relationship between itself and goal difference. Newcastle’s +5 goal difference typical related to 56 points over the season, not the 65 they accrued.

In the majority of cases where such atypical correlations occur, goal difference, rather than points is the better indicator of future performance.  So instead of using previous point tallies to predict those in the future, the use of goal difference may improve the accuracy of a model to predict future league points.

Predicting 2014/15 Premier League team points using goal difference

The simplest way of predicting goal difference for the upcoming season is to take the figure from the previous season and use the historical relationship between points and goal difference to project a point total for 2014/15.

This approach would have indicated that Newcastle were unlikely to repeat their 65 points in the 2012/13 season – they actually managed 41 and narrowly avoided relegation.

Bettors can attempt to improve matters further by taking a weighted average of the goal difference from more than one previous season, thereby increasing sample size.

This method excludes promoted or recently promoted teams, but does reduce the chances of their predictions being skewed by an attacking or defensive unit enjoying a particularly good or bad streak.

If bettors take the Premier League data since its inception and run a regression, goal difference in the next season is adequately predicted by multiplying the previous season’s goal difference by 0.658, the season prior to that by 0.208 and that from three seasons ago by 0.098. These three figures are then added together and finally 1 is subtracted.

Similarly, the general relationship between goal difference over a season and points gained in that season – obtained again by regressing of historical outcomes from previous Premier League seasons is:

League Points = (0.6455 * GD) + 52.

As an example, Liverpool’s goal difference in 2013/14 was +51, in 2012/13 it was +28 and in 2011/12 it was +7.

Therefore, Liverpool’s predicted GD in 2014/15 = (0.658*51) + (0.208*28) + (0.098*7) -1 = +39

If bettors now put our predicted Liverpool goal difference into the second equation, they can project how many points Liverpool might expect to gain in 2014/15.

Therefore the projected league points for Liverpool in 2014/15 =

(0.6455*39) + 52 = 77 points

2014/14 Premier League points prediction

Team

Expected Points 2014/15

Expected GD 2014/15

Actual GD 2013/14

Actual GD 2012/13

Actual GD 2011/12

Man City

87

+55

+65

+32

+64

Liverpool

77

+39

+51

+28

+7

Chelsea

76

+37

+44

+36

+19

Man Utd

70

+27

+21

+43

+56

Arsenal

69

+26

+27

+35

+25

Everton

63

+17

+22

+15

+10

Tottenham

57

+8

+4

+20

+25

Swansea

50

-3

0

-4

-7

Stoke

46

-10

-7

-11

-17

WBA

44

-13

-16

-4

-7

Newcastle

42

-16

-16

-23

+5

Sunderland

41

-16

-19

-13

-1

Aston Villa

38

-22

-22

-22

-16

The table above shows the numbers for all teams that have spent at least the three previous seasons in the EPL. By going back three seasons, the effect of one lucky or unlucky season is reduced, and by using goal difference to determine likely points, a season where a team gains points from goals arriving in a particularly fortuitous order, is therefore eliminated.

However, as mentioned here the estimate is still a probabilistic one and the relevant limitations must be considered.

A team may be as good as their projection suggests, but may still record less likely actual outcomes and possibly overshoot the model, as Liverpool perhaps did in 2013/14 and Newcastle in 2011/12.

The prediction also fails to consider managerial changes, tactical alterations, new transfers or fixture congestion. A first season under Louis van Gaal with no European competition to distract the team from the league or a second year under Jose Mourinho may elevate their respective sides above a purely statistical projection.

Bettors should therefore view their models, such as predicted league points, as the mid-point of a range of potential outcomes rather than rely solely on their findings.

Click here to bet on the 2014/15 Premier League futures.

If you have feedback, comments or questions regarding this article, please email the author or send us a tweet on Twitter.