The 2 point jumper is dead.
This must be distinctly understood, or nothing wonderful can come of this FanPost.
The beginning of the end was the mandated introduction of the 3-point line, adopted by the NBA in 1979, and the NCAA in 1986. Despite initial (and, honestly, ongoing) debate over whether the 3 pointer is little more than a gimmick, it raised scoring, excitement, and variability, sticking in both leagues and eventually being introduced at the lower levels of basketball, including high school, middle school, and AAU.
It started slowly, with the NBA averaging little more than three 3 point attempts per game for the first 5 or 6 years following its adoption (note – NBA references will be more prevalent for a little while. They’re much better with historical statistics, as easily accessible NCAAB records really only start in 2000, and, as a whole, are farther along in embracing the field of advanced statistics). Since then, the number of three pointers attempted has steadily risen – other than a lull/drop in 1997 – each year, with current season averages for 2013-2014 sitting at over 20 three point attempts per game.
The increase, especially in the NBA, has been driven by the proliferation (and subsequent embrace) of advanced statistics able to, among many other things, parse the shots on the floor that provide the most value, or expected points per possession, to a team’s offense. This movement has especially picked up in the last 10 years, beginning most notably with the hiring of Daryl Morey, an advanced-analyticals, MIT-educated pioneer in the NBA as the general manager of the Houston Rockets in 2007. Since his hiring, the trend towards analytically-bent front offices has increasingly accelerated - about 25 of the league’s 30 teams staff at least one full-time advanced stats analyst. 14 of the past 15 GM hires in the NBA have been non-players, with the hires either stemming from the Rockets' or Spurs' management tree, or being known for their work in the field of analytics.
Advanced statistics have been around for a long time in NCAA basketball – Dean Smith was using tempo-adjusted stats back in the 1970s – but, in general, have not evolved as a whole alongside the arc of the NBA revolution. A number of reasons, ranging from a smaller sample size in a season (30-35 games vs 82), much smaller budgets (the NBA makes $4 billion a year), and an enormously larger number of teams (350ish vs 30), with the corresponding extreme disparities in talent level from game to game, contribute to this difference. This is not to suggest it’s absent – Brad Stevens made heavy use of advanced statistics during his time at Butler, and I’ve seen mentions of Buzz Williams (Marquette) and Rick Pitino (Louisville) tracking statistics well beyond the box score. They’re certainly far from the only ones, and it’s likely almost every competitive team is tracking certain statistics we don’t have access to as the public. But, the culture is far less advanced (and prevalent) than that of the NBA.
One of the most tangible things (at least to the layman’s eye) advanced analytics has changed about the NBA is the shot selection profile of the league as a whole, and especially teams employing advanced statistical analysis. Analysis of shooting percentages from different areas of the floor has highlighted 2 types of shots – attempts from the restricted area (extending from the basket to 4 feet away), and 3 point shots (especially corner 3s – in the NBA, the 3 point line is 22 feet from the basket at the corners, and 23.75 feet from the basket at the top of the key. This difference, while it doesn’t exist in college, makes a huge difference in the value of corner 3s in the NBA), as those that have by far the highest effective field goal percentage, and thus the highest expected point generation per possession. This article by the excellent Tom Ziller of SB Nation highlights the ‘Green Triangle,’ composed of shots in the restricted area, corner 3s, and free throws, as the most effective spots to shoot from on the floor. This graphic, by the master Kirk Goldsberry (seriously, check out his stuff at his site, and Grantland, for some amazing visual representations of shooters, rebounding, and other basketball info), shows field goal percentage from all spots on the NBA floor in 2012-2013.
Also, on a slightly unrelated note, if you’re at all interested in next-level stats, please check out these articles from another of Grantland’s contributors – Zach Lowe – on the newly-installed-around-the-league SportsVU camera tracking system. The ghost dots–where statistics say players should be while playing defense – vs. actual player position are fascinating.
Basically, these shot selection profiles value attempts in the paint, and all attempts from beyond the arc, above other shots - other than free throws, of course. By taking a look at the graphics provided in the linked articles, you can see that all 2 point attempts that qualify as jumpers (non-restricted paint, mid-range) generally sit at below 40% conversion rates. While there are exceptions to this rule – Dirk Nowitzki is an excellent mid-range jump shooter, and Chris Bosh is another example that comes to mind – in general, the conversion rate for 2 point jumpers sits maybe 5% points above 3-point FG% as an aggregate – and that’s generous. The added value of hitting a 3 pointer means that it is a more efficient shot than any 2 point shot being converted at a rate less than 1.5 times the 3 point percentage – if a team is shooting 40% from inside the arc, and 30% from 3 – the effective field goal percentage (eFG%) shows us that shooting 30% from 3 is the same as 45% (30% x 1.5) from 2 – and that the 3 pointer is the more valuable shot in this case.
