APBRmetrics

Mike G wrote:Wouldn't it be about as simple as noting in the pbp, that A passed to B, who then attempted a shot?

Two problems with that. The PbPs don't currently record that info (at least the ones that I've seen). And even if they did, it's not enough to simply record who passed the ball; the key question is would that pass have been an assist, if the FG had been made. Only a highly hoopstat-oriented league will instruct its scorekeepers to record THAT information. Hence the need to get the info from video.

The other question is do I credit the passer for the potential assist or debit him?

I debit players roughly 0.74 points for missing a shot. On an assisted 2-pt FG, I give the passer (currently) 45% of the credit, and the shooter 55%. So, should I debit 45% of that 0.74 from the passer for the "missed" assist? Or should I give the passer the credit he would have gotten otherwise, and debit the shooter an additional amount for missing an easier shot? Neither situation seems entirely appropriate. I suppose that it could be decided by how much one way or the other improves out-of-sample predictions.

My prediction would be that potential assists correlate so well with actual assists, that it wouldn't add much to a model that estimates them as such.

A snippet from the past:

Harold Almonte



Joined: 04 Aug 2006
Posts: 616


PostPosted: Sat Jan 27, 2007 10:06 am Post subject: Reply with quote
Quote:
The other factor that most stands out to me is the general agreement that Nash's playmaking skills are what make the offense run, and that those skills are more important than any other player's. (Why this means that Stoudemire's and Marion's skills are any less important, I have no idea, nor do I understand why defense never enters the equation.) My thinking, which could be changed, is that even if a player's a skilled distributor, he has to rely on the skills of his teammates as well as his own, and that they deserve some measure of credit for his success.

I'm agree with the assessment that the passer deserves about 1/3 of an assisted basket points, Scorer's skills (shooting, get free, jump, etc.) deserves the rest. He also deserves 1/3 of FT's points, if the potential assists derived into a foul, but he also deserves 1/3 of penalization if he passed to a failed, blocked, stealed contested shot. To pass is as risky for loosing the possession as scoring, and ballhandling is as valuable basketball skill as height. Nash might be responsible of about more than 30% of Phoenix offensive, but defense downgrade him to normallity. Phoenix plays at a high pace, but they have a very good defensive efficiency, and that's no because Nash. Probably every top 5 PG would be MVP with Phoenix.

mtamada wrote:

Mike G wrote:Wouldn't it be about as simple as noting in the pbp, that A passed to B, who then attempted a shot?

Two problems with that. The PbPs don't currently record that info (at least the ones that I've seen). And even if they did, it's not enough to simply record who passed the ball; the key question is would that pass have been an assist, if the FG had been made. ..

Right. I'm suggesting that the scorekeepers could record that info.
With this very simple act, we could determine how effective each player's passes have been; how much a pass benefits a given shooter; the average value of an assist, team by team, etc.

When is a pass to a guy (who then shoots) not a potential assist?
If the same scorekeeper who is granting assists is recording these passes, then the standards should be consistent, no?

Hi all, I was interested in re-addressing this discussion and wanted to make a few points:

1. to Mike G above, the reason that doesn't work is that if A passes to B, who then holds the ball for some time (or dribbles) and then effectively creates his own shot, this wouldn't be an assits/potential assist even though play by play data would show A passes to B who attempts shot.

2. This link was provided earlier in the thread but just reposting it: http://82games.com/assisted.htm Does anyone know what kind of sample they were using for this data, what time period, sample size, etc.?

3. Using the data from 82games (or better data from a larger sample), I think you could conduct some useful analysis on assists/passing by looking at the eFG% of the players that Player X passed to over a given time frame and seeing whether there is statisically significant difference from league avg. Thus, if we were 10 games into the season and you thought your team's PG's assists were down, you could look at this data and if the players he was passing to in potential assist situations had eFG% in potential assist situations that were lower than avg (and not statistically significant) then you can reasonably assume that Player X's assists will increase in the near future (regression to the mean).

4. Found this formula in an old recovered threat "Some Rules of Thumb": Potential Assists = AST * (0.5 * PTS/FGM) / TS%
Does anyone know how this formula is derived and what it represents?

Hopefully that all made some sense. Thanks.

Knicks2832 wrote: 1. to Mike G above, the reason that doesn't work is that if A passes to B, who then holds the ball for some time (or dribbles) and then effectively creates his own shot, this wouldn't be an assits/potential assist even though play by play data would show A passes to B who attempts shot...

Never mind the pbp. Consider this analogy:
- LeBron shoots from beyond the arc. The ref holds up one hand, indicating that it's a 3-point attempt. If it goes in, the scorekeeper records the attempt and the make; if it doesn't go in, he just records the attempt.
- Wade passes to LeBron, who immediately shoots. Scorekeeper holds up his thumb.
If the shot goes in, he records a FG and FGA for LeBron, and an Assist for Wade; if it doesn't go in, it's a FGA for LeBron and a 'potential assist' for Wade.

The same situation that defines an Assist also defines a potential Assist. It just depends on whether the shot is made.

In that case yes, I agree with you. That information could potentiall be very useful. Does anyone know if the analysis done at 82games would be useful for this?

I did some work on potential assists for the Warriors a while back. Just gives an idea of what could be done and what kind of info would be obtained:

http://www.goldenstateofmind.com/2011/7 ... ed-spot-up