Methodology

On/off differentials

Further building on the foundation of possession counting and efficiency ratings, on/off compare a team's performance when a player is on the court versus off the court.

These metrics are specific to players, they are not relevant at the team level.

On-court possession counting

Because play-by-play data contains full possession attribution as well as full substitution patterns, we can very accurately track when a player is on or off the court and count only those possessions for which they are on the court.

Regrettably, there is some gray area here since a player can be subbed out mid-way through a possession. Rather than attempting to implement complex possession splitting logic Hoops Junkie considers the five on-court players at a possession's end to be "on" for that possession.

Off-court possession counting

A player could be temporarily subbed out of the game or could be missing a game entirely. Both circumstances are added to a players "off" sample.

As noted above, if a player is on-court for the start of a possession but is subbed out before the possession ends, they are considered to be "off" for that possession. This is less exact than we might like but is preferable to adding complicated partial possession handling logic.

Noise in the signal

There is inevitably some noise in an on/off signal since each individual player shares the court with their four teammates, not to mention the five players on the opposing team. In addition to this, bench players commonly share the court with less skilful players so their differential can be dragged down. It can be even more nuanced than that though, since bench players will usually play against an opposing team's second unit who themselves are usually less skilled than starters.

There is no real way around this noise but it's worth keeping in mind when interpreting the data. The larger a sample size is, the more the numbers tend to stabilise. The signal is still very powerful.

Offensive rating on/off differential (ORtg diff)

Offensive rating differential compares a team's offensive rating when a player is on the court with their team's offensive rating when that player is off the court. The result tells us how much better or worse the offense performs with that player involved.

For example, if a team has an offensive rating of 118.0 with a player on the court and 112.5 when they are off the court, that player's ORtg Diff would be:

ORtg Diff: 118.0 - 112.5 = +5.5

This suggests the offense is 5.5 points per 100 possessions better when this player is on the floor. A negative ORtg Diff means the team's offense performs worse with that player on-court.

Defensive rating on/off differential (DRtg diff)

Defensive rating differential follows the same logic but for the defensive end. It compares how many points per 100 possessions a team allows with a player on the court versus off the court.

Importantly, a negative DRtg Diff is a good thing. If a team allows 105.0 points per 100 possessions with a player on the court and 110.0 when they sit, that player's DRtg Diff would be:

DRtg Diff: 105.0 - 110.0 = -5.0

The negative value means the defense gives up fewer points when this player is involved, meaning they are making the defense better. Conversely, a positive DRtg Diff means the team allows more points with that player on the court than without them (eg. the team is worse, defensively).

Net on/off differential (NRtg on/off diff)

Net rating differential combines both sides of the ball into a single number. It is calculated as the player's on-court net rating minus their off-court net rating.

Continuing the examples above, let's say a player has an on-court ORtg of 118.0 and an on-court DRtg of 105.0. Their on-court net rating is:

On-court NRtg: 118.0 - 105.0 = +13.0

When they sit, their team has an ORtg of 112.5 and a DRtg of 110.0, giving an off-court net rating of:

Off-court NRtg: 112.5 - 110.0 = +2.5

The player's net rating differential is therefore:

NRtg Diff: +13.0 - 2.5 = +10.5

This is a powerful summary stat because it captures the full picture. A player who boosts the offense by +5.5 and improves the defense by -5.0 is contributing on both ends, and the +10.5 net differential reflects that combined impact. Of all the on/off metrics, net rating differential is generally the most useful single number for evaluating a player's overall influence on their team's performance.

Up next: Garbage timeß