As basketball's advanced stats community gathers for the 10th annual MIT Sloan Sports Analytics Conference on Friday and Saturday in Boston, the battle over the value of statistical analysis to teams is long over.
Although not everyone is convinced, all 30 NBA teams have some kind of statistical presence in their management. When teams ranked near the bottom in ESPN's 2015 Great Analytics Rankings, they didn't point to it as a badge of honor, as they might have in another era.
Instead, they vociferously defended themselves against the perceived criticism.
The results of this shift, which has taken place almost entirely in the decade in which the Sloan Conference has existed, can be seen on the court. Led by Sloan Conference co-founder Daryl Morey's Houston Rockets, NBA teams are attempting a higher percentage of shots from beyond the arc than ever before, and ignoring lower-value 2-point shots away from the basket.
However, NBA offenses are not more efficient than ever. In fact, on a per possession basis, the league is scoring no better than it was when a handful of die-hards met for the first Sloan Conference in 2007, and it is scoring far less efficiently than it was at its peak this century in 2008-09.
Why hasn't statistical analysis been able to solve offense?
The median is not the message
The simplest explanation for why better shot selection hasn't revealed better results is that we typically define shot quality by average shooting percentage, and that's not the right measure through which to explore changes in shot selection.
When a team turns midrange shots into 3-pointers, they're getting not 3-point attempts of average quality, but instead, more difficult ones they had previously chosen not to shoot.
Economists would call these marginal attempts, and there's a reason the term marginal is synonymous with "less valuable."
In other words, the more 3s a team shoots, the lower percentage they'll shoot on them, all other things being equal. Morey's Rockets exemplify this trend. Over the past three seasons, they have attempted nearly 700 more 3-pointers than the next most prolific team, the Golden State Warriors. Yet despite its focus on outside shooting, Houston has shot slightly worse than league average beyond the arc over that span (35.2 percent).
As Seth Partnow recently explored for Vice Sports, the Rockets and Philadelphia 76ers -- run by former Houston assistant GM Sam Hinkie -- have had the most efficient shot profiles in the league the past three seasons, based on the average value of shot attempts. (Golden State is slightly ahead of Philadelphia this season, but not over the three-year span.)
At the same time, the Rockets and Sixers have tended to get below average value from their shots, precisely because they are taking more marginal attempts.
More 3s are still better for offense
The most extreme version of the marginal value theory holds that as teams attempt more 3s, defenses will adjust to the point that 2-point jumpers -- currently the least valuable shots in the game, on average -- become more valuable than 3s, which leads to a midrange renaissance.
The key assumption here is that defenses can control the outcome of 3-pointers as much as offenses can. The evidence suggests otherwise -- at least at contemporary levels of 3-point attempts.
The evidence for this? The spread of 3-point percentages is far wider on offense than on defense, and when top-five shooting teams face top-five 3-point defenses, they still shoot far better (36.6 percent) than league average (35.3 percent) in those matchups.
Stephen Curry's season might be the ultimate evidence that there is no defense for truly elite 3-point shooters. Despite increasing his 3-point attempts from 8.1 per game to 11.0, which means a lot more marginal shots, Curry is shooting a career-high 45.9 percent from beyond the arc. Because of his virtually unlimited range, defenses have found it almost impossible to keep Curry from getting off high-value 3s.
At the team level, we'll know defenses have caught up to offenses' ability to shoot 3s when more 3-point attempts are no longer associated with better offense. That's not the case now.
The top five teams in percentage of shots taken from beyond the arc average a 106.5 offensive rating, which is far better than the league average of 103.5.
Better shots offset the trend toward better defenses
I buried the lead a bit with the premise of this column. The fact is teams are shooting more efficiently than they did in the past. The league's effective field-goal percentage, which values 3-pointers as 1.5 field goals, reached a post-merger high point in 2013-14 (.501) and has matched that this season.
The other areas of offense explain why it is down on a per-possession basis. Most notably, teams aren't benefiting from as many second chances. This year, offenses are rebounding 23.7 percent of available misses, which would mark the third consecutive season that the NBA has had its lowest offensive rebound percentage ever and would be the culmination -- for now -- of a downward trend that dates to the late 1980s.
Naturally, players spending more time away from the basket to space the floor is a factor. In fact, there's a strong correlation between increasing 3-point attempts and decreasing offensive rebound percentages at the league level.
Before blaming the 3, however, note that the relationship is muddier on a team-by-team basis. The teams that take the most 3s don't necessarily grab fewer offensive rebounds. Houston, for example, is above average in offensive rebounding. Instead, the lowest teams tend to have coaches (such as Mike Budenholzer of the Hawks and Doc Rivers of the Clippers) who tell their players to get back on defense instead of chasing offensive boards.
In addition to offensive rebounding, free throw rates (as measured by attempts per field goal attempt) are also near historic lows. Since the BAA renamed itself the NBA in 1949-50, the bottom six seasons in free throw rate include each of the past four years, according to Basketball-Reference.com. (The other two were 1972-73 and 1973-74.)
Again, it's tempting to blame this on players shooting so many outside shots. Here's the funny thing, though: Teams aren't shooting from the perimeter more. In fact, the ratio of shots inside and outside 10 feet has been virtually flat since 2000-01, the first season for which such data is available.
Instead, as Partnow noted earlier this week, what has really changed is the distribution of jump shots, with 3-pointers replacing long 2-point attempts.
Fifteen years ago, 3-pointers accounted for about 30 percent of all shots taken from beyond 10 feet. Now, more than every other shot from this distance is a 3.
Unless defenders are more likely to foul players shooting midrange jumpers, the growth of 3s can't explain the decline in foul rate. Instead, it's more likely that defenses have seen the data and have put increased priority on keeping opponents off the free throw line.
What it all means
Before the rule reinterpretations to limit contact on the perimeter prior to the 2004-05 season, the NBA's league-wide offensive rating had declined by an average of 0.6 points per year since 1994-95 to a non-lockout low of an even 100 points per 100 possessions.
The NBA's average offensive rating dropped a similar amount from 2008-09 to 2012-13, the first season the Rockets began playing what the media termed "Moreyball." In other words, absent league intervention, points per possession have tended to drift downward in the NBA since the mid-1990s, presumably reflecting coaches' increasing willingness to sacrifice scoring for better defense.
Since the rise of the 3-pointer, however, there has been pleasant stasis. Offensive ratings have hovered around 103 points per 100 possessions the past four seasons, varying from 104.0 in 2013-14 to 103.0 in 2014-15. The league-wide average in 2015-16 is a 103.3 offensive rating.
As such, maybe the question we should be asking isn't why optimized offenses aren't better but how much worse offensive ratings might be if teams weren't taking higher-value shots.