Double poisson home/away win system

[carlos50]

I am building a double poisson system for forecasting home and away wins. I won't pick draws as I find them hard to predict. I expect to have a complete and well tested system when Premier League starts, but as I am Danish I thought it would be fun to give picks for the Danish Superliga as well. The system is based on the Maher approach, with attack and defense parameters for each team. Furthermore I have a home/away parameter, and I use Pulleins success rate as an explanatory variable. I use a time parameter to weigh the matches, so that newer matches count more than older matches. These that new parameter estimates are needed for every round, hence the testing process takes longer time. I do all the parameter estimates in SPSS.

To test the system I have created League tables, and averages of home wins, draws, away win, home goals and away goals. Compared to actual results all of this looks reasonably good, a part from it seems that goal scoring are slightly underrated for the best teams and overrated for the bad teams. I have solved this by doing a simple regression of home and away goals with success rate as the basic variable. Based on this formula, I have made a manual correction to the expected goals in excel. It improved the league tables without altering the averages for the home wins etc. I consider this correction to be a temporary solution, and any ideas regarding this will be much appreciated.

I am still considering other parameters as ELO rating, average goals etc. I prefer match stats as variables, as they are easy to access and quantify. I am considering other softer values as top scorer not playing, 1st goalie not playing among others. They are just much more time consuming to create, and I want a solid basic system, before starting to use this kind of parameters.

I will only be using staking of 1 betting unit per bet, and a starting bank of 100 units.

All inputs will be much appreciated and I will update any changes in the system as well as test results here. Test results for Premier League:

	Count Home Win	Profit Home Win	Yield Home Win	Strike rate Home Win
PL 2004/05	63	7.35	11.7%	46%
PL 2005/06	92	12.4	13.5%	50%
PL 2006/07	90	17.18	19.1%	50%
PL 2007/08	98	14.69	15.0%	55%
PL 2008/09	80	1.97	2.5%	44%
PL 2009/10	81	21.40	26.4%	48%
Total	504	53.6	10.6%	49%

The first 5 years was used for modelling as well, so the only true test year is 2009/10, as that data wasnt used for anything else. It was a particilar good season as there were way more home wins than could be expected. I will try to get up test results for the Championship by the weekend.

[carlos50]

Re: Double poisson home/away win system Sunday 18th of July

Danish Superliga - 16:00 CET Randers v Brøndby - Home win (Pinnacle odds 3.26) - 1 betting unit

This bet kind of bothers me as my first bet as Brøndby is my team :D But it makes sence to me that the odds are to high on Randers.

Starting bank: £100 Less day stakes: £1 New bank: £99 Daily +/- £: New bank: Daily +/- %: Total +/- £: Total +/- %: Total bets: Total wins: Winning %: