New to the group.
I am a former NCAA Div I men head/asst coach for many years.
I'm not working with coaches as a mentor and work closely with several schools and their coaching staffs.
I'm looking for advice: What 3-5 things would be most useful for a college coach, staff, team?
I want to take the world of analytics/metrics/SportVU and condense it down to areas that coaches would look at and see value in.
In studying this field for many years, I get lost in the detail.
Any advice would be very helpful,
Randy Brown
Ames, Iowa
Former NCAA coach--Arizona, Marquette, Iowa State
NCAA Div I uses for Analytics--new to group
Re: NCAA Div I uses for Analytics--new to group
I know Vantage Sports is interested in providing their services to college teams. May want to get in touch with them, they have a number of statistics that could be useful to a coach. And you could pick and choose what to present to the coaches.
Re: NCAA Div I uses for Analytics--new to group
thanks for the tip--I will contact them--appreciate the follow up post,
RB
RB
Re: NCAA Div I uses for Analytics--new to group
Very biased obviously, but a player ranking system (that rates/ranks all D1 players) could help some D1 coaches when evaluating whether or not to consider taking in a transfer or grad transfer (especially from a small school to a mid or major school), since it adjusts for pace/SoS/etc. I rank all D1 players at the site in my sig - my next update should be this Tuesday morning.coachrb wrote:New to the group.
I am a former NCAA Div I men head/asst coach for many years.
I'm not working with coaches as a mentor and work closely with several schools and their coaching staffs.
I'm looking for advice: What 3-5 things would be most useful for a college coach, staff, team?
I want to take the world of analytics/metrics/SportVU and condense it down to areas that coaches would look at and see value in.
In studying this field for many years, I get lost in the detail.
Any advice would be very helpful,
Randy Brown
Ames, Iowa
Former NCAA coach--Arizona, Marquette, Iowa State
Here was a complete TRANSFERS ranking before this last season - based on each player's previous rating the last season they played:
http://classic.hoopsnerd.com/uploads/NC ... 013-14.pdf
Check back at my site during the summer - I should have the last 18 seasons of D1 players rated & ranked (June?) - and I'll be projecting player career curves for both college and the NBA (based on age and established statistical skillset breakdowns). Outside of the projection of possible transfer options, not sure how much it'd help college coaches - but the results might ineterest you in comparison to how you've evaluated these same players over the years.
BTW, I'm an Arizona alum - I often think fondly back to the days of Lute, and actually played pick up often w/ many past Arizona greats at the rec center (Jud, Khalid, Rooks, Mills, etc). Good times.
Re: NCAA Div I uses for Analytics--new to group
I'll try to answer your question... and then I will go further because I had more to say and it is easier and maybe better to keep most of it together here. (I put other stuff in the misc. notes thread.)
The four most important elements of basketball are field goal shooting, free throw shooting, turnovers and rebounding (for offense and defense), i.e. the Four Factors. Should be true whether one uses "analytic" language and tools or not. If some of the detail that follows makes your eyes glaze over (sorry, thought it was worth saying), remember this and work on improving the 4 Factors your way.
Good basketball is based on intelligent thinking and effective action and mainly in these four areas. "Analytics" should help players' and coaches' thinking and action. It use different tools and language but it should also be intelligent thinking and aimed at aiding effective and specific action on the court. They really should be considered in my opinion as part of the same approach rather than a different beast. Intelligent thinking about basketball involves looking at most or everything that the players and coaches are doing on the court and asking we do this BECAUSE and we have this information to show what results are associated with this action and what results are associated with the available options we have tried or thought of and we believe the option we have chosen to be the best one (in general or for this circumstance) for team performance on the 4 factors overall (not just one).
We try to enter the ball into the post (by pass or drive) BECAUSE we think it will increase offensive efficiency of that possession and we have this data that shows that plays where the ball goes into the post yield +X% more points than plays where the ball doesn't go into the post. We box out in this situation or in general BECAUSE we have data that shows that our rebound rate goes up when you have 2+ block outs. We defend a pick n roll this way BECAUSE we data that shows that is the best way for our team to do it, not just because this is the way our coach's coach or mentor did / does it. We have players heavily behind the 3 point line BECAUSE we have data that shows ouroffensive efficiency is better that way than not. We have a player run into the corner on a certain play 2 seconds before, simultaneous with another player movement or 1-2 seconds after BECAUSE we've looked to see which works best. We move the move side to side on average 3 or 5 times per play BECAUSE we've looked at the data for doing so against this team, this kind of man or zone, etc.
