APBRmetrics

I took a quick look whether in-game scoring momentum exists in the NBA
Specifically I wanted to know whether having scored in your last offensive possession(s) helps scoring in the current possession

One could certainly make the theoretical argument that some sort of momentum should exist: Teams are more likely to score after an opponent turnover or missed shot (+DefReb). Thus, not having scored makes it easier for your opponent to score (they don't have to take the ball out of bounds), which, in turn, makes it harder for you to score (and vice versa)

There are several ways of looking at this, I suppose, but as an initial analysis I simply checked for influence of
- having scored in your last possession
- having scored in your last 2 (or more) possessions
on the amount of points you will score this possession

The amount of points scored in those prior possessions were ignored (for now), just the fact whether points were scored was important.
I used the RAPM framework to control for strength of the 5-man-units on offense and defense.
I also adjusted for 'effect of leading', which, by itself, hints at slightly negative momentum

Initial results suggest that there is a positive effect of having scored in prior possessions -compared to not having scored-, but it's not huge:
- Having scored in your last possession (but not last 2) has the effect of scoring ~0.5 points (per 100 poss) more in your next possession
- Having scored in your last 2 (or more) possessions has the effect of scoring ~1 point (per 100 poss) more in your next possession

I think the 2nd finding is interesting, as it's not necessarily intuitive to say "the fact that you scored 2 possessions earlier makes it easier to score now"

As a next step I'll probably look at more than just the last 2 possessions, and I'll include offensive efficiency of those past possessions as well

Great work J.E. Does your RAPM model also work for timeouts? Lets say a team score 5 possessions in a row. Could RAPM work to gauge if a timeout called by the opponent can kill that momentum in comparison to not calling the timeout?

J.E. wrote:...

Initial results suggest that there is a positive effect of having scored in prior possessions -compared to not having scored-, but it's not huge:
- Having scored in your last possession (but not last 2) has the effect of scoring ~0.5 points (per 100 poss) more in your next possession
- Having scored in your last 2 (or more) possessions has the effect of scoring ~1 point (per 100 poss) more in your next possession

I think the 2nd finding is interesting, as it's not necessarily intuitive to say "the fact that you scored 2 possessions earlier makes it easier to score now"

...

This looks like a post hoc fallacy: Couldn't it be because some kind of in-game situation like garbage time or a favorable rotation affects consecutive possessions in the same way?

NateTG wrote:

J.E. wrote:...

Initial results suggest that there is a positive effect of having scored in prior possessions -compared to not having scored-, but it's not huge:
- Having scored in your last possession (but not last 2) has the effect of scoring ~0.5 points (per 100 poss) more in your next possession
- Having scored in your last 2 (or more) possessions has the effect of scoring ~1 point (per 100 poss) more in your next possession

I think the 2nd finding is interesting, as it's not necessarily intuitive to say "the fact that you scored 2 possessions earlier makes it easier to score now"

...

This looks like a post hoc fallacy: Couldn't it be because some kind of in-game situation like garbage time or a favorable rotation affects consecutive possessions in the same way?

Garbage time is, in a way, accounted for

As for your point about favorable rotation: definitely possible. So far, though, I've never seen anyone being able to quantify "favorable rotations".

Does your RAPM model also work for timeouts? Lets say a team score 5 possessions in a row. Could RAPM work to gauge if a timeout called by the opponent can kill that momentum in comparison to not calling the timeout?

I got timeouts in the PbP, so I can definitely check whether there's a positive effect of calling a timeout. I feel this is somewhat connected to NateTG's point about rotations. Maybe a coach takes a timeout to be finally able to sub someone in/out?

As for your point about favorable rotation: definitely possible. So far, though, I've never seen anyone being able to quantify "favorable rotations".

OK, let me try it a different way:
Looking at the previous 2 possessions a team had, there are four possibilities:
No score, no score
Score, no score
no score, score
score, score

Now, if I understand the momentum hypothesis correctly, if there is momentum, then there should be a significant difference in the chance to score between the two scenarios that mix score and no score. Is that correct?

OFC momentum exists on a team and individual level. However you should account for subs, timeouts, hack-a-shaq strategies; opponents' consecutive threes, power dunks, impossible shots, and-1's and opponent's momentum as momentum-breakers. Also momentum is not something you can get into with scoring 2 possessions in a row especially if you don't count what the opponent does on the offensive end.

JE, that's pretty cool, thanks for sharing.

Are you using auto-correlation at all to measure momentum?

I confess I'm not familiar with the 'leading effect'. What is it?

There's a cognitive science literature claiming that momentum doesn't exist, in particular that there is no such thing as a hot hand in basketball. But I'm not sure if that literature ever reached out to a whole team before, it may just be about individual players.

thanks again
Brian

"No Hot Hand" theorists have proven they can convince themselves of anything.