PFF recognizes that context is crucial in evaluating player performance, and that grading on a 9-point scale can overlook important contextual factors. To address this issue, raw player grades are adjusted to account for the expected performance in specific game situations using a mathematical framework.

For example, if a player is in a historically favorable situation, his grade may be adjusted down slightly, while a player in more unfavorable circumstances may get an adjustment in the other direction.

To determine the baseline expectation for each player on every play, PFF collects over 135 fields of data and normalizes the grades accordingly. For instance, headers generally receive lower grades than other shots because they are more difficult to execute, and are therefore normalized to reflect this difficulty.

To establish the expectation for each grade, PFF uses statistical models. These models learn what typically happens in a given scenario for each facet of play by using characteristics of the event as explanatory variables and the given play-by-play grade as the response variable.

Each facet has its own model, which is currently estimated for passing, crossing, shooting, ball carrying, challenge, clearance, carry defending, blocking, and shot stopping. For every grade, the model estimates its expected grade based on the circumstances of the event. The given play-by-play grade is then adjusted with the expected grade before the grades are converted to a 0-100 scale.

By adjusting raw grades for contextual factors, PFF is able to provide a more accurate and nuanced evaluation of player performance, taking into account the specific circumstances of each play.