The otherday, Ethan Strauss (@SherwoodStrauss) of ESPN/BleacherReport/HoopSpeak et. al, tweeted something interesting that piqued my attention:
Individual basketball stats will never agree on 1 eternal question: What's the value of taking a shot?
Depends on the shot, obviously.
My initial instinct is always to be snarky, so my first thought is that we don't have to "agree" here at all; whether two people agree about something will not change its truth. Of course, the issue is a little more complicated, as it always is whenever something like this comes up on twitter; 140-character messages have a tendency to simplify the subject, with the implicit understanding that the reader knows the nuances of the issue (as an aside, this is why it always ticks me off when someone says "I never said that, stop making assumptions about what I meant!" on Twitter. If no one made assumptions about "what you mean" in your 140-character tweets, Twitter would be a useless medium).
Let's start with a simple assumption: shots that do not go in (and do not result in free throws) have negative value. Almost every advanced statistic acknowledges this fact. Shooting without scoring is bad. What most statistics disagree about, though, is exactly how bad. In NBA's Efficiency Rating formula, for example, an eFG% of 25% is all you need to "break-even" on shooting. With GameScore, it's worse, at 20.2%. Hollinger's PER has a break-even eFG% of somewhere around 33% (I believe). In the WinsProduced formula, the "break-even" true-shooting percentage is a little over 50%. But then, there is a position adjustment too; shooting 52% as a center won't be good enough to make you "average", all else being equal.
In other words, of all the advanced metrics, Wins Produced is the harshest on missed field goals.
And here I have to take a minute to address the myth of shot creation. Let's set the record straight on one thing: shot creation, in an absolute sense, simply does not exist. If you put a group of high schoolers on an NBA court, they may indeed have trouble ever getting a shot off vs. an NBA defense that was trying. But historically in the NBA, every team, no matter how bad, has been able to "create" plenty of field goal attempts. Ask the 1990-91 Denver Nuggets, who managed to lead the NBA in FGA and FGM (by a fairly wide margin) with one of its worst offenses and certainly the worst overall team. Consider also that every team, no matter how bad, always seems to have guys who score in double figures and get off lots of shots. Last year the Bobcats were arguably the worst team in NBA history, yet they had 5 guys who were above average in FGA/48 minutes). Somehow, NBA defenses were not shutting them down so much that they couldn't get shots off.
Therefore, arguments that revolve around statements like "Well, someone has to shoot" are generally not worthwhile in discussing "shot creation". Arguments that a player has value because he can "create his own shot" are not helpful, because almost any NBA wing player can do that. If this skill were rare, we would see big variablities in the FGA of teams from year to year. If this skill were rare, would the Bobcats have 5 guys that could do it? The 11/12 Bobcats? No, the relevant skill is not getting a shot off at all, it is creating good scroing opportunites. The difference between those two things is dramatic.
And this is, I suspect, what most people mean when they talk about shot creation (one hopes); not the ability to get a shot off at all, but the ability to "create" good shots, by beating the defender of the dribble to get layups (LeBron, Paul, Rose, etc), or using screens effectively to get open 3 pointers (Allen, Korver, etc), by being taller/lengthier than the opponent so it's impossible to block or hinder their shot (Dirk, Dwight, LeBron again), and myriad other ways. And here's where it gets interesting and here is where it is, in my opinion, almost impossible to evaluate a player's ability to do this by the naked eye:
sometimes, bad shots go in. Players get lucky on very difficult shots
sometimes, good shots miss. Players get unlucky (or blow it) on easy shots
some shots are easier for some players due to ability/talent; LeBron can drive to the hoop and make a lot of shots that, for example, JJ Barea could never make
some shots are harder for some players; Wade is not very good at three-pointers, so a wide-open three for him isn't as good a shot as an open 3 for Allen, Battier, or Jones.
shot selection is important. It doesn't matter if a player can be extremely efficient; it only matters if he actually is.
That last point always reminds me of one of my favorite sayings:
Stupidity and Evil are the same thing if you are only counting the results.
Which, applied to this situation, means that it doesn't matter if you can't make your shots or if you just choose to take bad ones, the result is exactly the same (and equally bad) for your team. The above factors mean that when evaluating a game with the naked eye, coaches, fans, pundits, scouts, etc, are all incredibly prone to lots of cognitive biases. They see Melo's brilliant post up moves and label him an efficient scorer, despite the fact that he takes (and misses) a lot of bad shots. They see Monta Ellis' lightning quickness and label him a great "shot creator", even though he has a career shooting efficiency that is almost exactly average.
So, then, what constitutes a good shot? I think this is straightforward. We know the value of a point (as it correlates to wins, .033). We know the value of a lost possession (as it correlates to wins, -.034). A shot that does not result in points equates to a lost possession. Therefore, any shot that does better than "break-even" on this scale is a "good" shot. We could do some complex math but for our purposes we can see that .033 and -.034 are very close in absolute valude, and just put the over-under on just a little bit better than a point per field goal attempt (exactly one point per FGA would return -.001 wins). And we see, also, that the results in the NBA bear this out: the average team in the NBA has a true shooting percentage of about 52.7, meaning that the average team scores about 1.05 points per field goal attempt.
Then, of course, we need big samples. As I said, bad shots go in sometimes, good shots sometimes don't. But over time, if a player is consistently scoring more points than FGAs, he is, for all intents and purposes, shooting good shots. Essentially, Wins Produced assumes that a player who shoots below this average is "not good" at shooting, and players that shoot above this average are "good" at shooting. Let me re-emphazise that the model does not explain (or care) why. A player can be an inefficient scorer because they take bad shots or because he can't hit the broad side of the barn even when they are open; the model will treat both players the same. It's up to coaches, scouts, and others to determine the reasons and make adjustments if they can. In other words, you do not get any bonus points for talent. Many a coach (and organization) has lost a lot of games by trying to get talent to meet expecatation (we're looking at you, Michael Beasley).
But one thing is clear, to me at least: just because a player has great talent and is clearly capable of creating easy scoring opportunities, this does not make their bad shots "valuable". The simple fact is, Carmelo Anthony would be a more productive player if he simply stopped taking shit shots; so would Russell Westbrook. The idea that the bad shots that these players take create value for their team has no basis in evidence at all (nor is there any evidence that these players are reluctant shooters who are shooting so much because "someone has to take the shots"). You can choose to disagree with me on that, but it's rather like disagreeing with me about evolution and creationism -- as far as I'm concerned, prove it or move it.