Tuesday, August 09, 2011

Shoot First, Ask Questions Later

Our analysis of run-first quarterbacks was a resounding success, drawing praise from notoriously sharp critics like Wheelhouse Jerry and ... well ... no one else, but praise from Jerry is a big deal. As promised, our follow-on Part II analyzes the relative merits of shoot-first point guards (SFPG).

After searching basketball-reference.com for guards with more than 3000 assists since 1990, Mark winnowed out players who should not be considered to be point guards. This gave us 41 players. As before, the peculiarities of the website's search engine only gave players who amassed more than 3000 assists since 1990, so guys like Magic Johnson and Isiah Thomas weren't included. I then back-filled the career stats for players who started before 1990.

The traditional primary role of a point guard is to distribute the ball, so SFPG will be point guards who shoot more and pass less than other point guards. Passing is reflected by the assist statistic, so SFPG should have a higher ratio of field goal attempts to assits than other point guards. I divided everyone's career FGA by their career AST and got this:

(Right click on any image in this post and open it in a new window for a larger version.)

Here's the histogram and bar chart for the hard-core numbers junkies:

As an aside, John Stockton is the only Hall of Famer in this analysis. I don't know if that's because point guards haven't been good over the last 20+ years, or if it's hard to get into the Hall as a point guard, but I was surprised. By contrast, there were 6 HOF'ers in the RFQB comparison.

For purposes of this analysis, I'll define SFPG as a point guard with a FGA:AST ratio of 2.0 or higher. This gives us Rafer Alston, Baron Davis, Chauncey Billups, Derek Harper, Stephon Marbury, Gary Payton, Sam Cassell, Mike Bibby, Tony Parker, Steve Francis, David Wesley, Derek Fisher, and Anfernee Hardaway. I'll also include Damon Stoudamire, Nick Van Exel, and Terrell Brandon because they're practically at 2.0. None of this matters though because as in the RFQB piece I won't make any efforts to highlight these players in the graphs.

I converted everyone's FGA, AST, and PTS to a per-12 minute basis. I know that the standard convention is per-36 minutes but I like 12 better because it's one quarter. For these purposes it doesn't matter anyway because all of these variables will just be plotted along an axis and multiplying them by any random number won't change the plot. I think.

I assumed that players who take more shots have a higher FG%, particularly because all the players in this analysis stuck around in the league long enough to amass 3000 assists. Surprisingly, shot rate has nothing to do with FG%:

This is where Mark's theory starts to gel. Baron Davis takes about 5 shots per quarter and makes just over 2 of them. Those 3 missed shots are 3 wasted opportunities for someone else to score. Eric Snow is a similarly shitty shooter, but he takes about half as many shots and thus has half as many wasted opportunities.

It seems odd that GM's would continue to employ players who shoot and miss a lot. I would think that if a PG takes a lot of shots, you wouldn't keep him on your team unless he makes a lot of those shots. But what do I know.

Derek Fisher is interesting. His FGA:AST ratio makes him a SFPG, but he only takes about 3.5 shots per quarter. His FG% is a paltry 40.1%, so he misses about 2.1 shots per quarter. As we'll see in a bit, he generates less than 1.5 AST per quarter. Really an odd player.

FG% and FGA/AST are unrelated:

As we saw earlier, a PG's ability to shoot is not related to his propensity to shoot, and I guess it makes sense that a PG's ability to shoot is not related to his passing skill. Hence this blob of a graph.

Unsurprisingly, the players with the lowest FGA:AST ratio have the most assists per 12 minutes:

Also unsurprising is the fact that the players with the highest FGA:AST have the most shot attempts per quarter:

Derek Fisher, however, is surprising. Despite being a SFPG he only takes 3.5 shots per quarter. So dude doesn't get many assists nor does he take many shots, and when he takes a shot he isn't likely to make it. I wish I had his job and his $57 million in career earnings. To be fair, he shoots the 3 pretty well:

In fact, all SFPG shoot the 3 pretty well compared to all the other PG in this graph. I suspect that this is why they are able to stick around long enough to amass 3000 assists. I also suspect that this is why they shoot so much -- they are in love with their own 3-point-shooting ability to the point that they fail to see (or even ignore) open teammates.

