Update 11-14-22

Welcome back tout le monde!

Last time I did one of these was before the 2022 NBA season, and in the time since then I have hand-tracked and compiled shooting data from sources all over the internet (mostly basketball-reference).

But more importantly, The Celtics went to the NBA finals.

I was working a lot during the series so I couldn't record every game, but in the game I did, the C's shot profiles did not suggest winning basketball. Meanwhile Stephen Curry shot the lights out from all over the court. In the playoffs, he compared favorably against all point guards when shooting from all distances further than 3 feet away; ridiculous. Curry's shooting profile looks like what Chris Paul has put up most of his career. Curry had a total OVS+ of 1.02, good for 4th among all point guards who played at least 132 minutes in the playoffs. And he was behind only Kyrie Irving (1.44), Monte Morris (1.9) of Denver, and the aforementioned Paul, who came out on top by a considerable margin with a OVS+ of 5.9. Of those four, Curry and Paul were the only ones to play past the 1st round of the playoffs

I have completed the data base of NBA shooting data from 1997-current (from basketball-reference.com), and will post the excel sheets with the OVS+ attached to each of them, with frequency+ and fg+ coming in the next few weeks. They're fun sheets to play around with. You can compare players from different teams and those who have played a minimum amount of minutes to get more insight from the base data, or you can just look at distance-by-distance breakdowns. The world (numbers) are your oyster.

One thing that still happens is that players who have played a low amount of minutes, players who come in take one shot and check out again, have a higher degree of deviation in their scores. It does reflect the nature of the stat; in that they have demonstrated a threat from at least some range, but we are left without an idea of how they shoot from all other ranges.

For example, also in the 2022 playoffs, Nickeil Alexander-Walker had a OVS+ of almost 7, (a very good grade), but this came from a sample size of only two shots, both from the 3-10 foot range, and making them both. Which, based on the way OVS+ is calculated gave him a score of 12 from that distance, and even paired with zeros from all other ranges he still logged a high OVS+.

That more or less concludes the recap of what I've done, so let's look towards the future. In the past I have used excel macro commands to semi-automate the OVS+ process, but I still have to manually enter the formula. My fix for this is to code the calculation into R, so that the process will be fully automated. At the moment my lovely, old computer is not able to install R Studio, so I am working out of the console which just makes it a little tedious.

The next step is to start doing some modeling with the data. As I do not have anyone looking over my shoulder telling me what's right and wrong, I will just be trying things out. The frequency% and fg% portions resemble a multinomial distribution and a binomial distribution respectively, so I will start there and try to find a way to combine them. My goal is to be able to extrapolate past data to present day.

-Nick