NBA Chemistry Experiments
Home | Teams/Players | Commentary | Premium Services | Contact Us

Game Charting Insights:
The value of a good pass

One of the obvious problem areas currently in NBA statistics is passing. All we get are assist totals, which make no reference to the type of shot arising from the assist (a wide open dunk versus a contested jump shot!), and most significantly, we only get a record of the passes that succeed. So the 82games charting team set out to study the league wide effects of taking action off an "assist worthy" pass versus creating it all on your own.

Now what constitutes a potential assist? The NBA defines an assist in simple terms as:
Assist: A pass that leads directly to a basket.

Basically if the player after receiving the pass pauses or dribbles around for a while before taking action it's not an assist, but otherwise if the player takes the pass and immediately shoots (catch and shoot), drives to the basket, or has a little pump fake type move to throw off the defense and then goes up for the shot (with perhaps one small dribble even) then you're talking assist.

That's a pretty liberal definition, but the one we will go with for this project:
Potential Assist: A pass that leads directly to a possession event (shot, foul, turnover).

With turnovers it requires even more discretion from the charters in trying to foresee whether the receiving player for the pass would have indeed immediately tried to score (shoot, drive, draw foul) or was really just going to get the pass and then pause and decide what to do next...

Measuring FG% from a "potential assist" pass versus being unassisted
Shot Type
in FG%
of Att
 3-Pointers .379 .342 +3.7% 81%
 2-Point Jumpers .458 .363 +9.5% 52%
 Close Shots .613 .487 +12.6% 43%
 Dunks .910 .840 +7.0% 76%
 All Shots (excluding Tips) .502 .421 +8.1% 56%

So a shot that comes by virtue of an "assist worthy" pass had an over 8% better chance of going in for the sample we looked at! That is of course huge evidence for the value of an assist. In addition, more than half of field goal attempts occurred where our charters would have awarded an assist had the shot gone in.

Interestingly the shot type had quite a lot to do with the expectations -- three point shots were seldom taken without a 'potential assist' pass beforehand, and the gain was just 3.7% in accuracy. This might suggest NBA players seldom settle for a three when they have held the ball/dribbled for a while, or perhaps we should be more restrictive in defining a potential assist such that a "catch and shoot" doesn't automatically qualify.

Since dunks constitute a small sample and are always very high percentage shots to begin with, the heart of the matter then becomes the two point shots where the difference between taking something from a good pass and trying to create on your own are huge! On the surface it appears as if the least efficient shot a player can take is an unassisted two point jumper, with a terrible .363 accuracy and not a lot of fouls drawn either!

The value of the potential assist could also be stated in this fashion:

+.25 points for every Close Shot
+.19 points for Two-Point Jump Shots
+.14 points per Dunk Shot
+.10 points per Three-Pointer

...or +.16 points on average

Of course this is misleading in that the passer had numerous options when he had the ball in his hands (taking the shot himself, driving to the hoop, passing to a better situated or better shooting teammate, etc) and may not have made the optimal decision.

Moreover the passer may deserve far more of the credit for instance on a dunk attempt even though it doesn't improve the FG% as much as some other shot types, since the dunk wouldn't have happened at all without the pass!

Now another common complaint raised on assists is that a pass that leads to free throws gets no credit. Easy enough to rectify! By again tracking which passes are potential assists, we can also trace back the free throws. The same goes for turnovers, although as mentioned above, with the turnovers it's much harder to be sure that something was a potential assist since the intent of the player with the ball is less evident.

Free Throws From
 Shooting Fouls 41% 59%
 Personal/Loose Ball Fouls 18% 82%
 All Fouls Drawn 34% 66%
 Free Throw/FGA .19 .48
 Turnover/FGA .06 .32

This might be a little unexpected, but only a third of all free throws arise from "assist worthy passes" and only 40% of free throws from shooting fouls if we elect to discard the ambiguous/difficult to assess other types of fouls. Given that by our charting over 56% of all shot attempts were "assisted" it means you get a better rate of foul drawing on non-assisted plays. Indeed, it's far batter at over two times the FT/FGA ratio.

