[2级,继续招工]Seth Partnow专栏:NBA Offensive Styles Analysis, Part 4: In comparison to NCAA, WNBA, G Le...由JabariIverson 发表在翻译团招工部 https://bbs.hupu.com/fyt-store
Over the past several weeks, I’ve been digging into historical “play type” information collected from Synergy Sports. In Part One, I discussed some general trends in the NBA over the last decade and a half. In Part Two, I argued that the level of stylistic diversity in teams’ offensive attacks is as large as it’s ever been, and in Part Three I looked at how the 3-point revolution interacts with offensive style. As noted in those earlier pieces, the Synergy data set is imperfect. There is some degree of internal inconsistency built into a manually tracked dataset, and this variation only increases over a multi-season time frame as things such as changing definitions and adjudications of edge cases — how close to the basket does a player have to be before an iso becomes a post up? Further, successive new cadres of individual trackers naturally bring slightly different understandings with them.
These inconsistencies can make this data tricky to work with in terms of model building or other higher order statistical manipulation. But for the purposes of exploration and description such as these broad surveys of the data, the historical depth allows for identification of broad trends in a way that does not require multiple decimal point precision.
Another benefit of using Synergy play type data for this study is cross-league comparability. The exponentially larger space for exploration allowed by Second Spectrum/SportVU player tracking data has no real analogue in any non-NBA league, at least not in public. A smattering of college programs have tracking cameras available either as result of sharing an arena with an NBA team or their own initiative, while according to an informal survey, around half of the G League arenas have installed tracking capability in the last season or so. Of course, for our purposes even that degree of availability is immaterial as none of that data is available publicly.
Conversely, Synergy largely uses the same tagging conventions for the G League, NCAA and WNBA play and even tracks NBA Summer League play. All the usual caveats about the accuracy and consistency of manually tracked data apply, but if we’re interested in using aggregated statistics for a broad overview of how the styles of play differ in these various environs, those issues aren’t particularly damaging.
I’m going to briefly touch on NCAA and WNBA results, as those competitors are worth further investigation in their own right, before spending a bit more time on the closest analogues to NBA play we have: Summer League and the G League. Overall, it would appear that at least in the American game, similar trends are occurring at all levels of play, with pick-and-rolls replacing isolations as the go-to for offensive initiation.
NCAA Men’s Play
There is decent coverage of the college game going back to the 2008-09 season in the data set, at least if we’re mostly concerned with the “name” schools. For my sanity in data collection and assembly as anything else, I decided to focus on six prominent leagues, the ACC, Big 12, Big East, Big 10, PAC-12 and SEC. This captures most of the usually NBA-relevant programs. Though there are a few notable omissions such as Gonzaga, UConn and Syracuse, schools from those conferences account for roughly 80 percent of the players who came through the American college system and appeared in an NBA game in 2019-20.
The general trends observed for how the NBA style has changed apply to the high major NCAA game, as well, to a degree. There are some differences as might be expected while the NBA has seen a split of roughly 50-50 (generally around 48-52 most seasons) between on-ball and off-ball actions, the gap is wider in the NCAA, with closer to 40 percent of plays coming on ball and 60 percent from off the ball. At this point, any explanation for this difference would be largely conjecture, but plausible theories include the longer shot clock, the relatively lower skill level of even primary ballhandlers as well as the prevalence
But much like what has been observed in the NBA, that ratio has held steady, and further similarities are evident: While the ratio of individual off-ball play types have stayed roughly constant, pick-and-rolls have replaced isos and post-ups as the preferred on-ball actions.
As mentioned above, off-ball play types have been more prevalent in NCAA than NBA play:
With the bulk of the gap seemingly made up by the degree to which spot-ups are the most heavily used play type in high major college play. However, this extreme reliance on spot-ups is somewhat counter-cyclical to recent rules changes, with the longer 22.25-foot FIBA 3-point line being adopted prior to last season. Perhaps relatedly, efficiency on spot-up attempts took a nosedive last season:
In the future, I’ll come back and look into possible explanations for this dip in spot-up attempts, as well as investigate the relative diversity between teams. There are certainly some indications of different styles between leagues. For example, the majority of the power conferences have seen a steep decline in post up frequency. With one, very on-brand, exception:
Average Post Up%- Power 6 ConferencesSEASONACCBIG 12BIG EASTBIG TENPAC-12SEC200816.2%15.1%11.7%11.7%13.6%15.0%200916.7%14.7%12.6%12.6%12.4%14.4%201012.1%15.0%13.3%13.8%12.5%14.3%201114.0%13.9%12.4%12.0%12.8%13.2%201212.7%14.4%14.2%12.9%13.2%13.0%201311.8%13.0%14.2%10.8%12.5%12.2%201412.6%12.5%11.4%11.4%12.8%11.7%201511.9%12.8%12.3%13.6%12.2%12.2%201611.4%10.5%10.4%13.4%12.0%13.1%20179.7%11.0%11.2%13.6%10.7%12.3%20188.9%9.5%11.4%13.9%9.4%11.0%20199.6%10.2%10.0%13.9%11.5%7.8%
An additional avenue for exploration was suggested to me by Jordan Sperber and backed up with some good work by Pivot Analysis, that while the NBA is predominantly a “shooting” league in that eFG% for and against are by far the biggest determinants of success, the remaining three Four Factors have relatively more importance in the college games:
Pivot Analysis@Pivot_Analysis
The recent @hoopvision68 podcast with @SethPartnow touched on some great topics crossing the NCAA and NBA. One thing that we like to focus on is the importance of the #FourFactors. Based on this season, this is how the net four factors contributed to opponent-adjusted efficiency.