The largest and most important factor to basketball success is shooting / scoring. Therefore analytic efforts should focus on this heavily.
I'd start with defensive goal of impairing opponent scoring by limiting “good” shots (clear path layups, FTAS, other inside shots, 3s, really open 2s) and contesting good shots. Know where they get their good shots and realistically strategize what shots you can prevent, impair or nullify via steals or blocks.
The next goal would be increasing own scoring by maximizing the good shots your team can get and hit including uncontested and lightly contested shots. Review the tape of every bad / no shot plays and figure out how you could have gotten a good shot. Convince players that contested mid to long range 2 attempts except late in the clock are not good shots and not acceptable unless the team is completely failing to get any good shots. Everybody needs to know, agree upon and accept who your best 3 scorers are and who they aren’t. If you aren’t top 3 you mostly focus on something else- passing, picks or rebounding. Look at who is good or bad at shot types: pick n roll, pick n pop, spot-up, pull-up, post -up, drive, etc. players should do more of what they are good at and less of what they are below average at or below their teammate on the court can do instead.
Make sure lineups are balanced and players are familiar with each other and fill their roles. Best 3 scorers ideally should be most efficient but you probably want / need to get 60+% of the shots from them too.
Given the better offensive efficiency generally seen early in the shot clock and in higher pace style, play crisp but do not rush. Aim for 70+% of shots in first half of shot clock.
Teams and players need to analyze both their shot selection and their shooting efficiency and separately for clarity.
As for analyzing shot selection, one simple way for coaches and players to look at it might be to assign the following values to shot attempts (based on the relative historical average yields from these shots):
2 “pts” for clear path layup attempts
1.5 for possessions ending in FTAS,
1.5 for open 3s,
1 for inside shots,
0.5 for moderate to heavier contested 3s
0.5 really open outside 2s
0 contested outside 2s
Then. sum scores for shots taken = shot selection enhancement above minimum (i.e. the worst shot, contested outside 2s).
Sum scores for shots taken / shots taken = shot selection enhancement efficiency.
these two measures will help you know who are getting you more better shots and who is getting them more often.
Analyze your wins and losses, especially at team factor level. Every player needs to know the 1-2 things they most need to do to help make the team better on the 4 Factors. If they can’t / won’t do it better in next few games, get somebody else to the extent possible unless they are otherwise absolutely essential given your roster. Analyze your 3-5-10 most used linueps at the 4 Factors level as well.
For the college game I would especially look at which players and lineups are on the court when the team goes on an 8+ point run for or against. Know which PG really plays best with the other 4 starters and each individual player.
BECAUSE rebounding is probably the second most important part of the game after shooting/scoring make sure your team is looking at it carefully and correctly, that using rebound rates and looking at rebounding opportunities and capture rates and such and not just raw game totals. If one really wants to learn as much as possible and improve as much as possible I'd look at player and team rebounding performance by part of the shot clock, stage of game, with lead/ behind, home and away, in bonus/not, in man to man or zone or against the same, etc. Not just to collect and stare at numbers but to understand what really happened and what needs to be improved on the court to get better.
BECAUSE turnovers are the remaining major element of the game don't settle for a raw count. Know where they occur on the court, on theclock, what type of error they are and who contributed to the turnover. That is practical intelligent analytics aimed at more effective action.
If you want to look at individual player performance in a single metric, look at using an advanced statistical plus minus like DSMOK1's.
http://godismyjudgeok.com/DStats/aspm-and-vorp/
If there is the talent or money to do so it might be interesting to do a RAPM run exclusively for a conference (in conference games) and mainly focused on finding estimates for the starters. If the skill was available I'd take the analysis down to 4 Factor level. Sample sizes are small so maybe multi-season would help a bit.
If you have access to SportVU data on layer movement work with it until it gives you specific information that can help improve on the court action rather than reject it as esoteric.
Example (extended consideration, skip if not wanting all of it): Instead of scoffing as Stan Van Gundy did about the lack of usefulness of knowing that Paul George ran the most miles this season, drill down into the detail. SportsVU may only report total miles and average speed to the public but it has detail on every segment of time and how far and fast every player moved. If you divided player and team data that occurred in .25 mile or .10 mile increments one could look to see if player or team stats for eFG%, 3pt FG%, inside FG%, FGA, 3pt FGA, inside FGA, FTA, assists, overall usage, steals, blocks, reb%, personal fouls, etc. per possession or minute changed as total miles increased increment by increment thru a game or thru a stint on the court. You could look to see if performance for any of these declined if distance covered divided by time exceeded certain levels (.0x mile / Y minutes) and judge when the right time is to substitute each player and perhaps what the upper limit should be on team pace when that player is on the court. If you know how far a player has run and how he has performed, one could then look at tape as see how much of his movement was appropriate and where specifically it could be best streamlined or made more exerting.
