Airing of Grievances: Sixers Turnovers

I am generally pretty calm watching the Sixers play, with two major exceptions. First, when they play the Celtics, and second, after a stupid turnover (I did not handle that JJ bounce pass to no one with maturity). Following a pass thrown into the 10th row or a dribble off a foot, I turn into some extra from Hoosiers yelling about fundamentals. I’m pretty sure part of my hatred of turnovers comes from my other basketball team, the University of Michigan, and their near total avoidance of turnovers (while winning) during prime Process years. However, let’s take a look and see if my self-inflicted pain is just a regular overreaction, instead of an extreme overreaction.

Turnovers and Wins

From a broad NBA perspective, I assumed that a high number of turnovers would, in general, relate to losing. Figures 1 and 2 relate total turnovers and turnovers per 100 possessions to wins, respectively.

Hmm. Figure 2 seems pretty interesting. It looks like more turnovers equals fewer wins, aligning with my initial assumption. However, certain points that are extreme in both the x-and y-axis (Atlanta in this case) can have an outsized influence on the regression line. There are a multitude of ways to perform statistical testing for this, but this was entirely based on the eye test. Once we remove Atlanta from the plot, we can see that we get more or less nothing. Side note: rookie PG + new coach + tanking + poorly constructed roster = terrible.

So on a very high level, it does not appear that wins and turnovers have a linear relationship thus far in the 2018-19 season. Table 1 presents a variety of turnover data for the entire NBA, via the always great Basketball Reference.

76ers-Specific Turnovers and Wins

In order to delve a little deeper into the turnover issue, we need to break out turnovers into dead ball vs. live ball. Dead ball turnovers occur when the turnover results in a stoppage of play, such as throwing the ball out of bounds or traveling. Live ball turnovers occur when you have the ball stolen or you throw it to the other team. I scraped all the 76ers play-by-play data to isolate live vs. dead ball turnovers. Using some basic light modeling, we can see if the raw counts of total, live, and dead turnovers have an impact on winning or losing. These models are not predictive at this point (not that predicting based on an end of game total is particularly useful), but are for descriptive purposes.

Figure 6, the total model, approaches statistical meaningfulness, and does an okay job of predicting wins and losses. In particular, that 20-plus turnover cutoff appears instructive. Table 2 has all the Sixers’ raw turnover data for each game. You’ll note that the numbers do not perfectly add up in every instance. The Live and Dead columns are based on instances where the play-by-play data says “turnover”, whereas the Total column is a statistic given by NBA.com; there is likely some team vs. player situation involved, but I wanted to show the imperfections in the process as well. The total is off by 7 turnovers over 33 games, compared to the live plus dead sums.

Effect of Live Ball Turnover vs. Dead Ball Turnover

I always believed that live ball turnovers were worse to commit than a dead ball turnover, as a live ball turnover allows for a fast break opportunity. The canonical example would be a point guard picking the pocket of another point guard at half-court and coasting in for an easy layup (if TJ) or dunk (if Ben). On the other hand, losing the ball along the baseline would allow a team to get back on defense before the opposition gets into their sets.

I looked at the outcome of the possession for the opposition immediately following a live vs. a dead ball turnover. Via NBA.com, “[a] team is in possession when a player is holding, dribbling, or passing the ball. A team possession ends when the defensive team gains possession or there is a field goal attempt.” For this specific analysis, I included the results from foul shots as part of the possession. However, if a shot was taken and missed (without a foul), the possession ends, without regard to who rebounds the miss.

I’m going to include the play-by-play results of the analysis because I think they’re interesting. However, you will note that not every Opponent Possession entry makes perfect sense. This is due to some of the complexity of working with text-based play-by-play data and its lack of structural uniformity. Despite some of those not making perfect sense in context, the scoring columns are correct.

Now, let’s take a graphical look at points per opponent possession following each category of Sixers turnover by game in Figure 7. If a point (game) is above the 45 degree line, that means that the opponent scored more points per possession following a live ball turnover than a dead ball one — below the line is the opposite.

This was fairly surprising to me, as I expected that points per live ball turnover would be rather elevated compared to dead ball, but the scoring difference was 1.10 vs. 1.07, essentially the same. Obviously, you can’t just arbitrarily restrict turnovers, and not all are created equal. For example, I’m okay with Ben trying a creative and aggressive pass and failing, but less so with Amir traveling trying to euro step someone on a drive to the hoop. There isn’t exactly a huge takeaway from this piece — but it might lower my blood pressure slightly the next time a fan gets a souvenir from a terrible pass.

Fun and Slightly Stupid Exercise

What if we could magically go back and retcon the season so far, making the Sixers the best team in turnovers per game (12 vs. 16 with some bold rounding)? We are allowed to remove four turnovers per game and replace them with an average Sixers field goal attempt (eFG% of 53.2). With one simulation attempt, the results are the following.

There you go. I just got the Sixers two more wins. You’re welcome.