APBRmetrics

so as I've been building out some stats for my personal use I came across Poss in bball ref ( http://www.basketball-reference.com/about/glossary.html ) , used in both Pace and DRTG. The formula is defined as below:

0.5 * ((Tm FGA + 0.4 * Tm FTA - 1.07 * (Tm ORB / (Tm ORB + Opp DRB)) * (Tm FGA - Tm FG) + Tm TOV) + (Opp FGA + 0.4 * Opp FTA - 1.07 * (Opp ORB / (Opp ORB + Tm DRB)) * (Opp FGA - Opp FG) + Opp TOV))

This leads to the same # of possessions for both teams each game, so obviously not very accurate. I see an article on calculating individual ORTG/DRTG which has individual posessions in there ( http://www.basketball-reference.com/about/ratings.html ) , but I don't see anywhere on doing it for teams. If anyone could point me to an article/formula for this it would be greatly appreciated. Thanks!

I assumed a lot of people here would know about this it seems like an extremely common aspect of advanced stats. I guess I will phrase the question another way, is it possible for a team to get 16 points on a single possession? i.e.

make a 3, get fouled, miss FT, ORB
make a 3, get fouled, miss FT, ORG,
...
5x total + make final FT

would this all be considered one possession for that team? if not, then why are the team possession/pace defined as if teams should be exactly the same (or even within 1-2) in the game? If it is that way ( as described above ), I just don't see how this is useful in determining team efficiency which I thought was the foundation of the whole (per posession) stats

I dunno if a oreb off of fta continues possession or not.

there is some stuff about minor possessions that is relevant, I think. possession / pace counts can vary from site to site and maybe team to player because of this and the .44 fta estimator.

on the topic
viewtopic.php?f=2&t=8511&p=18816&hilit= ... ons#p18816
viewtopic.php?f=2&t=8450&p=18270&hilit=pace+possessions
viewtopic.php?f=2&t=8181&p=15064&hilit= ... ons#p15064
viewtopic.php?f=2&t=8160&p=14954&hilit=pace+possessions

Folks like DSMok1, Mystic, K Pelton, JE, etc. could probably explain more if they see this or are asked directly, if they have time and are inclined to.

yea not much there unfortunately, seems something nobody has thought/written much about yet, )!@4ing hollinger :-p. Just seems so inaccurate to me that both teams will have IDENTICAL # of possessions in every game, and then when you look at how Pace is calculated:

48 * ((Tm Poss + Opp Poss) / (2 * (Tm MP / 5)))

it seems to obviously be expecting the two Tm Poss and Opp Poss to be different

Guess i'll create my own based on pbp to provide something actually relevant/useful. Thanks for the help Crow

Some site has actual possessions not estimates. Nylon Calculus?

http://nyloncalculus.com/stats/

Basketballgeek.com dealt with this too awhile ago.

Possessions are defined as equal for both teams. You can calculate it for each team similarly and get a different answer. So they ( who estimate possessions) average the 2 and assign the same number to both teams. This is supposed to improve the accuracy of the poss. count.

As far as I am aware, nobody bothers with fractional possessions: those which start with less than 0:24 in the quarter.

An OReb is rather like a 1st down in football -- it continues the possession.

A lot of the varying definitions for possessions and pace are due to working from boxscores. As you mentioned, if you have play by play you can just count them.

The offensive rebound issue is, to my mind, more a matter of taste. Some/most people like to define possessions by switches; if one team had the ball and now they don't, then a possession ended. If one team takes fifteen shots, missing fourteen of them but rebounding it every time, then there was no switch and thus it's the same possession. Good ORB teams can look slow if you count this way, even though they might get an average number of shots (which, in my mind, does not mean they're a slow team).

If the stat or analysis you have in mind would be better described by defining possessions a different way, then you should define it a different way.

I do possession counting for our data at bigleagueinsights.com. We define a possession as anytime your team has gained control of the ball where you did not have control of it before. To determine the number for each team I count through the play by play. It seems easy to say, but it's not that easy to do, given all the weird things that can happen in game. I should note that we also count possessions where nothing happens that will not be captured in a box score. This is basically possessions at the end of quarters that don't end up in a shot being taken.

An easier way to calculate the number of possessions for each team in a game that I came up with is to the following calculation:

Tposs = FGA - OREB + TO + 2FT + 3FT

The only thing tricky here is that you need to search the play by play for any instances of "Free Throw 2 of 2" and "Free Throw 3 of 3" to get the counts of the 2FT and 3FT respectively. '

If we look at the TOR/PHI game for yesterday:

http://bigleagueinsights.com/box-scores ... 1-23-21143

You can do the above calculation and get a team possession for each side of 92 possessions.

This is the total number of possessions each team has that end with a shot being taken or a turnover. Essentially it's the number of possessions where some sort of offense took place.

I do not agree that the number of possessions have to be equal for both sides. If we look at a shortened play by play let's consider this game:

Q1
TeamA wins tip off
TeamA scores 2
TeamB misses
TeamA rebounds
TeamA scores
End of Q1

Q2 (TeamB gets ball because they lost tip off)
TeamB scores
TeamA scores
End of Q2

Q3 (TeamB gets ball because they lost tip off)
TeamB scores
TeamA scores
End of Q3

Q4 (TeamA gets ball because they won tip off)
TeamA scores
TeamB misses
TeamA rebounds
TeamA scores
End of Q4

I think we all would agree that TeamA had 2 possessions in the first, 1 in the second, 1 in the third, and 2 in the forth for a total of 6. For TeamB I think we can agree they had 1, 1, 1, and 1 possession in each quarter for a total of 4. This is the simplest case, but in any game a team can have at most 2 more possessions that the opposing team.

If there is interest I can make something that provides the possessions for each game that will update after the games automatically. I already have them all going back to 2000.

thanks a ton BLI, will consider this approach tomorrow when I'm more awake