I acquired all of the Division I team data since 2002, and from that we can observe some fascinating trends and data relationships in the data. This is a multipart series exploring some of that data.
OK. Things are really getting interesting! The last entry on this topic covered Percentage Loss of Ball Forced (%lob), and the conclusion was that it wasn't a helpful stat when trying to estimate Winning Percentage. How important is it to playing "good defense"?
There are a few good barometers for defensive play, and in my opinion, Dean Smith's Points Per Possession Allowed (ppp) is the best pure measure of defensive performance. The modern definition of this stat (ppm) subtracts offensive rebounds from the possession total, which has enormous effects on the final quotient. Here is an example why.
If Team A comes down the court and scores with a quick pass and layup, that team scores 2 points in the possession. However when Team B comes down the court, faces excellent defensive pressure, passes the ball for 25 seconds, misses the shot, gets its rebound, takes another 20 seconds to find a good shot, gets the rebound again, gets the ball knocked out of bounds by the defense, inbounds and finally takes a guarded 2 that goes in, Team B also gets 2 points in its possession.
According to the modern definition that Ken Pomeroy, Bart Torvik, and others use, these two defenses are rated the same because they each had one possession and allowed 2.00 points. Dean Smith's method has Team A's defender allowing 2.00 points per possession while Team B's defender allowed just 0.67 (2 points in 3 possessions). Clearly Team B played better defense, so the Smith Method better describes the defensive play than the Modern Method.
To appreciate the effects of forcing turnovers on the percentage loss of ball, I created a scatter plot of the ppp and %lob of the last 1,669 teams that made the NCAA Tournament. The results were quite a surprise.
The graph shows an expected inverse relationship between Points Per Possession Allowed (ppp) and Percentage Loss of Ball Forced (%lob). The orange trendline through the plot has an equation of:
From this we learn that for every single percentage a team suppresses its opponent, it can expect to reduce the opponent's points per possession by only 0.0068. Keep in mind that the difference between Dean Smith's offensive and defensive goals is 0.100, so forcing turnovers has little impact on the opposition's efficiency.
Shown in red is the current UNC team. From its point in the scatter we can see that it's %lob is among the lowest of the field and its overall defense (ppp) is fairly average. UNC can improve its defense in three ways: reduce opponent fg%, foul less, and increase turnovers forced. From the graph we can see that if UNC improved to being an average turnover forcer, it would slide to the right and down in the graph in a path parallel to the trendline, and find itself suppressing its opponents by a scant 0.025 more. The slope of this line is too flat for %lob to have much of an impact on ppp.
Furthermore, the scatter around this line is not tight. The Measure of Fit (R-square) is an extremely poor 0.108 (0.000-1.000 range). So while the slope is too flat, there is far too much variation as well to accept %lob as a meaningful stat.
This insignificance of %lob comes as an enormous surprise. It seems that denial defense makes an opponent score less easily. The real successes of it apparently come from forcing bad shots, not from forcing turnovers. So, there is way too much emphasis placed on forcing turnovers by coaches, announcers, and fans. We keep on learning!
Next up: Opponent Offensive Rebounding Percentage
OK. Things are really getting interesting! The last entry on this topic covered Percentage Loss of Ball Forced (%lob), and the conclusion was that it wasn't a helpful stat when trying to estimate Winning Percentage. How important is it to playing "good defense"?
There are a few good barometers for defensive play, and in my opinion, Dean Smith's Points Per Possession Allowed (ppp) is the best pure measure of defensive performance. The modern definition of this stat (ppm) subtracts offensive rebounds from the possession total, which has enormous effects on the final quotient. Here is an example why.
If Team A comes down the court and scores with a quick pass and layup, that team scores 2 points in the possession. However when Team B comes down the court, faces excellent defensive pressure, passes the ball for 25 seconds, misses the shot, gets its rebound, takes another 20 seconds to find a good shot, gets the rebound again, gets the ball knocked out of bounds by the defense, inbounds and finally takes a guarded 2 that goes in, Team B also gets 2 points in its possession.
According to the modern definition that Ken Pomeroy, Bart Torvik, and others use, these two defenses are rated the same because they each had one possession and allowed 2.00 points. Dean Smith's method has Team A's defender allowing 2.00 points per possession while Team B's defender allowed just 0.67 (2 points in 3 possessions). Clearly Team B played better defense, so the Smith Method better describes the defensive play than the Modern Method.
To appreciate the effects of forcing turnovers on the percentage loss of ball, I created a scatter plot of the ppp and %lob of the last 1,669 teams that made the NCAA Tournament. The results were quite a surprise.
The graph shows an expected inverse relationship between Points Per Possession Allowed (ppp) and Percentage Loss of Ball Forced (%lob). The orange trendline through the plot has an equation of:
From this we learn that for every single percentage a team suppresses its opponent, it can expect to reduce the opponent's points per possession by only 0.0068. Keep in mind that the difference between Dean Smith's offensive and defensive goals is 0.100, so forcing turnovers has little impact on the opposition's efficiency.
Shown in red is the current UNC team. From its point in the scatter we can see that it's %lob is among the lowest of the field and its overall defense (ppp) is fairly average. UNC can improve its defense in three ways: reduce opponent fg%, foul less, and increase turnovers forced. From the graph we can see that if UNC improved to being an average turnover forcer, it would slide to the right and down in the graph in a path parallel to the trendline, and find itself suppressing its opponents by a scant 0.025 more. The slope of this line is too flat for %lob to have much of an impact on ppp.
Furthermore, the scatter around this line is not tight. The Measure of Fit (R-square) is an extremely poor 0.108 (0.000-1.000 range). So while the slope is too flat, there is far too much variation as well to accept %lob as a meaningful stat.
This insignificance of %lob comes as an enormous surprise. It seems that denial defense makes an opponent score less easily. The real successes of it apparently come from forcing bad shots, not from forcing turnovers. So, there is way too much emphasis placed on forcing turnovers by coaches, announcers, and fans. We keep on learning!
Next up: Opponent Offensive Rebounding Percentage
Last edited:
