Advanced statistical analysis is becoming <a href="http://grantland.com/features/expected-value-possession-nba-analytics/" target="_blank">increasingly prevalent</a> in basketball.  Everything from the <a href="http://www.basketball-reference.com/about/factors.html" target="_blank">four factors</a> to lineup frequency is readily available.  Not so easy to find at the college level however, are shot charts and shot location data.  Some sites have this information on a game-to-game basis for teams in the Top 25 but, in general, this data isn't easy to track down.  My interest in these numbers, along with my love of basketball, has led me to endeavor to track this data for VCU.  If you'd like a peek behind the curtain at this process, continue on.  Otherwise, jump ahead to my amateur analysis of VCU's shooting performance through the out of conference schedule.
Hello, My Name Is Jeff
I am not a statistician.  In fact, I’m not particularly adept at working with numbers in any real capacity at all.  No, my finest achievements in any arena associated with mathematics are being first in my class to learn the multiplication table, and netting a minor in economics.  Everything in between, and everything since, has been a sort of Bermuda Triangle - my noodle is the place numbers go to die.
In the interest of full disclosure, I earned my econ minor the same way one earns a t-shirt for mowing down some über burger at a tacky diner.  It was a pathetic and disgusting display that, upon reflection, draws out more embarrassment than pride.
Ah, but therein lies one of the happy little accidents of sport.  Although I’m more likely to comprehend <i><a href="http://flavorwire.com/209104/10-of-literatures-most-notoriously-incomprehensible-classics" target="_blank">Naked Lunch </a> </i>than <i><a href="http://www.amazon.com/Naked-Statistics-Stripping-Dread-Data/dp/1480590185" target="_blank">Naked Statistics</a>,</i> there's something about box scores that transforms all those pesky numerals into a remarkably palatable snapshot of data, free from theory and abstraction.
Don’t get me wrong, my beautiful mind is still less <a href="https://www.youtube.com/watch?v=YWwAOutgWBQ" target="_blank">John Forbes Nash at a window</a>, and more <a href="https://www.youtube.com/watch?v=NkwJ-g0iJ6w" target="_blank">Johnny Nash when the rain is gone</a>.  I can see more clearly (admittedly I'm more interested) when the math is embedded in the context of sport.
All of this to deliver a word of caution: this shot chart project is more of an art than a science.
Process
If you’re at all familiar with the astonishing level of detail <a href="http://www.stats.com/sportvu/basketball_data.asp" target="_blank">SportVu</a> offers, you’ll be the first to know - this ain’t it.
As I watch the game, I’m simply charting where a shot is taken on the court.  I do this for both the team as a whole, as well as each player individually.  Unlike the NBA (with SportVU technology), at the college level, I believe this is how all shot chart data originates - someone watches and charts.
When humans perform tasks like this by hand, mistakes will be made.
Check out Melvin Johnson’s shot chart, as presented by cbssports.com, from VCU’s game against Villanova.
<a href="http://www.vcuramnation.com/wp-content/uploads/2015/01/Mel_BadChart.jpg"><img class="alignnone wp-image-13650 size-large" src="http://www.vcuramnation.com/wp-content/uploads/2015/01/Mel_BadChart-1024x445.jpg" alt="Mel_BadChart" width="618" height="269" /></a>
There are a couple of errors here, but just compare the box score three-point numbers against the chart.  They don’t add up.  The problem specifically is the left corner missed three-point attempt.  It never happened.  I can’t explain why the chart shows Melvin took that shot, but I can tell you that he wasn’t even in the game when the data implies he did.
The point here isn’t to deride cbssports.com, but rather to show that everyone makes mistakes when recording this kind of data, and I can’t promise that anything I do is without error.
So, I drop a little hexagon where each shot is taken, but how accurate is that?
In a <a href="http://kenpom.com/blog/index.php/weblog/entry/shot_chart_champions" target="_blank">2012 blog post</a>, Ken Pomeroy examined shot location data (from cbssports.com) and proceeded under the assumption that the margin for error was roughly 3-4’.  I’d like to believe that my work is comparable in that regard.