199:22 PM - Apr 25, 2020Twitter Ads info and privacySee Pivot Analysis's other Tweets
(Note that given the vagaries of schedule length and the number of games captured, I’ve done some manipulation of the sample to ensure individual teams don’t have too much or too little impact on the overall averages.)
WNBA
Synergy has relatively good data on WNBA play going back to the 2010 season. The trends over that time have been very similar to those in the NBA. The ratio of on-ball to off-ball scoring plays is in the same basic range, season by season, especially when compared to high major NCAA Men’s play:
Much like the NBA, the WNBA has seen a steady replacement of plays out of isolations with pick-and-rolls:
One other data point that caught the eye is the steep decline in half-court offensive efficiency across the WNBA last season. It went from a sample-period high in 2018 to the 3rd lowest mark, easily the largest single season change up or down over this time period:
In a later installment, I’ll examine these WNBA trends more fully to explore the causes and how the women’s game is changing in similar but not identical ways to the men’s game.
G League and Summer League Play
Though it is certainly notable that top college leagues and the WNBA have seen similar stylistic trends as the NBA over the last decade-plus, it is difficult to draw too much from the comparison given the differences in rule sets and player pools. There remain some differences: Summer League games are shorter, have non-standard foul out and overtime rules and tend to be officiated differently (at least in my observation). And of course, the overall talent level in both environments is far lower than in the NBA proper. Still, these remain the situations where we have the best chance to examine players in NBA-like game situations.
Direct translation of statistical accomplishment is hard enough. As my colleague John Hollinger put it recently, NBA level players tend to put up “video game numbers” in the G League, while Summer League can be a small sample size crapshoot, though it still has a small degree of predictive power for first-year players. The difficulty is exacerbated by the degree to which the NBA isn’t just a higher level of competition but also encompasses a different style of play.
For example, it is often said that evaluating bigs in Summer League is borderline impossible because of the extent to which the play tends to be guard dominated. In examining the comparative data, I would amend that statement to say that G League, and especially Summer League play, aren’t guard dominated so much as guard limited. That minor league play is a higher turnover environment than the NBA should be unsurprising, but the extent to which the additional turnovers accrue to on-ball play types show how and why teammate-dependent scorers might have a harder time shining than as a result of the less ball-secure offensive initiators found at that level.
The chart above demonstrates the degree to which the turnover rates in G-League and especially Summer League play are higher across all play types and categories. However, at these sub-NBA levels, turnover frequency increases to a much larger degree for the on-ball actions than off-ball. Though the data is insufficiently granular to definitively prove the point, this is at minimum an indicator of the difficulties faced by the poor guys who rely on getting the ball on time and in the right spots to be effective.
Overall, the trends in play utilization between the NBA and the G League are incredibly similar for the seasons with data for both (2008-09 on):
This result is honestly a little surprising to me. Not the part where G League teams play a recognizably similar style to their NBA brethren in recent years, but rather the degree to which it has long been so. There were only 16 teams in the D-League in 2008-09, and team-to-team affiliation where an NBA team assigned players to a single D-league squad as opposed to the overall league did not begin until 2009-10. Direct ownership and co-operation of D-league teams as a more traditionally minor league extension did not become commonplace until well into this decade. So the degree to which the G League hasn’t just ended up right around where the NBA is in terms of today’s style, but that it was already there to begin with, is not what I expected to find.
This was largely a quick survey of the various league comparisons. Drop some questions in the comments for deeper questions you might have. I currently have two more articles planned for this series, one flipping things around and looking at defense, and the other examining the play types we’ve knowingly excluded from what has so far been a half-court study — transition and offensive rebound/putback plays.
