The Betterbet model is a proprietary model that aims to model future team performance in league games in the below leagues. The model seeks to look at “predictive performance indicators” that are of more predictive value for a team’s performance than “simple stats” such as results and goal difference. We combine these indicators into team ratings for matches and then look at the difference between ratings to forecast “Model Odds” – and identify betting opportunities.
- English Premier League
- English Championship
- English League One
- English League Two
- French Ligue 1
- German Bundesliga
- Italian Serie A
- Scottish Premiership
- Spanish Primera División
The model uses CSV files downloaded from Football Data to gather raw data concerning matches in these leagues. It then produces a Team Rating using a combination of statistics, weighted according to their long term correlation, for each team in each match for that league’s season so far.
Each team’s Team Rating is then calculated as separate home and away rolling averages over the last 20 gameweeks. If the season has not reached Gameweek 20 in that league as of yet, rolling averages are calculated for the whole season so far, overall, and for home and away games separately. When a team plays at home, that team’s rating in home games is included in calculating their rating, and when a team plays away, that team’s rating in away games is included in calculating their rating.
These ratings are then adjusted for strength of schedule over the previous 20 gameweeks (or the whole season so far if the season has not yet reached its 21st gameweek). We refer to this as an Adjusted Team Rating. These are then combined (either Overall + Home rating or Overall + Away rating) in different weights to provide a Real Team Rating. A Real Team Rating is a match specific rating – if the home team was playing away that day, it would be different.
The differential in Real Team Ratings between Team A and Team B in a forthcoming match is then used to calculate Model Odds for each team. These Model Odds are then assessed against an average of bookmakers’ odds on that game.
(When we say points, we don’t mean pounds – you can adjust the stake on that scale – for example, if you went 0.1 pts per £, you might bet £1/£5/£10. Or £2/£10/£20. Up to you).
If the Model Odds imply a percentage chance of Team A or Team B winning the game which is 5% better than the average of Bookmakers’ odds suggest, then a bet of 10 points is recommended.
If the Model Odds imply a percentage chance of Team A or Team B winning the game which is 15% better than the average of Bookmakers’ odds suggest, then a bet of 50 points is recommended.
If the Model Odds imply a percentage chance of Team A or Team B winning the game which is 25% better than the average of Bookmakers’ odds suggest, then a bet of 100 points is recommended.
The below illustration provides an example of how this staking plan works:
This is a differential staking plan which has been back-tested using info from approximately 44,000 European football fixtures dating back to the year 2000.
How’s the model doing? – The Betterbet model so far this season as of the 8th May 2017 has had a +1.7% yield on all bets since launch in January. The results so far this season can be viewed below:
The model currently does not recommend bets on games in the first 6 gameweeks of a particular league season – this is because not enough data from the current season is available to feed to the model. This is a deliberate part of the design at the moment, although hopefully by next season (2017/18) there will be some “feeding in” of previous seasons data in order to try and make informed decisions on games in the first 6 gameweeks of a league season.
If you like what you’ve read so far, why not subscribe here.