Why TS rather than PPP?

[barborous]

I understand that the point of TS is to get a better "true" sense of efficiency by possession, where FG and eFG just measure it against attempted field goals, weighted and unweighted. So why not just use PPP (or PPP/2 for an analogous measure)? As far as I understand, the 0.44*FT in the denominator of the TS formula is intended to on average approximate how many possessions were taken to get those free throw attempts, so why not just count the possessions directly? Is it just a convenience thing so the number can be gleaned solely from commonly published box score statistics?

[Doctor MJ]

"True Shooting Percentage" is the name given for the approximation of points per actual shot because "Points Per Shot" as used by TV crews fails to factor in the free throw situation.

Why use an approximation? Because trackers historically pretend that shots missed due to a foul never happened, and hence there is no way to go back and figure out exactly how many shots were taken.

As for why we don't wise up and at least get things right going forward, why there isn't more of a push for this, a few reasons:

1) TS% is really, really close to the exact right answer in practice.

2) The advantage of having it "perfect" in the future is negated partially by the fact that historical comparisons are such a big deal and a change of system would make that ugly.

3) The decision makers here are not the statisticians.

[EvanZ]

So just to update what I went and did...My site (see my sig) now gives "true" TS%. It does away with the 0.44 multiplier on fta, and actually calculates points coming off shooting fouls, not including technicals and and1's.

To give an example, as of the time of this post, Stephen Curry's TS% drops from 56.7% to 55.6% (due to his making 20 out of 21 technical free throw attempts).

[barborous]

So just to update what I went and did...My site (see my sig) now gives "true" TS%. It does away with the 0.44 multiplier on fta, and actually calculates points coming off shooting fouls, not including technicals and and1's.

To give an example, as of the time of this post, Stephen Curry's TS% drops from 56.7% to 55.6% (due to his making 20 out of 21 technical free throw attempts).

Wow, that's actually a pretty significant drop. I'd imagine the 0.44 estimate is a lot better for players who aren't the best free throw shooters on their teams (and thus don't take technicals) though.

[EvanZ]

Ha, since I posted that, I made another change. The new formula is:

0.5*(2*2ptm + 3*3ptm + and1m + fta2m + fta3m)/(fga + (1/2)*fta2 + (1/3)*fta3)

With that fix, Curry's TS% is 56.3%, so only a drop of 0.4 %-points. Not such a big deal afterall. But still. Good to have more exact figures at least to remove doubt.