Over the past several weeks, I’ve been digging into historical “play type” information collected from Synergy Sports. In Part One, I discussed some general trends in the NBA over the last decade and a half. In Part Two, I argued that the level of stylistic diversity in teams’ offensive attacks is as large as it’s ever been, and in Part Three I looked at how the 3-point revolution interacts with offensive style. As noted in those earlier pieces, the Synergy data set is imperfect. There is some degree of internal inconsistency built into a manually tracked dataset, and this variation only increases over a multi-season time frame as things such as changing definitions and adjudications of edge cases — how close to the basket does a player have to be before an iso becomes a post up? Further, successive new cadres of individual trackers naturally bring slightly different understandings with them.
These inconsistencies can make this data tricky to work with in terms of model building or other higher order statistical manipulation. But for the purposes of exploration and description such as these broad surveys of the data, the historical depth allows for identification of broad trends in a way that does not require multiple decimal point precision.
Another benefit of using Synergy play type data for this study is cross-league comparability. The exponentially larger space for exploration allowed by Second Spectrum/SportVU player tracking data has no real analogue in any non-NBA league, at least not in public. A smattering of college programs have tracking cameras available either as result of sharing an arena with an NBA team or their own initiative, while according to an informal survey, around half of the G League arenas have installed tracking capability in the last season or so. Of course, for our purposes even that degree of availability is immaterial as none of that data is available publicly.
Conversely, Synergy largely uses the same tagging conventions for the G League, NCAA and WNBA play and even tracks NBA Summer League play. All the usual caveats about the accuracy and consistency of manually tracked data apply, but if we’re interested in using aggregated statistics for a broad overview of how the styles of play differ in these various environs, those issues aren’t particularly damaging.
I’m going to briefly touch on NCAA and WNBA results, as those competitors are worth further investigation in their own right, before spending a bit more time on the closest analogues to NBA play we have: Summer League and the G League. Overall, it would appear that at least in the American game, similar trends are occurring at all levels of play, with pick-and-rolls replacing isolations as the go-to for offensive initiation.
NCAA Men’s Play
There is decent coverage of the college game going back to the 2008-09 season in the data set, at least if we’re mostly concerned with the “name” schools. For my sanity in data collection and assembly as anything else, I decided to focus on six prominent leagues, the ACC, Big 12, Big East, Big 10, PAC-12 and SEC. This captures most of the usually NBA-relevant programs. Though there are a few notable omissions such as Gonzaga, UConn and Syracuse, schools from those conferences account for roughly 80 percent of the players who came through the American college system and appeared in an NBA game in 2019-20.
The general trends observed for how the NBA style has changed apply to the high major NCAA game, as well, to a degree. There are some differences as might be expected while the NBA has seen a split of roughly 50-50 (generally around 48-52 most seasons) between on-ball and off-ball actions, the gap is wider in the NCAA, with closer to 40 percent of plays coming on ball and 60 percent from off the ball. At this point, any explanation for this difference would be largely conjecture, but plausible theories include the longer shot clock, the relatively lower skill level of even primary ballhandlers as well as the prevalence
But much like what has been observed in the NBA, that ratio has held steady, and further similarities are evident: While the ratio of individual off-ball play types have stayed roughly constant, pick-and-rolls have replaced isos and post-ups as the preferred on-ball actions.
As mentioned above, off-ball play types have been more prevalent in NCAA than NBA play:
With the bulk of the gap seemingly made up by the degree to which spot-ups are the most heavily used play type in high major college play. However, this extreme reliance on spot-ups is somewhat counter-cyclical to recent rules changes, with the longer 22.25-foot FIBA 3-point line being adopted prior to last season. Perhaps relatedly, efficiency on spot-up attempts took a nosedive last season:
In the future, I’ll come back and look into possible explanations for this dip in spot-up attempts, as well as investigate the relative diversity between teams. There are certainly some indications of different styles between leagues. For example, the majority of the power conferences have seen a steep decline in post up frequency. With one, very on-brand, exception:
Average Post Up%- Power 6 ConferencesSEASONACCBIG 12BIG EASTBIG TENPAC-12SEC200816.2%15.1%11.7%11.7%13.6%15.0%200916.7%14.7%12.6%12.6%12.4%14.4%201012.1%15.0%13.3%13.8%12.5%14.3%201114.0%13.9%12.4%12.0%12.8%13.2%201212.7%14.4%14.2%12.9%13.2%13.0%201311.8%13.0%14.2%10.8%12.5%12.2%201412.6%12.5%11.4%11.4%12.8%11.7%201511.9%12.8%12.3%13.6%12.2%12.2%201611.4%10.5%10.4%13.4%12.0%13.1%20179.7%11.0%11.2%13.6%10.7%12.3%20188.9%9.5%11.4%13.9%9.4%11.0%20199.6%10.2%10.0%13.9%11.5%7.8%
An additional avenue for exploration was suggested to me by Jordan Sperber and backed up with some good work by Pivot Analysis, that while the NBA is predominantly a “shooting” league in that eFG% for and against are by far the biggest determinants of success, the remaining three Four Factors have relatively more importance in the college games:
Pivot Analysis@Pivot_Analysis
The recent @hoopvision68 podcast with @SethPartnow touched on some great topics crossing the NCAA and NBA. One thing that we like to focus on is the importance of the #FourFactors. Based on this season, this is how the net four factors contributed to opponent-adjusted efficiency.