The Four Factors, as explained in my last FanPost, are the advanced statistics most closely correlated with winning. The factor with the highest weight is eFG% - how well your team performs in this metric (basically, a measure of shooting from the field adjusted for the extra point value of a 3-pointer) contributes the most towards deciding the outcome of the game, relative to the other team. But, without getting too deeply into advanced stats, the takeaway is this – 3 pointers and shots within the restricted area are the best shots for a team to take. They’ll help overall shooting, and expected points per possession, more than any other.
For example – Daryl Morey’s team (as mentioned above), the Houston Rockets, is among the top 3 in both attempts at the rim (3rd, with 30.2 per game) and 3 point shots (1st, with 28.9 per game) in the NBA, per www.hoopdata.com – a trend from year to year. The organization, from management to coaches, as well as players, emphasizes taking as many efficient shots as possible – including getting to the free throw line.
It’s with this backdrop that we’ll dive into changes Villanova has made in its own shot selection profile this year, why this is a good thing, and how 3 point and paint attempts contribute to several measures of overall offensive success.
Changing the Game
Early this season, rather facetiously, I made several references to JDubs leading the advanced stats revolution in college, noting the preponderance of 3 point attempts and layups that made up Villanova’s shot selection. After digging a little deeper into the numbers, I’m removing the irony from that statement - it’s true (maybe not the leading part, but you'll get the idea).
On the previously-referenced Hoopdata.com, there’s a stat called expected effective field goal percentage – a stat that takes the shot selection profile of a team and weighs it against league averages from all those points – and spits out an ‘expected’ eFG%, or what a team should be shooting overall based on league-wide average from those points. This section owes a heavy debt to another article by Tom Ziller, exploring this concept in the NBA.
I took the aggregate NCAA team statistics from the past season – 2012-2013 – to generate this statistic (expected eFG%) for each individual team – though the only one we’ll really be concerned with is Villanova. All stats are based on data freely available from www.statsheet.com and www.hoop-math.com – both excellent sites for compiled statistics, if you don’t want to do all the leg work.
Oh, and one thing this reminds me of – I made reference to parsing the statistics in the last FanPost from box score data. This was inaccurate – I used play-by-play data, not box scores, to generate those statistics. Just a small note.
Here’s a graph of Expected eFG% (y axis) vs actual eFG%, in 2012-2013, for all NCAA teams. It’s split up into quadrants, showing good shooting vs bad shooting, and beneficial shot selection profiles (xeFG%) vs those less useful. The axes are drawn at the median value of both xeFG% and actual eFG%.
As you can see, Villanova landed on the bad side of shooting (the team’s eFG% last year came in at 47.44%, good for 240th or so in the country), and the mildly good side of shot selection (the team’s expected eFG% of 49.62%, based on the FG averages compiled from the data - 59.97% at the rim, 35.30% on 2 point jumpers, 34.05% on all 3 pointers – and Villanova’s shot selection profile, was good for 100th overall). Basically, the team didn’t select its shots all that well, and shot those it chose poorly – a lower eFG% than what was expected means the team’s percentages from at least one of those areas (ok, it was probably 3 point range) fell below nation-wide averages from the same spot. And this was fine – the team relied a lot on Mouph and his jump shooting abilities from the inefficient areas to generate offense on a team that was generally filled with immature, unready, or redshirting options on that end. But it still needed to change.
And it has. Take a look at this graph, using the same data from the 2013-2014 season. It should be noted this data may be missing the few latest games played, but can generally be considered accurate.
Villanova has increased its shot selection by a huge factor – their expected eFG% is 7th best in the country at 50.67%! And, encouragingly (for the season so far – certainly a regression candidate), Villanova has out-shot their expected eFG%, clocking in at an actual eFG% of 51.70%, which is 112th in the country at this point. They’re picking their shots extremely well, and making them at a clip higher than the country’s average – great signs for the young, promising season.
This increase in shooting, both from an expected and actually-hitting-the shots standpoint, has been driven by several factors. One not perfectly captured by the statistics is an overall improvement in offensive ability on the team. There’s no doubt this team has more capable shooters, both from outside and in terms of getting to the rim, than the roster we were looking at in 2012-2013. This obviously contributes to higher field goal percentages, offensive efficiency, and most typical stats.
But there are obvious changes in Villanova’s shot selection profile that have contributed to this jump that go beyond pure roster changes and point to either a deeper understanding of the most efficient spots on the floor, or just a ton of dumb luck (and chucking).
