Is there a correlation between the pace at which a team plays and their turnovers per possession? The reason I ask is that UW ranks last in the NCAA in possessions per game...but #2 for least turnovers per possession. OTOH, MU ranks #46 in possessions...and #77 in to per poss.
Also, is turnovers per possession the best way to measure the impact of turnovers?
Quote from: The Sultan of South Wayne on January 31, 2012, 08:41:31 AM
Is there a correlation between the pace at which a team plays and their turnovers per possession? The reason I ask is that UW ranks last in the NCAA in possessions per game...but #2 for least turnovers per possession. OTOH, MU ranks #46 in possessions...and #77 in to per poss.
Also, is turnovers per possession the best way to measure the impact of turnovers?
I've covered this a few times (including some old CS posts and a back and forth with bma), but wasn't able to find the exact answer on scoop. However, the answer is simple.
There isn't a correlation between pace and turnover rate. Not offensively or defensively. Not this year. Not for the entire span of Buzz's tenure.* Teams that play faster do not turn the ball over more, nor do they force more turnovers.
*Buzz has said a few times that part of the reason they're turning the ball over more is because they're playing at a faster pace. He also says that he thinks about numbers way more than anyone else out there. Just sayin'.Turnover rate might be the best way to measure the impact of turnovers. Turnover rate margin may be better. Another approach might be ability to force turnovers beyond normal rate. If SHU normally turns the ball over 18% of the time, how much more than that can we force?. Another way may be turnovers to points (like pick-six) or some derivative.
Honestly, I don't know what's best. Turnover margin is the best available data and the default. I spent a little time a few years ago looking at "forced turnover rate", with the hypothesis that the ability to force turnovers goes down as the season matures. I never finished that work, but still believe in the hypothesis and think it may be an area to improve on the idea.
Quote from: The Sultan of South Wayne on January 31, 2012, 08:41:31 AM
Is there a correlation between the pace at which a team plays and their turnovers per possession? The reason I ask is that UW ranks last in the NCAA in possessions per game...but #2 for least turnovers per possession. OTOH, MU ranks #46 in possessions...and #77 in to per poss.
Also, is turnovers per possession the best way to measure the impact of turnovers?
With respect to the latter of the two questions, what do you mean by "impact" of turnovers, i.e. for that specific possession, for the whole game, the entire season? On the surface it may seem remedial, but the impact of a given turnover is dependent upon what your opponent does in the next possession, right? In other words, a strong defensive team that turns over the ball may likely not be impacted at all whereas a poor defensive team likely would be impacted. But can this be extrapolated/averaged over a series of t/o's, say during an entire game or season? Consider the following:
Team A has an opp eFG% of 60% and t/o rate of 10/game.
Team B has an opp eFG% of 40% and t/o rate of 15/game.
(Assuming all else equal including possessions per game)
Simple math would indicate that both teams stand to give up 6 ppg off of turnovers. However, this wouldn't consider the intangible effects of turnovers. I won't mention specific examples, but most MU fans are all too familiar with how a turnover or two can completely change the momentum of the game or kill a team's momentum, thus changing the entire complexity of a game. On the other hand, if a "high turnover" team has a tendency to turnover the ball in the first half but is much more protective in the second half, it could still be quite successful as the likelihood of a first half t/o of being a game-changer pales in comparison to the likelihood of a second-half turnover being the same. I'd suspect you might even find at least one national champion over the past 20 years who was in the bottom quartile nationally in t/o's per possession or game.
In essence, I don't think t/o per possession or even points off t/o's truly measures the impact beyond a single possession... it may give you a general idea overall, but since even a single turnover has the ability to be a game-changer, intangibles play a significant enough role that at best, either of those two stats would be an gross oversimplification with a serious margin for error. But since most stats fall into that category, I'm not sure there's a better way to measure impact of t/o's... of course, other than something along the lines of running an ex-post simulation of a season using a KenPom-type model, comparing it to the actual results, and observing what correlations exist between those variances and actual turnovers, turnover rates, etc., if any.
Thank you both. Henry, I thought I had remembered you mentioning the lack of correlation between pace and turnover rate, but couldn't find anything.
And Benny, I do agree with you that there are turnovers..and then there are TURNOVERS! The reason I ask is that some people seem to really harp on turnovers, but I really don't know how important they really are.
So MU is ranked #77 in turnovers per possession....they are ranked #16 in opponents turnovers per possession at 23.8%. That means that nearly a quarter of their defensive possessions results in a turnover. Not bad. Especially with a eFG% of 45.4%. That means that the reward has been greater than the risk this year.
Quote from: The Sultan of South Wayne on January 31, 2012, 08:41:31 AM
is turnovers per possession the best way to measure the impact of turnovers?
Good responses above, so i will focus on your second question.
i would go further and say that turnovers per possession is the best team statistic in basketball because it is so clean.
