Official Poisson Discussion Thread

[SpodoKomodo]

So there are hundreds of Poisson betting articles out there, I thought it would be a nice idea to collate the information and tips we all have on using Poisson for sports betting, to see where the advantages/disadvantages and possible values lies. What I have done so far:

• For testing purposes, I am using this season's goals scored and goals conceded stats to create my expected numeric value (mean). I am doing the conventional: (Home team goals scored at home + Away team goals conceded away)/2. And vice versa.
• I am using these numbers to create the probability for each correct score (multiplying chances of each team scoring each number of goals).
• I then add the data of all the home win correct score odds to create the home win odds, and then the same for draw and away win.

My queries/worries so far:

• I am only calculating chances for scorelines up to 4 goals for each team, meaning that the odds for home win, draw and away win add up to around 98%, instead of 100%. Do other people calculate chances of scorelines up to, say, 10-9, just so the data is completely clean and not skewed?
• Is there a simple way of importing the "best odds available" for each match to allow easy comparison between "my odds" and bookmakers odds? (I'm aware I've asked this, and that is probably requires hard-coding, but I'm guessing it's something other people would like to do too!)
• And the question that is bothering me most: I am aware that by just using this season's scores as data, or the previous 20 matches etc, that I will not gain any edge over the bookmaker, as I am not taking into consideration any other factor. How would one go about incorporating a ratings system into the Poisson means? I cannot get my head around how you can use anything other than PURE statistical data in Poisson. For example, if I'm using elo ratings, and Chelsea and have a rating of 18 higher than Burnley, what would I do with this data to alter my mean goal predictions? One idea I had was that, for every point difference in rating between the 2, I could add 0.05 to the goal prediction for the favourite, and subtract 0.05 from the goal prediction of the outsider. Does this make sense?

I am hopeful of finding something that could work - at the moment I am paper trailing a CS value system with Poisson (using basic goal stats) and although my yield is crap, the scorelines have nearly ALL finished within 1 goal of the CS prediction. Obviously this requires more testing, but I still think this is a promising sign. For the foreseeable future, I am going to continue to experiment with Poisson and keep this thread updated with anything I learn/new systems etc. If anyone has any opinion, stories, facts, anything to do with Poisson, I hope you can contribute!

[SpodoKomodo]

Re: Official Poisson Discussion Thread

Match Odds		My odds	Odds
Home		2.331607	2.88
Away		3.414011	2.75
Draw		4.112086	3.40

These are my odds for Scunthorpe v Ipswich. The odds suggest my sums are correct. I just need to concentrate on deciding what data should be used for the Poisson variables.