Football modelling – Trying to gain an edge over the bookmakers is one of the most difficult things to do in the quest to become a profitable sports bettor. Let’s be honest, if it was easy, then we’d see far more people making a living from sports betting. In the Bet Advisor blog, we’ve already spoken a lot about value, finding value and how to try and beat the bookmaker, so in this latest blog post, we hope to take that one step further and offer a bit of an insight in to a popular and credible way to begin to compile your own odds – the first step to being able to accurately find positions in the markets that can offer value. We will give a relatively simplistic view on how you can get started with this process, but bear in mind it can take a long time to refine/tune your models to created a truly profitable tool.
Simply by coming here reading this blog, you’ve already got more desire than the majority of the public when it comes to trying to make money from sports betting. Betting is a hobby enjoyed by a large number of people, but whilst it remains recreational, there is a limit to your success. Taking it to the next level requires work, much like anything in life, but there are ways it can be done at a personally manageable level and give you an advantage, or at least a starting point. What is the starting point? Poisson Distribution.
Poisson distribution is a daunting prospect at first glance, with complex explanations and definitions, so we can try and simplify it. In the simplest sense, Poisson can be used to help estimate the percentage chances of certain outcomes. In football (the example we’ll use in this article), this can help us price up all kinds of markets – Asian Handicap, 1×2, Over/Under.
Arguably the biggest downfall of Poisson is the fact it is built on historical data, and using old data to predict future outcomes is always a topic of hot debate. But, considering the fact that none of us have a crystal ball, it’s a good starting point. Using previous results, you can collate a profile for each team, giving them an attacking and defensive strength value based on how they sit against the league average. For example, if Man City scored 15 goals in their last 5 home games, and the average for this was 10 goals in 5 home games, then Man City would be given a home attacking strength of 1.5 (50% greater than average of 1.00). If they’ve conceded 2 in 5 home games, and the average is 5, then their defensive strength is 0.40 (40% of the average of 1.00). These figures can be used in calculations against other teams to establish the likelihood of how many they will score and how many they will concede.
Here is an example fixture, that will create an expected score line using Poisson:
Man City – Attacking 1.50, Defensive 0.40
Tottenham – Attacking 1.08, Defensive 0.66
Home Team Goal Expectancy = HOME ATTACKING STRENGTH x AWAY DEFENSIVE STRENGTH x AVERAGE HOME TEAM GOALS (LEAGUE)
Away Team Goal Expectancy = AWAY ATTACKING STRENGTH x HOME DEFENSIVE STRENGTH x AVERAGE AWAY GOALS (LEAGUE)
Home Expectancy = 1.50 x 0.66 x 2.00
Away Expectancy = 1.08 x 0.40 x 1.25 (dummy data)
Home Expectancy = 1.98
Away Expectancy = 0.54
Total Expectancy = 2.52 Goals
So, this data alone doesn’t mean much, so taking it one step further, we can (using the Excel POISSON function) calculate a correct scores matrix, that includes the % chance of EACH combination of home and away scores. From this, we can then work out the percentage chance of a home win, draw, and away win. Compare these findings to the odds provided by the bookmakers, and you’re on your way to finding value. If you give Man City an 80% chance of winning (odds 1.25), and Pinnacle have them priced at 1.40. you’ve found a bet that shows good value for money.
Using Poisson, you get a percentage chance for each score line, so can therefore collate the percentages of all games that contain 3 or more goals to find the price the model thinks Over 2.5 goals should be. For Asian Handicap (say -1.50 Man City), you just collate the % chances for all score lines where Man City win by 2 or more, so 2-0, 3-0, 3-1, 4-0, 4-2,4-1 etc. This might give a total chance of 40%. So Man City -1.50 would be priced at 2.50 by your model. Again, compare with your bookmaker to see if it could be a good value bet.
Things to consider:
· Bookmaker Over-round – Bookmakers add a margin to ensure they profit regardless of result, consider adding this to your model to find the kind of price you’d need to pull the trigger
· Sample size of historical data – some people use last 5 games, others use 10, 20 or even 2-3 years of historical data to calculate attacking and defensive strengths. Use what you’re comfortable with, but most importantly use what is most profitable.
· Poisson uses the score as the only real indicator. A side can win 1-0 without deserving it, or can lose to a last minute penalty that shouldn’t have been given. Be sure to try and add some analysis and factor in these kind of things before placing bets
· Team news, changes in team dynamics, changes in management and other factors aren’t considered. They should be, so make sure you filter the bets suggested by your model based on these kind of factors.
It’s only a brief look in to the idea of pricing up your own events, in the future I am sure we will go in to more depth to show you a working example. In the mean time, if you have any queries, get in touch and we’d be happy to help answer any questions you have.