In simple terms, teams score an average of 1 point per trip down the court, so as soon as you turn them over you "take that point away" because the trip ends in zero points instead of the average 1.
While a 3 point field goal "adds" 2 points beyond the average, and a 2-point adds a point, you never know how much credit should go to an assist on the play. A blocked shot or other good defensive play that results in a missed shot (and lower eFG%) takes a point ONLY if there is also a defensive rebound to end the possession - so you really need to give .33 points taken away for the defensive rebound and .66 for the block or the stop (the latter of which is the stat we really need to record).
To get a little more advanced, the turnover rate is key because actually while teams are scoring 1.004 points per trip this year, think of it this way:
The 20.7% of time teams are turned over they score 0.00 points per trip
The 79.3% of the time teams are NOT turned over they score 1.27 points per trip down the court.
Therefore, to be really accurate a turnover is really worth 1.27 points until you figure that if you miss a turnover chance, your teammate may get it on the next pass (same trip or possession), so really it is probably worth about 1.2 points - meaning forcing a turnover by yourself is really more valuable than hitting a 2-point shot, and even slightly more valuable than hitting a 3-point shot off of an assist.
The only thing that would make the turnover% more valuable would be if you added transition credit for steals that result in breakaways and thus high percentage shots that help the eFG%.
If Vander Blue steals the ball at midcourt and streaks down for an uncontested dunk, then obviously that turnover is more valuable than forcing a travel, since he takes at least a point away from the opponent, and adds a point for Marquette (scoring 2 points rather than taking his chances on the average 1 point after a travel at the other end).
Now my head hurts.
One of the most interesting things that current predictive models miss is the dynamic interactions between the 4 factors, especially regarding TO rate. I'm sure Buzz is fully aware of these tradeoffs at an intuitive level but I wish there were better models. For example, when Buzz has a smaller team I think he intentionally increases the focus on steals. That is, he tries to increase defensive turnover rate, which has the likely side effects of increasing offensive efg (by getting more fast breaks), but also increasing defense efg% (by getting burned on a few gambles), and also increasing defensive free throw rate (steal attempts can be called fouls. Then if you get into foul trouble and lack depth, there can be a feed forward effect of all stats getting worse. Pretty interesting dynamical system.
Quote from: bamamarquettefan on January 31, 2012, 10:10:04 AM
Good responses above, so i will focus on your second question.
i would go further and say that turnovers per possession is the best team statistic in basketball because it is so clean.
In simple terms, teams score an average of 1 point per trip down the court, so as soon as you turn them over you "take that point away" because the trip ends in zero points instead of the average 1.
While a 3 point field goal "adds" 2 points beyond the average, and a 2-point adds a point, you never know how much credit should go to an assist on the play. A blocked shot or other good defensive play that results in a missed shot (and lower eFG%) takes a point ONLY if there is also a defensive rebound to end the possession - so you really need to give .33 points taken away for the defensive rebound and .66 for the block or the stop (the latter of which is the stat we really need to record).
To get a little more advanced, the turnover rate is key because actually while teams are scoring 1.004 points per trip this year, think of it this way:
The 20.7% of time teams are turned over they score 0.00 points per trip
The 79.3% of the time teams are NOT turned over they score 1.27 points per trip down the court.
Therefore, to be really accurate a turnover is really worth 1.27 points until you figure that if you miss a turnover chance, your teammate may get it on the next pass (same trip or possession), so really it is probably worth about 1.2 points - meaning forcing a turnover by yourself is really more valuable than hitting a 2-point shot, and even slightly more valuable than hitting a 3-point shot off of an assist.
The only thing that would make the turnover% more valuable would be if you added transition credit for steals that result in breakaways and thus high percentage shots that help the eFG%.
If Vander Blue steals the ball at midcourt and streaks down for an uncontested dunk, then obviously that turnover is more valuable than forcing a travel, since he takes at least a point away from the opponent, and adds a point for Marquette (scoring 2 points rather than taking his chances on the average 1 point after a travel at the other end).
Isn't a team's normal amount of turnovers included in the calculation of their points per possession. If so, then turnovers should not reduce their ppp until that opponent's average number of turnovers is exceeded. Or am I missing something?
Quote from: THRILLHO on January 31, 2012, 10:24:01 AM
One of the most interesting things that current predictive models miss is the dynamic interactions between the 4 factors, especially regarding TO rate. I'm sure Buzz is fully aware of these tradeoffs at an intuitive level but I wish there were better models. For example, when Buzz has a smaller team I think he intentionally increases the focus on steals. That is, he tries to increase defensive turnover rate, which has the likely side effects of increasing offensive efg (by getting more fast breaks), but also increasing defense efg% (by getting burned on a few gambles), and also increasing defensive free throw rate (steal attempts can be called fouls. Then if you get into foul trouble and lack depth, there can be a feed forward effect of all stats getting worse. Pretty interesting dynamical system.
Interesting question...