[CompSci87]

5.  Last point and a pet peeve of sorts.  The pull up 6-8 footer is your friend, especially when the defender gets to a spot and is obviously not in "contest shot" mode.

Amen! Seems to be largely a lost art across much of basketball - college and pro.

There is a reason for this. Basketball Analytics deprecates shots in the 3 to 10 foot range. They have the lowest expected value as a function of range from the basket.  See https://fansided.com/2015/08/17/nylon-ca...selection/
Granted, these numbers are for all types of shot, contested as well as uncontested. The raw expected value doesn't take into account angle to the backboard either. However, this low expected value in the general case has pushed these shots out of the game plan mainstream. If you don't practice taking that shot, you don't take it in a game, or if you do, you are more likely to miss. You may also get yelled at. The Analytics becomes a self-fulfilling prediction. Analytics says you either try to drive to the 0 to 3 ft area, or kick it out for a 3. That is the "new" basketball.

I'm glad you mentioned "contested as well as uncontested".  There a two things that bother me about the idea that analytics says you should only either go to the basket or shoot a 3.

* The pretty pictures that show the expected points from taking a shot from any spot on the floor are based on data that covers only *shots that were attempted*. That is, the expected points for shooting from spot X on the floor are really the average points gained by attempts that were actually made from that spot. This introduces a sort of selection bias. Basketball players are smart enough to attempt only shots that they think are a good idea to take, depending on how open the shot is, what the player's range is, and other factors such as time and score.

* In basic game theory, one often finds that the optimal strategy is to randomly mix multiple pure strategies, because if your opponent knows you always use the same strategy, your opponent can always use the strategy that is best against that one. Basketball is a lot more complex, but still, if your opponent knows you'll never do X (only Y or Z), they may be able to gain an advantage by setting up their defense to not bother guarding against X.  That may mean that if you did to X (at least occasionally), it would be much more successful, forcing your opponent to guard against X and weakening their defense against Y and Z.  With basketball and X as "midrange shot", this argument seems like it makes sense. (And in football, "make them guard the whole field", and "don't be too predictable" follow from this kind of argument.)

I suspect both of the above result in exaggerated conclusions about how everyone should learn to be a 3-point shooter and no one should ever shoot midrange.

[CompSci87]

5.  Last point and a pet peeve of sorts.  The pull up 6-8 footer is your friend, especially when the defender gets to a spot and is obviously not in "contest shot" mode.

Amen! Seems to be largely a lost art across much of basketball - college and pro.

There is a reason for this. Basketball Analytics deprecates shots in the 3 to 10 foot range. They have the lowest expected value as a function of range from the basket.  See https://fansided.com/2015/08/17/nylon-ca...selection/
Granted, these numbers are for all types of shot, contested as well as uncontested. The raw expected value doesn't take into account angle to the backboard either. However, this low expected value in the general case has pushed these shots out of the game plan mainstream. If you don't practice taking that shot, you don't take it in a game, or if you do, you are more likely to miss. You may also get yelled at. The Analytics becomes a self-fulfilling prediction. Analytics says you either try to drive to the 0 to 3 ft area, or kick it out for a 3. That is the "new" basketball.

I'm glad you mentioned "contested as well as uncontested".  There a two things that bother me about the idea that analytics says you should only either go to the basket or shoot a 3.

* The pretty pictures that show the expected points from taking a shot from any spot on the floor are based on data that covers only *shots that were attempted*. That is, the expected points for shooting from spot X on the floor are really the average points gained by attempts that were actually made from that spot. This introduces a sort of selection bias. Basketball players are smart enough to attempt only shots that they think are a good idea to take, depending on how open the shot is, what the player's range is, and other factors such as time and score.

* In basic game theory, one often finds that the optimal strategy is to randomly mix multiple pure strategies, because if your opponent knows you always use the same strategy, your opponent can always use the strategy that is best against that one. Basketball is a lot more complex, but still, if your opponent knows you'll never do X (only Y or Z), they may be able to gain an advantage by setting up their defense to not bother guarding against X.  That may mean that if you did to X (at least occasionally), it would be much more successful, forcing your opponent to guard against X and weakening their defense against Y and Z.  With basketball and X as "midrange shot", this argument seems like it makes sense. (And in football, "make them guard the whole field", and "don't be too predictable" follow from this kind of argument.)

I suspect both of the above result in exaggerated conclusions about how everyone should learn to be a 3-point shooter and no one should ever shoot midrange.