If you found out that Russell Westbrook for the season or the last month drove inside twice as much in his first mile of running as the second that would seem useful to learn and act on. Same for if you found out Durant increased his 3pt FGA rate by 50% if the distance travel in the prior 3-5 minutes was say 25% above average. If Perkins’ block rate went down 1/3 rd after the first half mile run or if .X miles / Y minutes exceeded Z, that also seems worth knowing. And so on in dozens of ways. If you knew Ibaka’s PF rate went up 40% when distance for a half of a game was 20% above normal that could affect when you should sit him, especially when near “foul trouble”. If your team depends on steals it might be worth knowing how successful your team is at getting steals when the opponent team or just the PG is running at above average, average and below average speed or the rate of travel in the last few minutes was high, medium or low.
If you have adopted a high pace, it would seem particularly important to know how your current players perform at high pace vs. average and low and how well available players do as well for these splits. Perhaps there is a better fit player for your style of play.
The analysis could extend to how player and team performance varies based on schedule variances (high number of games in a week, back to backs, month to month). Small changes in rotation management might be useful. The analysis gives information that can be considered. Judgment still must be applied. But a better informed decision-maker might do better. If they are open and able to handle the information. Instead of just looking for a way to dismiss and belittle it. I would assume that Mr. Van Gundy and most coaches try to gauge how players perform as they log miles and the speed of those miles vary. But they are doing that in a mini-window as the game progresses or the tape watching progress along with many other evaluations. Who is more likely to have accurate information and good insights and appropriate adjustments, the multi-tasking coach sometime and vaguely monitoring this or the data system and the analysts that have access to thousands or millions of observations and have the ability to study it carefully and find patterns to discuss with the coach? I know my answer but the kicker is that this is not necessarily an either or choice. Do it both ways and compare and discuss and study some more. Both ways.
If you had a long-range database about miles traveled and their intensity for the league, one could look at the possible correlation between them and injuries (to the leg or anywhere) and the possible impact of them to the “aging curve”. In general and for particular types of players and particular styles of play.
If you divide the distance traveled into the horizontal and vertical axises of the court one could see how far a player travels on each on average on a fast break (offense or defense) and within the first 2-4 seconds (vs. eventually). You could see who is running straight, fast, in lanes, who stays active and who quits, etc.
The sportsVU data could yield plenty of information in other areas to aid intelligent and effective action if you work it hard.
I know this rambled and got real specific in places but that is what came to mind in response to the question posed. I might revise and extend further later.
The four most important elements of basketball are field goal shooting, free throw shooting, turnovers and rebounding (for offense and defense), i.e. the Four Factors. Should be true whether one uses "analytic" language and tools or not. If some of the detail that follows makes your eyes glaze over (sorry, thought it was worth saying), remember this and work on improving the 4 Factors your way.
Good basketball is based on intelligent thinking and effective action and mainly in these four areas. "Analytics" should help players' and coaches' thinking and action. It use different tools and language but it should also be intelligent thinking and aimed at aiding effective and specific action on the court. They really should be considered in my opinion as part of the same approach rather than a different beast. Intelligent thinking about basketball involves looking at most or everything that the players and coaches are doing on the court and asking we do this BECAUSE and we have this information to show what results are associated with this action and what results are associated with the available options we have tried or thought of and we believe the option we have chosen to be the best one (in general or for this circumstance) for team performance on the 4 factors overall (not just one).
We try to enter the ball into the post (by pass or drive) BECAUSE we think it will increase offensive efficiency of that possession and we have this data that shows that plays where the ball goes into the post yield +X% more points than plays where the ball doesn't go into the post. We box out in this situation or in general BECAUSE we have data that shows that our rebound rate goes up when you have 2+ block outs. We defend a pick n roll this way BECAUSE we data that shows that is the best way for our team to do it, not just because this is the way our coach's coach or mentor did / does it. We have players heavily behind the 3 point line BECAUSE we have data that shows ouroffensive efficiency is better that way than not. We have a player run into the corner on a certain play 2 seconds before, simultaneous with another player movement or 1-2 seconds after BECAUSE we've looked to see which works best. We move the move side to side on average 3 or 5 times per play BECAUSE we've looked at the data for doing so against this team, this kind of man or zone, etc.