The most demonstrable change is the enormous drop in 2 point jumpers taken by Villanova in 2013-2014 vs 2012-2013. Stacking up these two Villanova teams against each other…
|% of Shots
|% of Shots 2-Pt Jumpers||3PM||3PA||% 3P's||eFG%||O Rtg||Expected eFG%|
Where have all those non-money makers gone? Converted into the most efficient shots in basketball – shots from the paint, or 3 pointers.So far this year, Villanova is only taking 14.50% of its shots from 2-point jumper range, which is basically anything that doesn’t qualify as a layup, tip-in, or dunk. This is the 5th lowest rate in the country – and, when compared to the 29.30% rate they clocked in at last year, it's even more impressive. Doing away with the 2 point jump shot is an excellent thing, for many reasons. As we touched on in the development of shot selection profiles, two-pointers away from the rim are literally the most inefficient shots in basketball, college or pro. There’s barely an appreciable difference in field goal percentage between jump shots inside the arc, and those outside – which means it’s MUCH more efficient to shoot a 3 pointer over a 2 point jumper nearly every time. For all those fawning over Arch’s shot fake from 3, then step inside for a very long 2-pointer: I would much rather see Arch continue to bomb away from 3 than start turning those 3s into long range J’s from inside the arc – unfortunately, we all know he’s not getting to the rim.
Last year, Villanova took 35.50% of its shots at the rim – that’s shot up to exactly 40% so far this year. The rest of the shots have been distributed outside the arc – Villanova is currently taking over 45% of its attempts from beyond the arc, an increase of over 10% from the previous year. This shot selection profile is comparable to the NBA’s most advanced analytical teams, and is a great indicator for this year’s team sustaining its improved offense so far. The team is emphasizing scoring from the most efficient areas of the field, and minimizing its attempts from the worst spots on the floor.
Why This is a Good Thing
Disclaimer – skip this section if you don’t like graphs, or math, even though none of this math is complicated, and you can do everything I did with Excel in about 5 minutes once you have the data
It should be relatively obvious why attempting more shots from the most efficient spots on the floor is a good thing – but let’s dig a little deeper.
In the following series of charts, I plotted various aspects of the shot selection profile against offensive indicators like eFG% and O Rtg, based on data from the 2012-2013 season. A few disclaimers – this isn’t an enormous sample size, as only one season of NCAA basketball is taken into account. Drawing incredibly concrete conclusions on one season of data isn’t the optimal situation, but the sample is also far from meaningless. Also – correlation, which will be discussed quite a bit as part of this section of the analysis, does not necessarily imply causation. The findings presented below will be presented as evidence to support the ideas that taking FG attempts at the rim and 3 pointers are generally beneficial for offense – not that only shooting 3s will lead to the world’s most efficient offense. Positive indicators, rather than across-the-board conclusions, are what I’d like you to take away.
First, let’s look at 3 graphs summarizing aspects of shooting in the 2012-2013 college basketball season, and the conclusions we can draw from this data.
In order, these graphs will look at % of shots taken at the rim vs eFG%, % of shots that are 2 point jumpers vs eFG%, and 3 point attempt % (of shots) vs eFG%.
The percentage of shots taken from the 3 areas will always be presented on the x-axis.
Fitting a linear trendline to each of these graphs gives us an interesting slice of analysis. The first is that shots at the rim have a low correlation with effective field goal percentage – the r value (Pearson correlation coefficient, a measure of the linear dependence between two variables) is 0.1473.
In simplistic terms, the r value is a measurement of how linear change in one variable affects change in another, either in a positive or negative direction. An r-value between 0 and 0.3 (positive or negative) is a weak linear relationship; between 0.3 and 0.7 indicates a moderate relationship, and anything over 0.7 (extending to 1, a perfect linear relationship) indicates a strong correlation. Squaring the r value gives the approximate percent of variation in one variable explained by the other; if two variables are found to have an r value of 0.40, approximately 16% of the change in one of the variables is caused by movement of the other.
The r value of 0.1473 (getting back to the analysis) between % of shots at the rim and eFG% indicates there’s a weak connection between the two – taking more (or less) of your shots at the rim won’t have an enormous effect on your eFG% - again, an overall measure of effectiveness shooting from the floor. The relationship is positive, indicating that the more shots you take at the rim, the higher the eFG% will probably go. Again, though, the relationship is shaky.
There’s a more obvious association between 2 point jumpers and eFG% - namely, the more 2 point jumpers a team takes, the lower its eFG% will go. The r value of this relationship is -0.3844 – a moderately strong correlation in the negative direction. As mentioned, these shots are simply inefficient, and there are pretty strong indicators that taking a lot of them will hurt you in the most important of the 4 factors.
