APBRmetrics

I'm a little bewildered by the somewhat arbitrary use of 0.44 in the TSA formula:

(From basketball-reference.com)
True Shooting Attempts; the formula is FGA + 0.44 * FTA.
True Shooting Percentage; the formula is PTS / (2 * TSA).

But assuming that we have play-by-play data (which we do), why are we using this arbitrary 0.44, instead of doing it properly?

Example: If all a player did for the game was to attempt a 3FG, get fouled on the attempt (without making the basket), go to the FT line to shoot 3 FT's, we should just say that he made one (1) TSA. And not 0.44 * 3 = 1.32 TSA.

The only difficulty I see with my "proper approach" is how to account for Technical FT's. But this difficulty is not confronted, but merely evaded/hidden in the arbitrary use of 0.44.

I can see why 0.44 was used for data from the old days, where we do not have play-by-play data. But it is strange to see it still being used for games played yesterday.

Many of the "advanced" box score stats we use can be improved or at least made more precise using play by play data. The. 44 coefficient used to estimate possessions is just one example. usage and ast% assume that there isn't any change in pace when the player is on/off the court, not necessary if you calculate it with PbP. Id be interested to see if any kind of player is getting under or over valued because we're estimating and assuming and using c oefficient derived numbers most of the time.

Its something that used to come up in the comments of Neil paine's blog, I wonder what he'd say about it now.

Obviously when it comes to ts%, players who get an above average rate of and-ones and three-ft-trips are more efficient than box score TS gives them credit for. Players who don't generate bonus fta and shoot a poor ft % appear more efficient than they are as well.

There might be certain players for whom TS% isn't altogether accurate, but over the course of even 20 or so TSAs, it would be pretty tough to say that their percentage is skewed, considering that the R^2 value of

"Possessions~FGA-OR+0.44*FTA+TOV"

is pretty high (Around .97 IIRC).



Either way it's a much more important number than FG%, FT%, and even eFG%, and is close to ORTG.
Which is even pretty worthless without a usage stat next to it, however.

Hm I don't think it's just "one example" of an advanced stat that could be improved. Here's why.

1) TS% is a concept/idea that is sufficiently simple that even the casual fan can quickly grasp AND agree is a very sensible definition of the term "true shooting %".

2) And we have a very clear sense of the proper way to calculate it. (0.44 is not the proper way.)

3) Moreover, we have all the data to do calculate it properly.

The other advanced stats do not have the above properties, or at best have them to a lesser extent, especially (1).

I imagine we're stuck using the .44 for any pre-pbp seasons though, yes?

In the old days, there were 3-to-make-2 FT situations and 1-FT fouls. Either would require an adjustment to the .44 FTA multiplier.

I've decided to use a .50 multiplier on FTA and call it "effective shooting %". It's worse to miss 2 FTA than to miss a FGA, because not many missed FT are rebounded offensively.

Mike G wrote:In the old days, there were 3-to-make-2 FT situations and 1-FT fouls. Either would require an adjustment to the .44 FTA multiplier.

I've decided to use a .50 multiplier on FTA and call it "effective shooting %". It's worse to miss 2 FTA than to miss a FGA, because not many missed FT are rebounded offensively.

I've always done the same.