The largest and most important factor to basketball success is shooting / scoring. Therefore analytic efforts should focus on this heavily.
I'd start with defensive goal of impairing opponent scoring by limiting “good” shots (clear path layups, FTAS, other inside shots, 3s, really open 2s) and contesting good shots. Know where they get their good shots and realistically strategize what shots you can prevent, impair or nullify via steals or blocks.
The next goal would be increasing own scoring by maximizing the good shots your team can get and hit including uncontested and lightly contested shots. Review the tape of every bad / no shot plays and figure out how you could have gotten a good shot. Convince players that contested mid to long range 2 attempts except late in the clock are not good shots and not acceptable unless the team is completely failing to get any good shots. Everybody needs to know, agree upon and accept who your best 3 scorers are and who they aren’t. If you aren’t top 3 you mostly focus on something else- passing, picks or rebounding. Look at who is good or bad at shot types: pick n roll, pick n pop, spot-up, pull-up, post -up, drive, etc. players should do more of what they are good at and less of what they are below average at or below their teammate on the court can do instead.
Make sure lineups are balanced and players are familiar with each other and fill their roles. Best 3 scorers ideally should be most efficient but you probably want / need to get 60+% of the shots from them too.
Given the better offensive efficiency generally seen early in the shot clock and in higher pace style, play crisp but do not rush. Aim for 70+% of shots in first half of shot clock.
Teams and players need to analyze both their shot selection and their shooting efficiency and separately for clarity.
As for analyzing shot selection, one simple way for coaches and players to look at it might be to assign the following values to shot attempts (based on the relative historical average yields from these shots):
2 “pts” for clear path layup attempts
1.5 for possessions ending in FTAS,
1.5 for open 3s,
1 for inside shots,
0.5 for moderate to heavier contested 3s
0.5 really open outside 2s
0 contested outside 2s
Then. sum scores for shots taken = shot selection enhancement above minimum (i.e. the worst shot, contested outside 2s).
Sum scores for shots taken / shots taken = shot selection enhancement efficiency.
these two measures will help you know who are getting you more better shots and who is getting them more often.
Analyze your wins and losses, especially at team factor level. Every player needs to know the 1-2 things they most need to do to help make the team better on the 4 Factors. If they can’t / won’t do it better in next few games, get somebody else to the extent possible unless they are otherwise absolutely essential given your roster. Analyze your 3-5-10 most used linueps at the 4 Factors level as well.
For the college game I would especially look at which players and lineups are on the court when the team goes on an 8+ point run for or against. Know which PG really plays best with the other 4 starters and each individual player.
BECAUSE rebounding is probably the second most important part of the game after shooting/scoring make sure your team is looking at it carefully and correctly, that using rebound rates and looking at rebounding opportunities and capture rates and such and not just raw game totals. If one really wants to learn as much as possible and improve as much as possible I'd look at player and team rebounding performance by part of the shot clock, stage of game, with lead/ behind, home and away, in bonus/not, in man to man or zone or against the same, etc. Not just to collect and stare at numbers but to understand what really happened and what needs to be improved on the court to get better.
BECAUSE turnovers are the remaining major element of the game don't settle for a raw count. Know where they occur on the court, on theclock, what type of error they are and who contributed to the turnover. That is practical intelligent analytics aimed at more effective action.
If you want to look at individual player performance in a single metric, look at using an advanced statistical plus minus like DSMOK1's.
http://godismyjudgeok.com/DStats/aspm-and-vorp/
If there is the talent or money to do so it might be interesting to do a RAPM run exclusively for a conference (in conference games) and mainly focused on finding estimates for the starters. If the skill was available I'd take the analysis down to 4 Factor level. Sample sizes are small so maybe multi-season would help a bit.
If you have access to SportVU data on layer movement work with it until it gives you specific information that can help improve on the court action rather than reject it as esoteric.
Example (extended consideration, skip if not wanting all of it): Instead of scoffing as Stan Van Gundy did about the lack of usefulness of knowing that Paul George ran the most miles this season, drill down into the detail. SportsVU may only report total miles and average speed to the public but it has detail on every segment of time and how far and fast every player moved. If you divided player and team data that occurred in .25 mile or .10 mile increments one could look to see if player or team stats for eFG%, 3pt FG%, inside FG%, FGA, 3pt FGA, inside FGA, FTA, assists, overall usage, steals, blocks, reb%, personal fouls, etc. per possession or minute changed as total miles increased increment by increment thru a game or thru a stint on the court. You could look to see if performance for any of these declined if distance covered divided by time exceeded certain levels (.0x mile / Y minutes) and judge when the right time is to substitute each player and perhaps what the upper limit should be on team pace when that player is on the court. If you know how far a player has run and how he has performed, one could then look at tape as see how much of his movement was appropriate and where specifically it could be best streamlined or made more exerting.
