Assisted vs. Unassisted FG%

[Italian Stallion]

I've been able to find data on the percentage of shots each player takes that are assisted or not assisted, but I haven't been able to find data with the actual eFG% of each category for each player.

Does that exist at one of the public sources for individual players and for the league?

[Mike G]

Assisted shots are 100% made.

[Italian Stallion]

Assisted shots are 100% made.

Damn I need a coffee.

[Italian Stallion]

What I'm trying to get at is the value of players that take a lot of unassisted shots and make them at above average rates vs. players that mostly take assisted shots.

One of the things that became apparent in the Denver series was that when their plays/passing broke down they were kind of lost because they didn't have someone that could create a shot with a decent probability of going in very easily.

I don't think the ability to create like that is necessarily a positive unless the player is producing at an above average rate in that situation and is only doing it when a superior shot is unavailable, but if he is, it has value.

I don't have the data that tells me when a player got a great pass and would have gotten an assisted shot, but he missed it. That way I could really tell the difference between let's say Melo's FG% "assisted vs. unassisted" and vs. the league average at both.

[Crow]

Evan's pbp database could probably help with parts of this.

[Mike G]

I've got a ridiculously complex formula that estimates the % of a player's FG which are unassisted.
For the Nuggets in the playoffs, JR Smith's .56 (56%) is tops. Arron Afflalo is 29%. These are just estimates based on other stats.

This unassisted% is used to derive a factor by which I multiply the player's scoring per 36 rate (which is affected by other factors as well).
JR Smith's per36 scoring rate was 21.3, assuming his uAst% was .54
If it were half that, just less than Afflalo's, and all other stats are the same, his Sco rate is 19.1 .

So within this range, a player's scoring contribution (Sco) can vary by some 10%.
JR Smith's total contribution, measured by eWins per 480 minutes, drops from 1.03 to .83, a loss of about 20%, with the lower uAst%.

In this particular series, it seems shot creation carries a higher than normal weight. Shooting% carried a slightly low weight.

Oh, yeah -- you can get individual assisted% here
http://www.82games.com/1011/10DEN6.HTM

JR Smith was 51% assisted, or .49 uAst% during the season. (I had him at .41).
Carmelo was .49 uAst% in Den (.53 by my estimator) and .54 in NY (.54!).
In playoffs, I had him at .51, 2nd to AC (.57) for the Knicks.

[Italian Stallion]

I'm not sure I'm following you, but I'm sure it's me. So be patient.

If a player took 100 shots, made 50, and 30 of his 50 made shots were assisted (30 of 50 = AS% of 60%), that means he made 20 of the other 70 shots.

I could back calculate that kind of thing based on assisted shot% data at 82 games and on Hoopsdata.

Here's the problem:

I know he created the other 20 made shots because they weren't assisted.

But out of the 50 he missed I don't know how many were among the shots he created vs. those where he got a great look form a teammate and just missed.

[DSMok1]

This should help: http://82games.com/assisted.htm

We really have no way of knowing unless somebody tracks it.

[Mike G]

...of the 50 he missed I don't know how many were among the shots he created vs. those where he got a great look form a teammate and just missed.

Right. And for this, we need for 'potential assists' to be counted. I don't believe pbp provides this.

[Italian Stallion]

OK, so we really don't have that data. It's time for the NBA to catch up to us.

Thanks for that link. At least I can make an estimate.

[EvanZ]

Based on the 82games study, I assumed that jump shots have a 9.5 %-point advantage when assisted. I used that to calculate a hypothetical unassisted FG% for 16-23 ft jumpers:

http://thecity2.com/2011/03/20/monta-el ... long-shot/

[mtamada]

OK, so we really don't have that data. It's time for the NBA to catch up to us.

This is one of the basic and obvious things that STATS's SportVU video data ought to be able to provide (for those who have access to the data). It will fill one of the gaping holes in basketball data that even the PbP data can't fill.

Actually any video data could provide this, so maybe Synergy has it already? And come to think of it, I suppose human judgement is needed to determine whether a pass and subsequent missed FG (or fouled shooter) would have led to an assist if the FG had been made. Or maybe SportVU already has algorithms to decide this? Imagine: assists awarded by a computer, instead of a human scorer.

In fact, the computer instead of making a binary yes/no decision about the assist, could (if it has a sufficiently sophisticated algorithm) assign probabilities to the play: a fabulous pass that threads the needle half the length of the court for an easy dunk raised the probability of scoring from say 50% to 99%. Assist man gets +49% x 2 points, or thereabouts. Dunker gets only +1%, although I suppose we should give him some credit for running the floor and getting open.

In contrast, if some mediocre point guard simply hands the ball the LeBron 35 feet from the basket, and LeBron proceeds to drive through traffic for a lay-in, the passer contributed nothing (or maybe we give him a +1% or 2% for not turning it over), and LeBron would get all the credit.

Those are the extreme examples of assisted and unassisted baskets, but with a good algorithm the computer could assign proper probabilities, and proper assist credit, for the in-between cases too.

That's a tall order, my guess is that SportVU does not currently have such an algorithm and it'd be a pretty titanic task to accomplish -- maybe not as hard as teaching Watson to win at Jeopardy, but tough. But video recognition is making huge strides these days, thanks to exponential increases in computing power.

[Mike G]

Wouldn't it be about as simple as noting in the pbp, that A passed to B, who then attempted a shot?
Splitting credit between passer and scorer is another can of worms entirely. It may be my opinion that the 7-footer who sprints the length of the court and leaps 2' above the rim, and dunks, has done a lot more than the guy who flings the ball up there.

[knarsu3]

...of the 50 he missed I don't know how many were among the shots he created vs. those where he got a great look form a teammate and just missed.

Right. And for this, we need for 'potential assists' to be counted. I don't believe pbp provides this.

Has anyone come up with an estimate for 'potential assists'?

I have no clue how you would approach this but I thought I saw a thread from awhile back where Dan Rosenbaum came up with an estimate. (The thread may have been lost when the forum was hacked)

[Crow]

Perhaps the assist thread started by Dan R. on this page of the miscellaneous recovered threads thread?

http://sonicscentral.com/apbrmetrics/vi ... 9&start=60