At the conclusion of every game, I overlay a graphic that breaks the court into 17 different sections.  I then record makes and misses for every player at each location.  The resulting chart is theoretically easier to digest as it helps declutter, and organizes a number of smaller data points.
<a href="http://www.vcuramnation.com/wp-content/uploads/2015/01/ETSUBins_Final.jpg"><img class="alignnone wp-image-13651 size-medium" src="http://www.vcuramnation.com/wp-content/uploads/2015/01/ETSUBins_Final-300x285.jpg" alt="ETSUBins_Final" width="300" height="285" /></a>
Since other shot location data isn’t broken into these same bins, I can’t compare VCU’s conversion rate at a specific location to anything else.  What I can do is compare how they shoot, at a certain spot, to the national average for the value of the field goal attempt.  In other words, the left corner three-point attempt is compared against the national average for <i>all</i> three-point attempts - the right side elbow jumper is compared against the national average for <i>all</i> two-point attempts.
So there it is.  It isn’t perfect, but remember - it’s more art than science.  My hope is that the charts are a nice supplement to the game and if they, in any way, help us understand the sport we love, I’ll consider it a success.
VCU - OOC Shot Chart  
VCU begins Atlantic 10 conference play tomorrow, but before they do I thought it’d be fun to take a look back at how the Rams did offensively during OOC play.
Here is the VCU team shot chart for games through December 30, 2014.  The data at each location shows FGM/FGA, FG%, and distribution percentage.  Essentially it tells us at what volume did the Rams shoot at any given spot, and the rate of conversion.  I did not include information for locations that account for less than 3% of all field goal attempts.
<a href="http://www.vcuramnation.com/wp-content/uploads/2015/01/OOC_Final_2.jpg"><img class="alignnone wp-image-13652 size-large" src="http://www.vcuramnation.com/wp-content/uploads/2015/01/OOC_Final_2-1024x973.jpg" alt="OOC_Final_2" width="618" height="587" /></a>
The first thing that jumps out to me is actually captured with the graph in the bottom left.
<a href="http://www.vcuramnation.com/wp-content/uploads/2015/01/FGA_Dist.png"><img class="alignnone wp-image-13653 size-large" src="http://www.vcuramnation.com/wp-content/uploads/2015/01/FGA_Dist-1024x500.png" alt="FGA_Dist" width="618" height="302" /></a>
About 40% of VCU’s attempts are at the rim, and about 40% are from beyond the arc, with the other 20% being pretty evenly split between mid-range attempts and long twos.
I love this.  The Rams are evenly distributing field goal attempts between shots that are statistically the most efficient (those at the rim) and attempts that have the highest value (3s).
Speaking of even distribution, check out the percentage of attempts around the perimeter.  VCU gets three-pointers up at a relatively even rate at every spot around the arc - just slightly favoring the top of the key and the deep left corner.
At 48.2%, the conversion rate in that deep left corner is outstanding, whereas the 25.4% at the top of the key (where the Rams take the most threes) is significantly under the national three-point average.
The other sore spot from beyond the arc for the Rams is the left wing.  VCU’s two left side dominant three-point shooters, Jordan Burgess and Terry Larrier, have combined to account for more than half of the attempts (19) at that location, but have made just 3.
Burgess is a slightly above average three-point shooter and is responsible for all 3 makes at that spot.  Larrier, on the other hand, is a freshman making the adjustment to the college game right now, and has yet to find a consistency from range.  He is fantastic in the deep corners, shooting a combined 46.2%, but struggling everywhere else.  Outside of those corners, Larrier is 2 of 18.
As a rule of thumb, long two-point attempts are bad news.  They are just as valuable as attempts at the rim, but share a likelihood of conversion closer to a three-point shot.  Needless to say, they aren’t efficient attempts.
If you take that into account, the field goal percentages shown at that range aren’t awful - although they are below the national average for two-point attempts.  But keep an eye on shots of this kind.  For instance, are they wide open looks in the soft part of a zone, or are they contested efforts to beat the shot clock?