If you found out that Russell Westbrook for the season or the last month drove inside twice as much in his first mile of running as the second that would seem useful to learn and act on. Same for if you found out Durant increased his 3pt FGA rate by 50% if the distance travel in the prior 3-5 minutes was say 25% above average. If Perkins’ block rate went down 1/3 rd after the first half mile run or if .X miles / Y minutes exceeded Z, that also seems worth knowing. And so on in dozens of ways. If you knew Ibaka’s PF rate went up 40% when distance for a half of a game was 20% above normal that could affect when you should sit him, especially when near “foul trouble”. If your team depends on steals it might be worth knowing how successful your team is at getting steals when the opponent team or just the PG is running at above average, average and below average speed or the rate of travel in the last few minutes was high, medium or low.
If you have adopted a high pace, it would seem particularly important to know how your current players perform at high pace vs. average and low and how well available players do as well for these splits. Perhaps there is a better fit player for your style of play.
The analysis could extend to how player and team performance varies based on schedule variances (high number of games in a week, back to backs, month to month). Small changes in rotation management might be useful. The analysis gives information that can be considered. Judgment still must be applied. But a better informed decision-maker might do better. If they are open and able to handle the information. Instead of just looking for a way to dismiss and belittle it. I would assume that Mr. Van Gundy and most coaches try to gauge how players perform as they log miles and the speed of those miles vary. But they are doing that in a mini-window as the game progresses or the tape watching progress along with many other evaluations. Who is more likely to have accurate information and good insights and appropriate adjustments, the multi-tasking coach sometime and vaguely monitoring this or the data system and the analysts that have access to thousands or millions of observations and have the ability to study it carefully and find patterns to discuss with the coach? I know my answer but the kicker is that this is not necessarily an either or choice. Do it both ways and compare and discuss and study some more. Both ways.
If you had a long-range database about miles traveled and their intensity for the league, one could look at the possible correlation between them and injuries (to the leg or anywhere) and the possible impact of them to the “aging curve”. In general and for particular types of players and particular styles of play.
If you divide the distance traveled into the horizontal and vertical axises of the court one could see how far a player travels on each on average on a fast break (offense or defense) and within the first 2-4 seconds (vs. eventually). You could see who is running straight, fast, in lanes, who stays active and who quits, etc.
The sportsVU data could yield plenty of information in other areas to aid intelligent and effective action if you work it hard.
I know this rambled and got real specific in places but that is what came to mind in response to the question posed. I might revise and extend further later.
Re: NCAA Div I uses for Analytics--new to group
Too much obsession with SportVu, when Vantage is just better. 
Also, I'm not even sure if SportVu provides anything to college teams. They don't have the cameras so how are college teams supposed to use them?
Also, I'm not even sure if SportVu provides anything to college teams. They don't have the cameras so how are college teams supposed to use them?
Re: NCAA Div I uses for Analytics--new to group
If college has willing to pay somebody would probably be willing to provide the service, at least for home games.
Re: NCAA Div I uses for Analytics--new to group
I suppose. Or you could just pay for the Vantage stats which would provide for both home and road games + superior stats. A few examples being: on contested shots, SportVu doesn't track whether your hand is up or not. Big difference. Another difference: on contested rebounds, SportVu tracks within 5 feet, Vantage tracks within 1 foot. Other differences include the ability to track moves- like shot fakes, jab steps, etc. (maybe SportVu tracks this but haven't heard about it) Good article on that here: http://blog.cacvantage.com/2013/02/alwa ... ckets.html
Additionally, rather than having to worry about 80% accuracy on on-ball screens, you get 100% accuracy (well near 100%, think NBA audit found they were 99.7% accurate) because Vantage tracks on-ball screens.
You get more stats and don't have to worry about tracking only road games.
Additionally, rather than having to worry about 80% accuracy on on-ball screens, you get 100% accuracy (well near 100%, think NBA audit found they were 99.7% accurate) because Vantage tracks on-ball screens.
You get more stats and don't have to worry about tracking only road games.
Re: NCAA Div I uses for Analytics--new to group
There are a few college teams with SportVU and many more to come. The most notable current teams are Duke and Louisville.knarsu3 wrote: Also, I'm not even sure if SportVu provides anything to college teams. They don't have the cameras so how are college teams supposed to use them?