With the series against the Brooklyn Nets over, capped off by a 22-point waxing, it’s time to look back on the series through a non-traditional numerical lens. If you’re looking for X’s and O’s thoughts, check out these previous posts by our other writers.

Shot Attempt Density and Field Goal Percentage

No free throw data included

Our first item on the list examines where the shots came from and if they went in or not compared to the regular season. Note: a five-game sample size can produce some extreme values in the FG% chart, so keep that in mind near both the physical edges of the surface, as well as any tiny spots that are extremely red or extremely blue. For the FG% charts, blue corresponds to better in the playoffs, and red corresponds to worse in the playoffs.

What we can see upon a visual inspection is that the Sixers were more tilted towards 3s than usual in terms of their shot distribution (not that necessarily took more in absolute numbers). The Nets appeared relatively similar, except the right corner 3. As always, these are smoothed estimations of point processes. Don’t get super wrapped up in minute fluctuations and, instead, focus on the trends.

Points per Shot

Team Surfaces

No free throw data included

Points per shot surfaces are something I am still experimenting with in order to nail down the exact best method. Therefore, consider these the future first-round pick of graphics — full of limitless potential, but might be a bust in a year or two. Keep in mind that “blue is better” for the following plot in relation to playoff vs. regular season performance.

Generally speaking, the Sixers did a great job defensively, outperforming their regular season in most locations on the floor except the long-two range, which is pretty much ideal. It was interesting to me that just based entirely on the eye test, the Sixers seemed to be better in all phases (except dribbling ability of guards), and it wasn’t as if they needed to shoot 50 percent from 3 for the series to get the 4-1 series victory.

Expected vs. Actual

No free throw data included

This here is another slightly experimental section based on a comment section discussion I had with user bubqr in my recent piece on “momentum shots” that you can see by scrolling all the way to the bottom. If we can create points per shot surfaces for a player over the course of the season, we can then estimate points expected based on locations of shots from a given game. Do that for each game in the series and you have an expected number of points scored that can be compared with an actual number of points scored. One absolutely gigantic caveat is that this does not include any sort of context regarding who is covering the shooter, level of contest, or other pertinent information that NBA teams actually use to do vastly more complex versions of this analysis.

I only used information on the top-seven players for each team in minutes played (prior to Game 5), as they’re likely to have the most shot attempts to ensure that this actually does something.

Of note, I weighted the average of the expected points surface so that all prior Nets games were 25 percent of the average (instead of ~5 percent) and the remaining games were the other 75 percent. This was entirely arbitrary, but I wanted to include at least some component of matchup-related play. Unsurprisingly, this paints Joe Harris and D’Angelo Russell as negatives and Caris LeVert as a major plus. In fact, via Sixers PR, Brooklyn head coach Kenny Atkinson stated that “...their defense on Joe Harris. That really slowed us down. They [were] top-locking him and he’s kind of our engine”.

For the Sixers, and I hate to say this, but Mike Scott had a rough offensive series. The last column was put in there to prevent the Mike Scott Hive from ruining my life. Also, Tobias Harris is a high-end professional scorer and should be paid as such in the near future. Figure 3 shows the same information in bar chart form, if you prefer.

It’s interesting, that according to Sebastian Pycior at Nylon Calculus, JJ Redick had one of the worst series of all playoff performers, while Joel Embiid had one of the best. His methodology is based off of a tweak on RAPM, and is certainly more all-encompassing than what I did here, although mine is explicitly field goal production only.

Similarity Scores

One of the most popular activities across sports is to compare players. It’s great because it allows for endless (possibly stupid) debates. I also like comparing similarity, but this is going to be done slightly differently than usual. What I’m going to do is compare a player’s metric (will get to how I do that in a second) in the playoffs to all players in the regular season. If a player’s closest comparison is themselves, then they probably played the same both in play style and quality. If their closest comparison is Kevin Durant, they probably played better.

I chose to measure similarity by using (surprise) spatial data. Two players can have identical true shooting and attempt numbers, but be very different players. Here are two explanations of the method:

TL;DR Explanation

Each player has a very long (~180k) list of numbers that encapsulates their efficiency, usage, and assisted rate, all by every location on the floor. The sets of numbers are compared, and if Player A compared with Player B has a smaller score than Player A with Player C, Player A is more similar to Player B than Player C.

Vague Technical Explanation

Each player has a points per shot surface multiplied by their relative FGA density surface and the percent of total shots they took in their games. Another surface is created in the same manner, but with percent of shots assisted instead of points per shot. The two adjusted surfaces (PPS and %Ast) are averaged at each of 180k-plus locations on the court. Each player’s pair of surfaces at each of the locations is compared to every other player’s surfaces. The results are transformed, weighted 75 percent/25 percent towards the points per shot surface, combined, and then scaled from zero to one.

1. Playoff Ben Simmons has played offensively like regular season Montrezl Harrell (which is good)
2. Playoff Caris LeVert is, in a technical phrasing, hella nasty
3. Playoff JJ playing like two other old guys is a hilarious accident
4. Playoff Tobias doing that scoring work
5. Playoff Joe Harris as Larry Nance Jr. and Mario Hezonja is an extremely bad look

Wrap-Up

Well, that was a truly shocking amount of work to get those similarity scores. Not sure if the methodology actually improves on anything, but it was a fun challenge. Onwards and upwards now for the Sixers. May your weekend have Jimmy Butler in bunny ears vibes.