Our approach can best be characterized as data mining. Assessing the breadth of statistical data we have determined which factors, and/or which combinations of factors, are most relevant to formulating an ‘odds-on' assessment of a game's outcome. By testing different ideas about ‘what matters,' we've found things which are both supportive of -- and surprisingly contradictory to -- some widely held beliefs about estimating game odds and fair value in different sports. This is because:
a) It can be difficult to differentiate between which basic statistics or relative value metrics are already "baked into" others, and which aren't
b) Understanding the difference between (i) the ways in which individual contributions contribute to team performance, and (ii) how/when/which information is valuable in formulating a given line.
This last point is precisely why winning at fantasy sports does not equate directly to winning at the Las Vegas sports-books. The sum of the parts is not necessarily equal to the whole (e.g. Phillip Rivers throwing for 350+ yards and 2 touchdowns...and the Chargers still losing a game).
We care about matchups, and we care about our LA rankings. They are based purely on teams' quantitative production, and those aspects of it we find MOST predictive.