Football odds Generation

In the last blog we used the Poisson Distribution to predict the outcome for Man Utd vs Swansea.  How accurate was our prediction?  Let's compare with the book maker odds for the match.

When a bookie offers 9/1 they really mean 9 unfavourable outcomes to 1 favourable outcome - so that is 1 in 10 chance or a 10% chance of a favourable outcome.  We can translate all the bookies odds to an Implied Probability this way and compare with the probabilities that we generated.  Be careful here to realise that the bookies implied probabilities will add up to over 100%, this is where they make their margin.  For Paddy Power for the premier league matches that we are looking at, the probabilities add up to 108%.

I have applied my calculations to all the first round matches that I have enough data on.  The table below compares my calculated odds vs paddy power's published odds.

You can see that there is a large discrepancy in some case, eg about 4/1 vs 7/1 for Swansea against Man Utd.  But I do think that the bookies would quite happily offer some of my other odds.  For example the odds for Liverpool to draw, Man Utd to draw, Hull to win, Newcastle to draw, Stoke to win.

Obviously the bookmaker takes more into consideration than simply the teams past records - for example new manager, new players, injuries, even the weather.

Also I quite arbitrarily based my predictions on the last 3 seasons.  This could be tuned to more games (to get a more accurate long term average) or less games (to include more relevant data).  The optimum number of games to consider is not obvious but we could work our estimates better by trial and error.

Finally I discarded alot of information in creating my estimates.  So for example the Man Utd prediction I never considered Man Utd's away games.  This would have skewed my prediction (average away goals is not a good predictor of home goals), but they could be included provided some realistic weighting is given to them.

Match OutcomesFoGPaddy Power
HomeAwayOutcomeOddsProbabilityOddsImplied Probability
Arsenal FCCrystal Palace FCArsenal FC5/70.58362/70.7778
Arsenal FCCrystal Palace FCDraw13/40.23564/10.2000
Arsenal FCCrystal Palace FCCrystal Palace FC9/20.18059/10.1000
Liverpool FCSouthampton FCLiverpool FC17/200.53994/110.7333
Liverpool FCSouthampton FCDraw13/40.23647/20.2222
Liverpool FCSouthampton FCSouthampton FC7/20.22327/10.1250
Manchester United FCSwansea City AFCManchester United FC3/40.56974/110.7333
Manchester United FCSwansea City AFCDraw17/50.22507/20.2222
Manchester United FCSwansea City AFCSwansea City AFC39/100.20387/10.1250
Newcastle United FCManchester City FCManchester City FC6/50.45858/130.6190
Newcastle United FCManchester City FCDraw3/10.250414/50.2632
Newcastle United FCManchester City FCNewcastle United FC12/50.29124/10.2000
Queens Park Rangers FCHull City FCQueens Park Rangers FC3/20.403013/100.4348
Queens Park Rangers FCHull City FCHull City FC21/100.32792/10.3333
Queens Park Rangers FCHull City FCDraw27/100.26909/40.3077
Stoke City FCAston Villa FCStoke City FC21/200.488511/100.4762
Stoke City FCAston Villa FCDraw29/100.257023/100.3030
Stoke City FCAston Villa FCAston Villa FC29/100.254923/100.3030
West Bromwich Albion FCSunderland AFCWest Bromwich Albion FC6/50.453813/100.4348
West Bromwich Albion FCSunderland AFCSunderland AFC12/50.29302/10.3333
West Bromwich Albion FCSunderland AFCDraw29/100.25379/40.3077
West Ham United FCTottenham Hotspur FCTottenham Hotspur FC7/40.36136/50.4545
West Ham United FCTottenham Hotspur FCWest Ham United FC31/200.390721/100.3226
West Ham United FCTottenham Hotspur FCDraw3/10.248223/100.3030

Comments

Popular posts from this blog

Bet Unfair

Deconstructing WDL and O/U 2.5 goals odds.

Chrome Extension 1000