Once you think about it, this all makes sense, since assists often lead to wide open shots or simple catch and shoots where the chance of drawing a foul is slim. On the other hand, when a player elects to freelance with a bit of one on one (or in some cases, one on five!) it's more often the case that he's trying to beat his man off the dribble or attack the basket and thereby draws more contact. Moreover, the best "create their own shot" types also happen to be the majority of the scoring superstars and thus have a knack for earning the trip to the charity stripe.

With turnovers the ratio's are so skewed it might be deemed a charter bias since it's pretty hard to determine on many turnovers whether a shot was imminent or not, and consequently it's hard to classify plays as potentially assisted.

If you conjure up a simple points per 100 possessions (omitting the offensive rebound implications), you find it looks like this:

110.4 - w/Potential Assist
  81.9 - not assisted

Again though the turnover issue plays a large part in the discrepancy, and it could be argued for instance that all passing turnovers should be treated as 'potential assists'. The case is thus made for a better definition of how to chart the "assistedness of turnovers" and that's something we will address.

Throw out the turnovers then and just look at a points per shot number and it's:

1.17 - w/Potential Assist
1.04 - not assisted

One more interesting league wide avenue to explore is to break it out by the shot clock timing:

Measuring FG% from a "potential assist" pass versus being unassisted
Shot Clock Usage
in FG%
of Att
 0-10 seconds .522 .454 +6.8% 53%
 11-15 seconds .528 .436 +9.2% 56%
 16-20 seconds .480 .397 +8.3% 54%
 21+ seconds .439 .368 +7.1% 54%

It might have made more sense to separate out the quick 'follow up' shots coming from offensive rebounds since they merge with the fastbreak/transistion/quick shot possessions in the 0-10 seconds usage, but it can be seen that there's a fair amount of consistency across the shot clock spectrum in how a shot that's potentially assisted improves the expected outcome.

With the biggest difference coming in the 11-15 second range it has us wondering how this translates to individual players! How are the assists of Steve Nash and other top dime-men distributed across the seconds? Is there a common pattern with the best passers or does it vary?

That's really the drawback to digging deeper into the numbers -- sometimes you find more new questions than answers to old ones!

Ah, and we can bet you are wondering what the numbers look like for individual players in terms of how their passes improve FG%, lead to free throws, and increase points per possession efficiency.

Well, rest assured we will have more articles expanding on better ways to measure passing, including of course analysis of individual players in the not too distant future...

Game Charters for the "Potential Assist" Project:
Adrian Lawhorn, Allyn Wright, Andre Warner, Anthony Cerminaro, Brad Burnett, Brett Steele, Brian Cole, Brian Ganster, Cameron Tana, Chad Casarotto, Charles Floyd, Chris Hancock, Craig Ward, Dana Henderson, Daniel Kelly, David Mintz, Dmitri Salcedo, Dwayne Killings, Eric Patten, Eric Wallace, Frank Mantesta, Gabe Farkas, Greg Humphreys, Gregg Calvin, Husamettin Erciyes, Jerry Hardin, John Magee, Josh Braby, Lorenzo Pascucci, Mark Reyes, Mike Raak, Mike Wolf, Narbeh Avanessian, Noah Libby-Haines, Noah Purcell, Patrick Clark, Patrick Sheehy, Phil Edwards, Philip Wong, Raj Kannan, Rich Schmidt, Rob Ireland, Rob Stewart, Sachin Gupta, Scott Castiglia, Sean Campion, Shawn Krest, Thomas Lore, Tom Powers, Tyner Wilson, Zach Ellin

Thanks as always to the noble efforts of the charters -- they are the ones who will be pushing NBA statistical analysis to new heights!

Interested in doing some game charting? Send a message to: [email protected]

Rate this Feature
Poor   Fair   Good   Excellent

Enter your comments in the box

Email (optional)

We want your feedback! Tell us your thoughts

Copyright ?2006 by, All Rights Reserved