Can I predict the rest of the Premier League Season?

Of cause I can't, however I have been messing around over the Christmas break and I have come up with a Prediction model for the premier league. As I like to put my neck on the line I am going to publish my prediction of the final league table now.

So I am showing that Manchester United will win the Premier League with Arsenal Second, Man City Third and Spurs retaining forth place at the expense of Chelsea. Other points of note are that LFC finish 9th and so do not qualify for European Football and West Ham avoid relegation at the expense of London Rivals Fulham in the last week of the season.

So how did I come up with these predictions? Basically I built three values based on Home/Away and Combined performance.

I then applied a sliding prioritising to these values based on the time between the game and the current date so that the previous games ratio was 1, the game before applied at 0.9 down to all games over ten weeks ago based at 0.1 these then combined to give a rating for the home and away teams, Then combining the Home/Away score with .25 score from the total combined score I applied these to the results of the last match between the two teams to come to the predictive final score.

So for example tonight’s game against Stoke I have Man Utd with a score of 0.95 and Stoke with an Away score of 0.39 these values are applied to the reverse fixture a 1-2 victory for Man Utd and so the model predicts a 2-0 win for Man Utd.

Now before I head down to William Hill to make my fortune I should say that the model has a predictive accuracy of 53% (On Correct results rather than direct scores) which may sound like a wide margin of error (And it is) but it is considerably better than the first instance of the model which ran at 14% accuracy.

The main limitations of the model are that it will always produce a very low amount of draws preferring to predict direct Home and Away wins (Mostly Home) and the scores predicted are usually low as it can not correctly predict when a team will collapse to a 6-0 defeat.

Finally due to the ratings that I calculate the rating will evolve and change with each week so even after one game the prediction model will adjust all future games and therefore the weekly predictions I am planning to publish (If I get the time and remember) will be highly unlikely to reflect the current predictions as this is based on if the teams play the rest of the season in their current form. For example when I built the model last week it predicted Arsenal Third, Chelsea Forth but now predicts Tottenham Fourth and Chelsea fifth based on the change in Chelsea’s rating after the Aston Villa game.

So in short all of this means basically that there is no point me listing the league table as it has come from the model but it seemed like a good idea at the time.

I will however be publishing my predictions each week so that you can see how I am doing over all, just the results though as the scores are seldom actually correct.

So here goes for this weeks

Blackpool V Birmingham – Blackpool Win
Fulham V West Brom – Fulham Win
Man Utd V Stoke City – Man Utd Win
Everton V Tottenham Hotspur – Tottenham Win
Arsenal V Man City – Draw
Aston Villa V Sunderland – Aston Villa Win
Newcastle V West Ham – Newcastle Win
Wolves V Chelsea – Draw
Blackburn V Liverpool – Blackburn Win
Bolton V Wigan – Bolton Win