Overall, I’d maybe like to see a slight uptick in the rate of conversion inside (a combined 56.4%), but VCU is shooting threes well and minimizing inefficient looks.
With the distribution of field goal attempts balanced the way it is, I think the recipe is right.  If and when more shots start to fall, it could spell danger for the rest of the conference as VCU begins A10 play.
Hello, My Name Is Jeff
I am not a statistician.  In fact, I’m not particularly adept at working with numbers in any real capacity at all.  No, my finest achievements in any arena associated with mathematics are being first in my class to learn the multiplication table, and netting a minor in economics.  Everything in between, and everything since, has been a sort of Bermuda Triangle - my noodle is the place numbers go to die.
In the interest of full disclosure, I earned my econ minor the same way one earns a t-shirt for mowing down some über burger at a tacky diner.  It was a pathetic and disgusting display that, upon reflection, draws out more embarrassment than pride.
Ah, but therein lies one of the happy little accidents of sport.  Although I’m more likely to comprehend <i><a href="http://flavorwire.com/209104/10-of-literatures-most-notoriously-incomprehensible-classics" target="_blank">Naked Lunch </a> </i>than <i><a href="http://www.amazon.com/Naked-Statistics-Stripping-Dread-Data/dp/1480590185" target="_blank">Naked Statistics</a>,</i> there's something about box scores that transforms all those pesky numerals into a remarkably palatable snapshot of data, free from theory and abstraction.
Don’t get me wrong, my beautiful mind is still less <a href="https://www.youtube.com/watch?v=YWwAOutgWBQ" target="_blank">John Forbes Nash at a window</a>, and more <a href="https://www.youtube.com/watch?v=NkwJ-g0iJ6w" target="_blank">Johnny Nash when the rain is gone</a>.  I can see more clearly (admittedly I'm more interested) when the math is embedded in the context of sport.
All of this to deliver a word of caution: this shot chart project is more of an art than a science.
Process
If you’re at all familiar with the astonishing level of detail <a href="http://www.stats.com/sportvu/basketball_data.asp" target="_blank">SportVu</a> offers, you’ll be the first to know - this ain’t it.
As I watch the game, I’m simply charting where a shot is taken on the court.  I do this for both the team as a whole, as well as each player individually.  Unlike the NBA (with SportVU technology), at the college level, I believe this is how all shot chart data originates - someone watches and charts.
When humans perform tasks like this by hand, mistakes will be made.
Check out Melvin Johnson’s shot chart, as presented by cbssports.com, from VCU’s game against Villanova.
<a href="http://www.vcuramnation.com/wp-content/uploads/2015/01/Mel_BadChart.jpg"><img class="alignnone wp-image-13650 size-large" src="http://www.vcuramnation.com/wp-content/uploads/2015/01/Mel_BadChart-1024x445.jpg" alt="Mel_BadChart" width="618" height="269" /></a>
There are a couple of errors here, but just compare the box score three-point numbers against the chart.  They don’t add up.  The problem specifically is the left corner missed three-point attempt.  It never happened.  I can’t explain why the chart shows Melvin took that shot, but I can tell you that he wasn’t even in the game when the data implies he did.
The point here isn’t to deride cbssports.com, but rather to show that everyone makes mistakes when recording this kind of data, and I can’t promise that anything I do is without error.
So, I drop a little hexagon where each shot is taken, but how accurate is that?
In a <a href="http://kenpom.com/blog/index.php/weblog/entry/shot_chart_champions" target="_blank">2012 blog post</a>, Ken Pomeroy examined shot location data (from cbssports.com) and proceeded under the assumption that the margin for error was roughly 3-4’.  I’d like to believe that my work is comparable in that regard.
At the conclusion of every game, I overlay a graphic that breaks the court into 17 different sections.  I then record makes and misses for every player at each location.  The resulting chart is theoretically easier to digest as it helps declutter, and organizes a number of smaller data points.