Finally, the graph (plus associated trendline and r value) of % of shots as 3 pointers vs eFG% shows us that there is a positive correlation between the two – not quite as strong as the negative correlation between 2 point jumpers and eFG%, but substantial in its own right. The r value of this correlation is 0.341 – again, (mildly) statistically significant. 3 pointers as a higher share of your overall shots has the strongest POSITIVE correlation of any of the three shot types (as %’s of your overall shots) with eFG%.
Finally, a similar exercise was conducted with short two point ATTEMPTS, long two point jumpers, and 3 point shots versus offensive rating, the adjusted per-possession offensive efficiency of a team. The graph of long two point attempts vs offensive rating was omitted, as it had a very small negative correlation that’s just not significant. We already know they aren’t great shots – not much more analysis to provide.
The r value of the best-fit linear trendline of 2 pointers at the rim attempted versus O Rtg is 0.323, an indicator that shows the value and efficiency of attempts near the rim more effectively than the earlier graph comparing eFG% to % of shots at the rim. More at-the-rim attempts generally generate more foul throws, in addition to the shot’s already high effective field goal percentage, which helps push its correlation with offensive rating higher (eFG% does not account for free throws).
But, a stronger correlation can be found between 3 point attempts and offensive rating. An r-value of 0.424 – indicating that 18% of the positive change in O Rtg as 3 point attempts increase is directly related to the increase in 3PA – shows a strong correlation between the two. Generally, an offense needs to take more threes to become more efficient offensively – and Jay’s boys are doing an excellent job so far.
An added value of replacing long two-point jumpers with more 3s and rim attempts is an associated increase in offensive rebounds. This excellent graphic by Kirk Goldsberry (Jordan Sperber, who has written for VUHoops, did something similar for college rebounds, but Goldsberry’s graphic is interactive, easy to understand, and generally correlates with what Jordan found for NCAA rebounds, per the article’s text) charts the offensive rebounding percentage for shots from each section of the floor for all the shots in the 2011-2012 NBA season. Generally, the offensive rebounding percentage is slightly lower for midrange shots than for 3s (from similar areas of the floor). Offensive rebounds are also plucked most often from shots near the rim (35.4%). Extrapolating the increase in Villanova’s offensive rebounding numbers directly relatable to their change in shot selection (per these percentages – this is a quick and dirty calculation), based on field goal attempts so far this season, is a simple exercise.
In 10 games, Villanova has attempted 588 shots this season. Extrapolating this to a full season, Villanova will take approximately 1,750 shots, and miss 981 of them. We’ll start with the assumption Villanova rebounds 25% of its long two attempts (a made up number). Assuming a conservative 0.5% increase in offensive rebounding percentage for shots from 3 versus 2 point jumpers, and a 7.5% increase in offensive rebounding percentage on shots at the rim (increases are based on Goldsberry’s graphic), Villanova would have rebounded 4 more of its misses, simply based on its shot selection profile, over the course of a season. Statistically insignificant, sure. But another thing to remember when memorializing the 2 point jumper.
The Variability Argument
A common narrative, both in the wider basketball world and our own little corner of discussion, is that teams who take a ton of 3 pointers ‘live by the three, and die by the three.’
Of course, for all the positive indicators that taking more 3s increases offensive efficiency, it does introduce an element of variability not present with less of the long attempts. The variability introduced, however, is overblown, and doesn’t hold up incredibly well to a detailed look.
In the article linked here, which attempts to help Bracketologists pick upsets by identifying factors that predict variability in a team’s performance, the author finds a present, but overall weak, relationship between variance and 3 pointers attempted. Data from the 2011-2012 NCAAB season is used to determine the relationships. The standard deviation of a team’s point totals increases with the number of 3 pointers attempted – but this increase is between 1 and 2 overall points of standard deviation when you move from 20% of your shot attempts as 3 pointers to 40%. An approximate increase in standard deviation from 11.5 or so to 13, or 13.5, isn’t really all that significant, and the correlation between the two was already weak (r value of 0.22 – less than 0.30).
Furthermore, in this article exploring the same narrative in the NBA, and its actual applicability to what’s happening, Kevin Pelton – currently writing for ESPN – finds that the Orlando Magic (in 2011-2012, one of the most 3 happy teams in the League), are almost exactly average in terms of their standard deviation of scoring margin.
This is a narrative more supported by confirmation bias and old-school thinking than actual data. And, it ignores the fact that attempting more 3 points should move the needle in the positive direction (per the graphs above) for the team’s offensive rating. While the standard deviation may increase, the average points scored is also moving higher up the scale – meaning even the worst case scenarios for a 3-bombing team may almost exactly equal those for a team more oriented to 2 pointers, if they started from the same baseline (scoring average).