The extra possibility of making 3 straight free throws #2 seems to tilt the calculus to the 3-pt shot.
- UConn seems to get around 11 FTMs / game; even if 8 FTMs were due to 2-pt attempts, 8 vs 3 is not so different from 2 vs 1 (2FGA vs 3FGA in UConn’s schemes);
- UConn’s 2015-16 team shot 60% from 2 and 38% (not 40%) from 3, and that was optimal #2 (see below);
===
The binary decision to go for a two or for a three at a granular level is straightforward:
- Note that not all two’s are the same level of difficulty (e.g. an uncontested layup is generally easier than an uncontested jumper);
- also note that almost all three’s are the same level of difficulty, without regard to contestabilty;
- for the same player, a step-back three is better than the closest long two;
- KML would take an uncontested 3 vs driving to the basket;
- Azzi for 3 is probably better than Ayanna for two at the foul line but not necessarily better than Sarah near the basket (the Pomeroy feature you highlighted); note that paint opportunities are not always available;
- and so on.
With UConn’s versatile (all mostly 3-level) players in a read-and-react motion offense to find the best shot:
- practices will tell the coaches the relative team efficiency in two’s vs. threes;
- with such knowledge, coaches can set the percentages of threes vs twos, as a way to optimize the offense — generally 31-33% although it dipped (28%) in the 2009-10 team (Maya-Tina);
- if UConn, as a (homogeneous — sort of) team can make two’s at a 60% rate #1, make free throws at an 80% rate, gets fouled with FTs awarded (mostly in the act of shooting a two) 10% of the time and gets fouled in the act of shooting a three 10% of the time, the team must be making three’s close to 38.00% #2, for such fact pattern to be optimal;
- the fact-pattern above matches the 2015-16 team, with 64.6 FGA/game at 1.32 pts per FGA (modeled points of 85 pts / game vs 88 pts / game (actual));
- This episodic optimization will apply to this hypothetical situation and mostly not necessarily apply to others.
For teams with similarly well-defined two’s vs three’s profile, the binary decision team optimization could be well-defined:
- Teams with various credible 3-point threats but with a measly post offense will most often prefer the three to a two;
- And vice versa (e.g. pre-Tehina Pao Pao/Tessa Johnson South Carolina years).
Another thread also attempted to discuss this.
#1 At 50% 2-pt accuracy, 3-pt accuracy must be close to 31% to be optimal. Counter intuitively, 31% is lower than 33%.
The possibility of making 3 straight FT’s tilts the calculus towards the 3-point shot.
#2 Comparing Bayesian means of 2-pt attempts vs 3-pt attempts: made FGA with no foul, made FGA with foul, missed FGA with foul.