I was playing around with some numbers yesterday and i started to come up with this theory:
turnovers can be avoided by making a player shoot a very large percentage of the time he's on the floor.
Now that i've given the conclusion, i'll pull a Tarantino and show evidence after the fact...
While I used a relatively small sample size (only one of my teams, which runs a motion offense), it seems that passer rating has pretty strong negative correlation with turnovers per minute, which makes sense and most of us would have assumed. However, passer rating has no correlation at all with turnovers per FGA. Which also makes sense.
Generally, turnovers per FGA increases as you go down the list of the high scorers on the team. At the same time, FGA/minute decreases, meaning FGA appears to have no effect on TO (as a side note, the poorer scorers that also shot 3's appeared to have another boost in turnovers per FGA, so much so that it could even be theorized that 2pt FGA rarely turn into turnovers compared to 3pt attempts, but i digress...)
It is assumed that that increasing a player's distribution, and thus increasing a player's FGA, does in fact decrease the player's FG%. As mentioned before, increasing a player's distribution appears to have no effect on turnovers per FGA (at least compared to the other factors i discussed above).
On each trip down the floor, an individual player has the following potential outcomes:
1. The player attempts a field goal, leading to either a FG, missed FG, or FTA's
2. The player gets an assist
3. The player turns the ball over
4. The player does nothing statworthy
I can think of two ways that the engine can run for each possession, each of which would lead to a different conclusion based on the above data:
1. invisible (meaning not showing up in the play-by-play) passes take place, each of which has a % change of a turnover. each pass takes up a variable amount of time, and each possession includes a random number of passes concluding with a FGA or TO.
2. all possible possession outcomes, including possession time, are mutually exclusive.
If #2 is true, which is the lazier way to set up the simulation and I would expect it of whatif (no offense intended to whatif):
By increasing the distribution of a player with a low passer rating, you decrease the chance of turnover on any given possession, and potentially offset any negative effect you may experience from said player's decrease in FG%.
This theory is supported by the fact that if you subtract minutes from a player's total playing time based on 2pt FGA (e.g. 20 seconds per 2pt attempt), the correlation between passing and TO/min increases. So when a player shoots a 2, the chance of turning the ball over is taken out of the equation.
If #1 is true, then the theory is false. This is supported by the fact that if you subtract minutes from a player's total playing time based on all FGA (20s per FGA), then the correlation between passing and TO/min does not increase. (it may also be possible that turnover potential is added for players looking for 3pt shots as a way to mitigate teams that shoot a lot of 3's. I have yet to investigate this, but if true it would point back to conclusion #1).
I've probably lost everyone by now, but anyway...thoughts?