Our Models
Predicting Football
Like a lot sports fans, football is right at the top of our list. We became obsessed with perfecting match prediction so we decided to leverage our knowledge and experience to re-frame how we form our opinions on game outcomes, to let them become fully data driven.
Tried and Tested
We reduced games to what we found to be the key predictive factors and optimised our models around these. Each model has it's own unique characteristics and parameter set, but all share that they have been trained on over 20 years of match data across 56 leagues, totaling over 400,000 matches.
Win, Lose or Draw ?
Our models are broken into 2 distinct categories. We use betting parlance of "back" and "lay" to classify them. Our "back" models select a definitive outcome from a game. Our "lay" models select a single outcome that we don't think will happen
Our current models are:
Eagle: Backs a single team to win a match
Falcon: Backs a draw outcome from a match
Hawk: Lays a draw outcome from a match
Robin: Lays a single team from a match
F.A.Q.
Learn a little more
A concise daily table of upcoming match selections alongside bet type, target price and suggested staking
When we say edge, we mean the difference between the market price/probability and the output of MODEL-1. Edge can be positive or negative in our universe
In certain models, counter-intuitively MODEL-1 supports the execution of bets with negative edge. This is rare and may be attributable to inefficiencies in some more illiquid markets
MODEL-1 outputs it’s own price/probabilities for selections. We can compare this to market price and determine whether selection is executable based on the edge rules. If it is, it gets added to the YTD table and the outcome is stored. The overall performance will be the sum of all these outcomes (as a %)
Obviously this is proprietary information. It is a form of multivariate factor analysis that we have tailored for our needs