<a href="http://www.vcuramnation.com/wp-content/uploads/2015/01/ETSUBins_Final.jpg"><img class="alignnone wp-image-13651 size-medium" src="http://www.vcuramnation.com/wp-content/uploads/2015/01/ETSUBins_Final-300x285.jpg" alt="ETSUBins_Final" width="300" height="285" /></a>
Since other shot location data isn’t broken into these same bins, I can’t compare VCU’s conversion rate at a specific location to anything else.  What I can do is compare how they shoot, at a certain spot, to the national average for the value of the field goal attempt.  In other words, the left corner three-point attempt is compared against the national average for <i>all</i> three-point attempts - the right side elbow jumper is compared against the national average for <i>all</i> two-point attempts.
So there it is.  It isn’t perfect, but remember - it’s more art than science.  My hope is that the charts are a nice supplement to the game and if they, in any way, help us understand the sport we love, I’ll consider it a success.
VCU - OOC Shot Chart  
VCU begins Atlantic 10 conference play tomorrow, but before they do I thought it’d be fun to take a look back at how the Rams did offensively during OOC play.
Here is the VCU team shot chart for games through December 30, 2014.  The data at each location shows FGM/FGA, FG%, and distribution percentage.  Essentially it tells us at what volume did the Rams shoot at any given spot, and the rate of conversion.  I did not include information for locations that account for less than 3% of all field goal attempts.
<a href="http://www.vcuramnation.com/wp-content/uploads/2015/01/OOC_Final_2.jpg"><img class="alignnone wp-image-13652 size-large" src="http://www.vcuramnation.com/wp-content/uploads/2015/01/OOC_Final_2-1024x973.jpg" alt="OOC_Final_2" width="618" height="587" /></a>
The first thing that jumps out to me is actually captured with the graph in the bottom left.
<a href="http://www.vcuramnation.com/wp-content/uploads/2015/01/FGA_Dist.png"><img class="alignnone wp-image-13653 size-large" src="http://www.vcuramnation.com/wp-content/uploads/2015/01/FGA_Dist-1024x500.png" alt="FGA_Dist" width="618" height="302" /></a>
About 40% of VCU’s attempts are at the rim, and about 40% are from beyond the arc, with the other 20% being pretty evenly split between mid-range attempts and long twos.
I love this.  The Rams are evenly distributing field goal attempts between shots that are statistically the most efficient (those at the rim) and attempts that have the highest value (3s).
Speaking of even distribution, check out the percentage of attempts around the perimeter.  VCU gets three-pointers up at a relatively even rate at every spot around the arc - just slightly favoring the top of the key and the deep left corner.
At 48.2%, the conversion rate in that deep left corner is outstanding, whereas the 25.4% at the top of the key (where the Rams take the most threes) is significantly under the national three-point average.
The other sore spot from beyond the arc for the Rams is the left wing.  VCU’s two left side dominant three-point shooters, Jordan Burgess and Terry Larrier, have combined to account for more than half of the attempts (19) at that location, but have made just 3.
Burgess is a slightly above average three-point shooter and is responsible for all 3 makes at that spot.  Larrier, on the other hand, is a freshman making the adjustment to the college game right now, and has yet to find a consistency from range.  He is fantastic in the deep corners, shooting a combined 46.2%, but struggling everywhere else.  Outside of those corners, Larrier is 2 of 18.
As a rule of thumb, long two-point attempts are bad news.  They are just as valuable as attempts at the rim, but share a likelihood of conversion closer to a three-point shot.  Needless to say, they aren’t efficient attempts.
If you take that into account, the field goal percentages shown at that range aren’t awful - although they are below the national average for two-point attempts.  But keep an eye on shots of this kind.  For instance, are they wide open looks in the soft part of a zone, or are they contested efforts to beat the shot clock?
Overall, I’d maybe like to see a slight uptick in the rate of conversion inside (a combined 56.4%), but VCU is shooting threes well and minimizing inefficient looks.
With the distribution of field goal attempts balanced the way it is, I think the recipe is right.  If and when more shots start to fall, it could spell danger for the rest of the conference as VCU begins A10 play.
