[[1819:predictions_pl]]

We compare predictions for the outcome of Premier League season 2018/19. Included are only popular sites and those who have a web page for reference of the prediction and published before 10.08.2018. Eventually, at the end of the season, we compare all predictions like as it was done for the Bundesliga 2017/18.

The following 28 sites or institutes have published a prediction or rankings:

Not all predictions are really predictions but rankings based upon results, market value, and other sophisticated models. Those are included as references. Some of them like Goalimpact or Club Soccer Index are rankings, but based on their models a prediction was published.

  • AFC Bournemouth = #AFCB
  • Brighton & Hove Albion = #BHAFC
  • Cardiff City = #CardiffCity
  • Crystal Palace = #CPFC
  • FC Arsenal = #AFC
  • FC Burnley = #Clarets
  • FC Chelsea = #CFC
  • FC Everton = #EFC
  • FC Fulham = #FFC
  • FC Liverpool = #LFC
  • FC Southampton = #SaintsFC
  • FC Watford = #WatfordFC
  • Huddersfield Town = #HTAFC
  • Leicester City = #LCFC
  • Manchester City = #MCFC
  • Manchester United = #MUFC
  • Newcastle United = #NUFC
  • Tottenham Hotspur = #COYS
  • West Ham United = #WHUFC
  • Wolverhampton Wanderers = #WWFC

Below the predictions are tabulated in alphabetically order of their abbreviations.

Team BBCBICDRADSTECIEFPELOEPIESPEXPFAEFBMFDBFORFSTFTEGOIGOPLSLMIRMISSKASKYTDSTHETNSTMVYAH
AFCB 14 13 14 19 13 11 12 14 17 13 12 18 13 17 15 16 19 9 11 19 18 14 13 17 12 16 18 16
BHAFC 15 20 17 14 17 16 16 13 16 15 16 10 14 16 18 18 12 16 16 15 16 15 18 15 18 18 14 17
Cardiff 20 18 20 17 19 19 19 20 20 19 15 20 19 19 20 20 20 17 20 10 20 20 20 19 20 20 20 20
CPFC 13 11 11 15 15 9 9 7 8 12 19 12 7 13 14 7 10 12 14 16 8 13 14 18 15 14 12 12
AFC 6 6 5 5 6 6 6 6 5 4 5 5 6 6 4 5 2 6 6 3 5 6 3 6 6 6 6 6
Clarets 12 16 13 13 9 10 10 12 15 11 8 17 9 15 10 14 11 19 15 12 15 8 10 10 14 13 13 10
CFC 5 4 3 3 5 5 5 5 6 6 3 3 5 5 5 4 4 3 4 5 4 2 5 4 3 4 2 5
EFC 8 7 8 10 8 7 7 8 10 9 14 13 8 7 7 10 7 10 7 8 9 9 11 7 8 7 7 7
FFC 11 14 18 9 12 15 15 15 9 18 20 19 18 12 13 15 14 13 17 13 17 16 8 16 16 12 11 14
LFC 2 2 2 2 4 2 2 3 2 1 2 6 3 2 2 2 6 2 2 1 2 5 1 5 2 2 3 2
Saints 16 12 10 12 11 14 14 16 13 16 9 15 15 9 17 9 8 15 9 14 11 12 9 12 13 15 10 15
Watford 19 17 16 11 18 17 17 18 17 14 18 16 16 18 16 12 17 14 18 17 13 17 15 11 17 17 16 18
HTAFC 17 19 19 16 20 20 20 19 19 20 13 11 17 20 19 19 16 20 19 18 19 19 19 20 19 19 19 19
LCFC 10 10 7 8 7 8 8 9 14 8 7 7 10 10 9 8 15 8 8 9 7 11 16 8 7 10 8 11
MCFC 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 3 1 1 2 1 1 2 1 1 1 1 1
MUFC 3 3 4 6 2 4 4 4 4 3 4 2 4 3 6 6 1 5 3 4 6 3 6 2 4 3 4 3
NUFC 18 15 9 20 14 13 13 11 7 17 10 14 11 14 12 11 18 7 13 11 10 18 17 13 10 11 15 8
COYS 4 5 6 4 3 3 3 2 3 5 6 4 2 4 3 3 5 4 5 6 3 4 4 3 5 5 5 4
WHUFC 9 8 15 18 10 12 11 10 11 7 11 8 12 11 8 13 9 18 12 7 12 7 7 9 9 8 9 9
WWFC 7 9 12 7 16 18 18 17 12 10 17 9 20 8 11 17 13 11 10 10 14 10 12 14 11 9 17 13

Graphically all predictions are ordered by the average prediction. Shown are the average prediction as bars together with standard deviation as error bars in grey. The colored points mark the single predictions of the above mentioned sites. Asterisks ($\star$) mark betting-predictions, triangles ($\triangle$) votings, bullets ($\bullet$) rankings, diamonds ($\diamond$) experts and squares ($\square$) algorithms.

The following chart shows the final Premier League table together with the average predictions as $+$. The red circles mark the final standings. As well, the three best predictions are included (see discussion below), based on the best mean-absolut-error prediction.

For a better comparison we use the difference: $d(m):=\mathsf{rg}_{p}(m)-\mathsf{rg}_{38}(m)$ with $\mathsf{rg}_{p}(m)$ as the average prediction rank of team $m$ and $\mathsf{rg}_{38}(m)$ the rank of team $m$ after match day 38. This deviation from prediction, is a kind of measure, how well they have done! The larger $d$ the better they have performed, negativ $d$ means they have under performed, compared to the predictions. Sorting all teams accordingly from high $d$ to small $d$ is shown in the chart below.

On top are those teams which have over performed and on the bottom those which have under performed. To judge whether it is really over- or under performed so far, those errors must be taken into account. All teams which are placed within this error bars, i.e. from 3.-17. rank, are roughly in the rang one could expect.

So, the really positive surprising teams are really Wolves. The disappointing teams are Burnley, Southampton and Fulham.


The overall deviation of all predictions is best measured by the mean absolut error: $$ mae := \frac{1}{20}\sum_{m=1}^{20} \Big| \mathsf{rg}_{p}(m)-\mathsf{rg}_{38}(m)\Big| = 2.33 \pm 0.50. $$


Below the prediction sites are listed in order of best mean average error mae, together with the correlation r in the first column. Two experts with pretty good predictions so far are leading the chart: Experts from Express (EXP) and Alex Conrad from THE18 (THE), the betting sites Freesupertips.co.uk (FST) and @BettingsIsCool (BIC).

In the above chart you can compare the actual standings in comparison to average prediction and the prediction of EXP, THE, FST and BIC.

A more detailed analysis will be done later!

  • Zuletzt geändert: 2019/05/12 17:02
  • von admin