Do Your Picks Account For Weather And Injuries?
We explain what we do when something unique, outside of our models, occurs. This could include recent injuries or severe weather.
A common question we get is:
How do you account for [angle]?
The angle could be any one of a number of things (and sometimes more than one of these):
- A star player is injured
- It’s a conference/division/rivalry game
- The weather’s going to be bad
- One team is more motivated than the other due to [various reasons].
Current Week Game Prediction Models
If is about a specific statistic, then there’s a good chance our models account for it. We use hundreds of different stats as input to our models, and we look at rolling averages over several different time frames, as well as season averages and previous season data.
However, if it’s something more difficult to quantify — most often an injury, or extreme weather — then our models that predict current-week games rely on betting market info to figure out the proper adjustment.
To do that, the models use betting lines (spreads, totals, and money lines) as inputs. When the markets think one team is favored by a lot more or less than we expect, our models essentially view that as a signal that there may be non-statistical info that should affect our game prediction.
To take the case of an injury, if one team is favored by 3 points, and then a star player gets injured, and the spread moves to -1, our win odds estimates for that game will usually shift lower for the team suffering the injury. The spread provides a kind of “anchor point” for our game winner prediction. Even if our models really like a team, we’re usually only going to project them as a few percent more likely to win than whatever the current market estimate is.
So, while the current week prediction models don’t directly account injuries and other non-statistical factors, we indirectly account for them via betting market info.