The real test of the value of a SFPG is whether you can win games with one. So I plotted FGA/AST vs. win shares per game:

Only three of the top 10 players with respect to WS/game are SFPG: Billups, Payton, and Parker. However, only four of the top ten players with respect to WS/game won championships: Billups (1), Payton (1), Parker (3), and Kidd (1). All but Kidd are SFPG.

Of the 16 SFPG identified in this analysis, five won titles: the aforementioned Billups (1), Payton (1), and Parker (3), and Derek Fisher (5) and Sam Cassell (3). Winning a title is a huge deal and I don't want to belittle Fisher's accomplishments, but playing with Kobe and Shaq/Gasol made it a lot easier. And as we already saw, Derek Fisher doesn't do anything, so his SFPG status is pretty questionable. Sure, he shoots more twice as often as he gets an assist, but both events are so rare that I'm not sure he matters in the grand scheme of things. Payton was a backup when he won his ring, and he went 0-3 from the field during the meager 32 minutes he played in the playoffs that year. He chipped in 3 assists in the playoffs for a tidy 1.0 FGA:AST ratio. Cassell was a backup on all three of his title teams but he played an important role, especially in Houston. Billups and Parker were major contributors to their championship teams.

With the exception of Billups and Cassell in 2008, all the SFPG who won a title won their rings as teammates with arguably the league's best player that year or that year's playoffs: Payton with Wade (Finals MVP), Parker with Duncan (league MVP, Finals MVP) Fisher with Kobe (Finals MVP) and sometimes Shaq (league MVP, Finals MVP), and Cassell with Olajuwon (league MVP, Finals MVP). Cassell played with Garnett, Allen, and Pierce (Finals MVP) in 2008; although none were as dominant as Shaq or Duncan in their primes, it can't hurt your title chances to play with three just-past-their-prime Hall of Famers.

So what does it all mean? At least 13 of the past 21 championship teams included a SFPG, but these particular SFPG either didn't do much, didn't do anything, or played second banana to the best player in the league or the playoffs. Only Chauncey Billups really stands out, at least to me.

If you ever get hired to build an NBA roster, I suggest that you avoid SFPG unless you have the best player in the league. You might also want to avoid the best player in the league right now but that's a different post. I say this depsite failing to find any worthwhile trends in the data, and based entirely on the previous conjecture. I eagerly await your contrary analysis in the comments.


Almighty Yojo said...

have to drive my kids to soccer camp, so i haven't finished yet, but it looks like another bang up job. my only complaint? no use of the term "chucker."

T.J. said...

I wonder what cocaine bear thinks of all this?

Jerry said...

Nice work. You're starting to arrive at some of the fundamental concepts of advanced basketball stats.

I suspect that the Derek Fisher issue could be partially cleaned up by incorporating free throw attempts into the shooting volume figures because he doesn't shoot many and some of the other guys on the list probably shoot a lot.

There's a stat called Usage Percentage that combines a players shots, free throws, and turnovers and basically divides that sum by the # of possessions that the player was on the floor for.

For perimeter players, Effective FG% can be fairly useful. It incorporates the fact that 3 pointers are worth 50% more than 2 pointers.

Mark said...

Effective FG% and usage rate are the two advanced stats that I like (and believe) most.

Dave said...

i must confess, i am in love with my own three point shot, to the point that i ignore open teammates.

i am also in love with doing "the spider." i wish you had a stat for this ability.

T.J. said...

Snoop and the Good Doc back with a brand-new sack
Shit's wrong, money gone, I'll blast
Out of town, out of bounds, no pass
Runnin' up, talkin' shit, get smashed
(Shoot first) Ask questions last

zman said...

Thanks guys. I suspected I wasn't using the right metrics. I should've run this by Mark first.

I completely ignored FT. You can get 1, 2, or 3 FTA depending on the situation. Each situation may be more or less desirable than another. Is there a metric that takes this into account? For example, making a shot and getting fouled seems more desirable than missing a shot and getting fouled, for obvious reasons. And is there a way to include this in the FGA stat? How can you tell how many FT correspond to missed shots?