I’ve been tinkering with the algorithm model over the past week or so.  The result is lower risk and higher yield - the perfect combination.  This is of course based on the results output from the new algorithm.  In some way this goes without saying but I think it’s always good to remember this point.  If the model’s wrong then performance will be worse than predicted.  How much worse?  The trouble is you cannot measure what you don’t know!
Continued

When I tell people about how I bet on the Premier League the question I am most often asked is, “Er, what’s an algorithm?”.  Well, it’s nothing more than a series of calculations.  They don’t have to be complex, they just follow one after the other.  The output from one calculation is used for the input to the next and so on.  Excel is a great platform for entering and holding the orginal input data, performing calculations and displaying and analysing the results.

Continued

Hindsight is a wonderful thing.  Indeed it’s almost up there with seeing into the future.  Whilst I do not know what the current algorithm model will do in the future I can see how it performed in the past with the benefit of hindsight.  A somewhat fruitless task but fun nonetheless.

So using the existing model I use week to week I amended the stake sizes (same fixed % for all games) to maximise the fund value as at today.  The model begins betting from the 2000/1 season and so includes 7 1/2 seasons including 2007/8 (5718 bets on 2859 matches!).  The average gain per season (counting this 1/2 season as a full season) is 2774%; i.e. profits of over 27x the starting fund each season.  The maximum return in a season was 9594% in 2003/4 and minimum was 137% in 2002/3; i.e. no losing seasons. 

Continued

It’s a nervous weekend of games after last week’s losses.  Although 2 losing weeks on the trot is not an unknown for the model it is a remote event and one I do not want to see played out this weekend.  Therefore I have more than my fingers crossed that we get back on track with this weekend’s selections.

I have made some changes to the algorithm, albeit minor.  It was just a bit of tweaking of the input variables which has resulted in improved performance over all seasons and more importantly, this season. 

The best value today is backing Birmingham away offering 90% value at Bolton whereas the closest odds produced by our model versus the market is Arsenal with just 4% value.   One point of interest is the model lays Portsmouth which is the first time in recent times the market odds have been priced too short to offer value in backing Pompey.  This is likely due to Portsmouth’s loss last week combined with the very high rating the model has for Liverpool.

DATE HOME AB Odds Ex Odds Value  AWAY AB Odds Ex Odds Value
22/12/07 Arsenal 1.49 1.55 4%  Tottenham 6.44 7.80 21%
22/12/07 Aston Villa 1.31 2.04 56%  Man City 9.71 4.80 -51%
22/12/07 Bolton 3.55 1.88 -47%  Birmingham 2.64 5.00 90%
22/12/07 Fulham 1.46 2.22 52%  Wigan 6.73 3.90 -42%
22/12/07 Liverpool 1.33 1.58 19%  Portsmouth 9.12 7.80 -14%
22/12/07 Middlesbrough 2.01 2.30 14%  West Ham 3.91 3.50 -10%
22/12/07 Reading 1.32 1.88 42%  Sunderland 9.29 5.20 -44%

The 2 Premier League funds diverged this weekend.  Whilst the Alpha Fund took a slight loss, the Beta Fund managed a moderate gain.   As the Beta fund is an enhanced version of the Alpha fund it was good to see the Beta fund continuing to outperform.   See the stats below for relative performance of the 2 funds.

It was actually a very pleasing result for the Beta fund as it overcame the setback of laying Bolton v Man U (Bolton winning 1-0).  At market odds of 8.4 the win for Bolton was a rare event; even rarer at the Algorithm’s price of 11.85.  Of course, these rare events are by definition expected and therefore cannot be excluded as ‘unlucky’ or ‘one-offs’.  It’s just good to see the other match bets more than offsetting the loss in the same week.  The weekly gain is however easy to comprehend once you note the Algorithm backed all 5 winning away teams this week. View the detailed Beta fund results here.

PERFORMANCE OF ALGORITHM BETTING PREMIER LEAGUE FUNDS:

Weekly gain/loss in the fund (movement in the fund since last week): Alpha -2.9%, Beta +4.8%

Winning weeks (number of weeks showing gains/ total number of weeks since fund started): Alpha 12/14, Beta 7/7

% gain to date (% increase since the fund started): Alpha 37.2%, Beta +121.6%

Risk (max weekly gain/max weekly loss, insufficient data for standard deviations): Alpha 6.9%/-2.9%, Beta +49.6%/+4.8%