Making Connections: D-TOV% vs. O-Eff
Editor’s Note: This post is a collaboration between myself and Jeremy Conlin. He’s the man behind SuiteSports.com, as well as a contributor to Knickerblogger, ClipperBlog, BuzzFeed Sports and right here at Hickory-High. You can find Jeremy on Twitter, @jeremy_conlin.
The relationships between the Four Factors and efficiency have been fairly well studied and established, however this has usually been done by looking at aggregate, league-wide data across multiple seasons. This helps establish how these statistical categories work together and overlap at a macro level, but we wanted to drill down to the micro.
We grabbed the game logs for each team from last season and calculated offensive and defensive efficiency, as well as the Four Factors and pace for each team in each game. This lets us look at that way these statistics relate to each other on an individual team level, which can often be very different from the league-wide trends. For example, in our first analysis we found that the Memphis Grizzlies, one of the best offensive rebounding teams, actually had a negative correlation between Offensive Rebound Percentage (OREB%) and Offensive Rating (ORTG). Although hitting the offensive glass was a huge strength for the Grizzlies it was a way of covering up for their other offensive holes. The more they were hitting the glass, the more other things weren’t working, and their overall efficiency was suffering.
Today we’re looking at the relationship between Defensive Turnover Percentage (DTO%) and Offensive Rating (ORTG), exploring how opportunities to get out in transition can effect overall offensive efficiency. The individual team graphs are below.
This next visualization combines all the individual team information into one graph. The height of each bar represents how strong the correlation was between DTO% and ORTG. The color of each bar represents the team’s ORTG. The width of each bar represents their DTO%.
At the beginning of most NBA broadcasts, there will be a short segment right before tip-off where the color commentator outlines “Keys to The Game,” often sponsored by an automobile (because “keys,” get it? Hilarious). If you watch a whole bunch of games, it becomes an inevitability that the analyst will mention “avoiding turnovers” as one of the keys to the game.
The reason for this is fairly straightforward. For one, it’s slightly more creative than “score more points than the other team.” But for the most part, it’s because there is a dual-faceted negative to committing turnovers. First, committing a turnover on offense precludes the team from scoring a basket on that possession. That’s just common sense. And second, a turnover often creates a transition opportunity for the other team, which often leads to an easy layup.
Turns out, maybe not. According to the numbers we put together here, most teams actually see a NEGATIVE correlation between DTO% and ORTG. As in, the more turnovers a team forces on offense, the worse their offense performs. When we charted all 82 games (or 81 in the case of Boston and Indiana) for all 30 teams, we found that 22 of the 30 teams saw trend lines with negative slopes, and the league average correlation from defensive turnover rate to offensive rating was -8%.
Even more surprising was which teams finished where. For example, the Heat, perhaps the most explosive open-court team of recent memory, saw effectively zero correlation (-0.6%) between the rate at which they force turnovers on defense and their offensive efficiency. Other teams that are often seen as explosive in transition, like Oklahoma City, Houston, and the Clippers, also saw negative scores. Oklahoma City actually ranked 22nd in the league with a correlation of almost -15%.
The obvious question here, is why? Why is it that forcing turnovers on defense seems to cause an adverse reaction for an offense? The simple answer is that we don’t know, but we can hazard a few guesses. The first and most obvious guess is that we’re assuming causation where there isn’t any. Because all of the correlation percentages are so low, the soundest explanation might be that defensive turnover rate and offensive efficiency just don’t correlate at all, and that it’s merely a coincidence that 22 of 30 teams saw negative correlation percentages. We could try to expand the data set to include more seasons and see if that trend continues.
This lack of correlation in general could be explained by the difference between live-ball and dead-ball turnovers. All live-ball turnovers are recorded as steals. All dead-ball turnovers are not. So it’s pretty easy to calculate how many turnovers actually lead to a transition opportunity for the other team. In 2013, the league average for turnovers was 1192, and the league average for steals was 639. Basic math would then tell us that 53% of turnovers were of the live-ball persuasion.
If it’s roughly a 50/50 proposition, it might be possible that the advantage gained from a live-ball turnover (a transition opportunity in the other direction) is offset by the disadvantage of a dead-ball turnover (the stop in play allows the defense to set up). And that doesn’t even account for the fact that not all live-ball turnovers immediately lead to a fast break – like the defense stripping the ball in traffic under their own basket, or a scrum for a loose ball that ends up with more bodies on the floor than leaking out.
We had one other idea, that is perhaps more far-fetched, but still worth mentioning. Generally speaking, forcing a turnover requires the defense to expend more effort than they might normally – it’s pretty rare to see an offensive player simply pass the ball directly to a defender (unless it’s Fred Brown throwing the ball to James Worthy) or dribble the ball off his foot under no duress. So assuming that to be true, perhaps it’s the case that the extra effort expended to force a turnover makes it harder to perform on the subsequent offensive possession.
In any event, the only thing we seemed to have learned is that forcing turnovers doesn’t correlate to offensive performance the way many of us might have assumed. Possible next steps include expanding the data set to include other seasons (to see if the negative correlations continue to occur), or perhaps looking at just steal rate as opposed to turnover rate – looking only at live-ball turnovers might yield different results.