199:22 PM - Apr 25, 2020Twitter Ads info and privacySee Pivot Analysis's other Tweets
(Note that given the vagaries of schedule length and the number of games captured, I’ve done some manipulation of the sample to ensure individual teams don’t have too much or too little impact on the overall averages.)
WNBA
Synergy has relatively good data on WNBA play going back to the 2010 season. The trends over that time have been very similar to those in the NBA. The ratio of on-ball to off-ball scoring plays is in the same basic range, season by season, especially when compared to high major NCAA Men’s play:
Much like the NBA, the WNBA has seen a steady replacement of plays out of isolations with pick-and-rolls:
One other data point that caught the eye is the steep decline in half-court offensive efficiency across the WNBA last season. It went from a sample-period high in 2018 to the 3rd lowest mark, easily the largest single season change up or down over this time period:
In a later installment, I’ll examine these WNBA trends more fully to explore the causes and how the women’s game is changing in similar but not identical ways to the men’s game.
G League and Summer League Play
Though it is certainly notable that top college leagues and the WNBA have seen similar stylistic trends as the NBA over the last decade-plus, it is difficult to draw too much from the comparison given the differences in rule sets and player pools. There remain some differences: Summer League games are shorter, have non-standard foul out and overtime rules and tend to be officiated differently (at least in my observation). And of course, the overall talent level in both environments is far lower than in the NBA proper. Still, these remain the situations where we have the best chance to examine players in NBA-like game situations.
Direct translation of statistical accomplishment is hard enough. As my colleague John Hollinger put it recently, NBA level players tend to put up “video game numbers” in the G League, while Summer League can be a small sample size crapshoot, though it still has a small degree of predictive power for first-year players. The difficulty is exacerbated by the degree to which the NBA isn’t just a higher level of competition but also encompasses a different style of play.
For example, it is often said that evaluating bigs in Summer League is borderline impossible because of the extent to which the play tends to be guard dominated. In examining the comparative data, I would amend that statement to say that G League, and especially Summer League play, aren’t guard dominated so much as guard limited. That minor league play is a higher turnover environment than the NBA should be unsurprising, but the extent to which the additional turnovers accrue to on-ball play types show how and why teammate-dependent scorers might have a harder time shining than as a result of the less ball-secure offensive initiators found at that level.
The chart above demonstrates the degree to which the turnover rates in G-League and especially Summer League play are higher across all play types and categories. However, at these sub-NBA levels, turnover frequency increases to a much larger degree for the on-ball actions than off-ball. Though the data is insufficiently granular to definitively prove the point, this is at minimum an indicator of the difficulties faced by the poor guys who rely on getting the ball on time and in the right spots to be effective.
Overall, the trends in play utilization between the NBA and the G League are incredibly similar for the seasons with data for both (2008-09 on):
This result is honestly a little surprising to me. Not the part where G League teams play a recognizably similar style to their NBA brethren in recent years, but rather the degree to which it has long been so. There were only 16 teams in the D-League in 2008-09, and team-to-team affiliation where an NBA team assigned players to a single D-league squad as opposed to the overall league did not begin until 2009-10. Direct ownership and co-operation of D-league teams as a more traditionally minor league extension did not become commonplace until well into this decade. So the degree to which the G League hasn’t just ended up right around where the NBA is in terms of today’s style, but that it was already there to begin with, is not what I expected to find.
This was largely a quick survey of the various league comparisons. Drop some questions in the comments for deeper questions you might have. I currently have two more articles planned for this series, one flipping things around and looking at defense, and the other examining the play types we’ve knowingly excluded from what has so far been a half-court study — transition and offensive rebound/putback plays.
推荐
评论 (5)
收藏
分享
举报
只看楼主