Raptors Net Rating Question

[Other Scott]

Hey if there's any stat nerds on this site, can you help me out with this discrepancy?

So I was under the impression that NBA.com's net rating is simply points scored/100 possessions (Offensive RTG) - points allowed/100 possessions (Defensive RTG).

The problem is that it doesn't match up with the actual average margin of victory that's posted in ESPN's standings.

So for instance ESPN has an average point differential of +8.7 for the Raptors and +8.6 for the Warriors. Whereas NBA.com's net rating has a net rating of +10.6 for the Warriors and +8.3 for the Raptors.

Now that's okay because the net ratings are adjusted for pace of the team so you'd expect the numbers to be different. Except - the Warriors play at a quicker pace than the Raptors, at about 102.5 possessions a game to about 100.5 possessions a game. So actually when you adjust for pace, the point differential change is in the Raptors favour, if they win by 8.7 points in 100.5 possessions, with two more possessions to match the Warriors pace on average they will win by more points per game.

So there's something going on here that I'm missing explaining a difference of like 2.5 points of net efficiency. Can someone explain what that is?

[DanH]

Hey if there's any stat nerds on this site, can you help me out with this discrepancy?

So I was under the impression that NBA.com's net rating is simply points scored/100 possessions (Offensive RTG) - points allowed/100 possessions (Defensive RTG).

The problem is that it doesn't match up with the actual average margin of victory that's posted in ESPN's standings.

So for instance ESPN has an average point differential of +8.7 for the Raptors and +8.6 for the Warriors. Whereas NBA.com's net rating has a net rating of +10.6 for the Warriors and +8.3 for the Raptors.

Now that's okay because the net ratings are adjusted for pace of the team so you'd expect the numbers to be different. Except - the Warriors play at a quicker pace than the Raptors, at about 102.5 possessions a game to about 100.5 possessions a game. So actually when you adjust for pace, the point differential change is in the Raptors favour, if they win by 8.7 points in 100.5 possessions, with two more possessions to match the Warriors pace on average they will win by more points per game.

So there's something going on here that I'm missing explaining a difference of like 2.5 points of net efficiency. Can someone explain what that is?

A few things.

First, the ratings use an estimate of possessions. Same goes for the pace calculation. They don't count the possessions, they use a formula - I believe it is FGA+TOV+0.44*FTA-OREB. Sometimes this throws the possession count off, because of things like technical free throws, or and-ones, or fouled three point attempts, or flagrant fouls or clear path fouls... They also don't do a great job of accounting for team rebounds (ie when a missed shot goes out of bounds).

Second, teams can end up with imbalanced possession counts on offence and defence. For example, if a team manages the clock such that they get the final possession of every quarter, then that game they will end up with two more offensive possessions than defensive possessions. Usually this balances out over time but it can have an effect for some teams in terms of point differential vs net rating. It's far less a thing for teams than it is for lineups or players, who can have far greater possession count mismatches in a game based on their sub pattern.

NBA.com does also have a page for points per 100 possessions that simply adjusts based on pace rather than trying to count (by estimation) individual offensive and defensive possessions.

[Other Scott]

First, the ratings use an estimate of possessions. Same goes for the pace calculation. They don't count the possessions, they use a formula - I believe it is FGA+TOV+0.44*FTA-OREB. Sometimes this throws the possession count off, because of things like technical free throws, or and-ones, or fouled three point attempts, or flagrant fouls or clear path fouls... They also don't do a great job of accounting for team rebounds (ie when a missed shot goes out of bounds).

Second, teams can end up with imbalanced possession counts on offence and defence. For example, if a team manages the clock such that they get the final possession of every quarter, then that game they will end up with two more offensive possessions than defensive possessions. Usually this balances out over time but it can have an effect for some teams in terms of point differential vs net rating. It's far less a thing for teams than it is for lineups or players, who can have far greater possession count mismatches in a game based on their sub pattern.

NBA.com does also have a page for points per 100 possessions that simply adjusts based on pace rather than trying to count (by estimation) individual offensive and defensive possessions.

But if the way of counting possessions for both pace and efficiency is the same, why is there a significant difference between the actual net rating that we see on the site and the pace-adjusted one - in theory the formula should count the same number of possessions whether that number is being used to give you pace or efficiency ratings, right?

[DanH]

But if the way of counting possessions for both pace and efficiency is the same, why is there a significant difference between the actual net rating that we see on the site and the pace-adjusted one - in theory the formula should count the same number of possessions whether that number is being used to give you pace or efficiency ratings, right?

Pace takes an average of offensive possessions and defensive possessions, treating them as though they are equal, and adjusts per-game plus-minus to fix for that average possession count.

Net rating uses the individual offensive and defensive possessions (and points) counts (estimates), figures an efficiency for each end, and finds the difference of those efficiencies: ie it is in theory (with errors because of the possession estimation) the point differential for the team if both teams had exactly 100 offensive possessions every game.

[Other Scott]

Okay I think I got it. So basically the difference is explained by unequal numbers of offensive and defensive possessions.

2.5 points worth of difference means those numbers have to be pretty heavily unequal though - I'm still fairly perplexed by it.

[DanH]

Okay I think I got it. So basically the difference is explained by unequal numbers of offensive and defensive possessions.

2.5 points worth of difference means those numbers have to be pretty heavily unequal though - I'm still fairly perplexed by it.

The estimation aspect can skew the difference between offensive and defensive possessions for individual teams as well, if they consistently, for example, get more (or less) of those poorly-counted scenarios than their opponents do.

[Other Scott]

Okay so it really seems like the pace adjusted ratings are better right? Like it doesn't seem too bad an assumption that teams get approximately as many offensive possessions as defensive possessions, whereas there seems to be fairly major blind spots in the possession formula.

[DanH]

Okay so it really seems like the pace adjusted ratings are better right? Like it doesn't seem too bad an assumption that teams get approximately as many offensive possessions as defensive possessions, whereas there seems to be fairly major blind spots in the possession formula.

Both are probably about equally wrong for team numbers. Net ratings are probably better for lineups, players, single game numbers, any small sample (like clutch time)... So all things considered I just use the ratings for everything.