Last summer, after a wildly successful World Cup, I set out to demonstrate that a statistical approach to Barclays Premier League punting can, with minimal time invested, lead the armchair fan to significantly more profitable outcomes than they would otherwise enjoy.
At the time, all I had was a database of results, cut and pasted from somewhere online, and a set of assumptions on which to start creating ELO ratings. Along the way there has been a lot of tinkering and changes of direction, but having profitably paper-traded a relatively stable set of ratings for a few months now, I’m increasingly certain that a healthy profit can be returned by using ELO.
Although I was always relatively confident that the ratings would work, there was always the worry that they wouldn’t necessarily identify any value bets: that the Premier League would be so well-known and well-analysed that the odds available would be a nearly perfect representation of chance in any match. This hasn’t been the case, though, with the League just as prone to baseless discrepancies in odds as any other sporting contest.
For anyone wanting to create their own Elo ratings in time for next season, I certainly wouldn’t put you off. If, as I do, you enjoy playing around with spreadsheets (it’s a wonder I managed to convince someone to marry me), a bit of work upfront will provide you with a relatively easy-to-maintain set of ratings.
Having said that, there are a few websites, such as Football Database, that provide ratings for free. And although my ratings aren’t identical to theirs, and I would question one or two of their assumptions, the differences are not significant enough to affect long-term profitability.
Much more difficult for most punters is how to convert these Elo ratings to “pure” odds that they can use as a basis for identifying value bets. Even a quick search online will reveal that this is a subject that creates huge confusion, with lots of conflicting (and, in my opinion, just plain wrong) advice as to how to do it.
My solution for converting ELO ratings into odds is described below. My spreadsheet for doing this is a little more nuanced than suggested below, but the differences in odds returned aren’t significant enough to worry about. You should find it relatively easy to create a formula in your spreadsheet that does the following:
1. Find the difference in the Elo rating between the two teams.
2. Add 7% to the figure of the home team to account for home advantage.
3. Take the strongest team as having a 35% chance of winning, and then add 1% to that figure for every 10 Elo points they are superior.
4. Convert that into their odds of winning by dividing 1 by their chance of winning (e.g. 1/0.56). You now have an idea of what odds they should be to win the match.
4. Now assume 30% as the chance for a draw. Ignore the first 100 Elo points that a team is superior and then take away 1% for every 20 Elo points that a team is superior above that.
5. Convert that into odds of a draw in the same way as described in 3, above.
6. Finally, finding the odds for an away win is simple: add up the chance of a home win and a draw and subtract this number from 1. This can then be converted into odds in the same way as before.
There you have it! Please post any questions if the above isn’t clear, but I hope it provides a better solution than most of the tosh that you’ll find elsewhere online.
Best of luck with it.