Basketball Analytics

Hi everyone, my team and I are currently working on a project that aims to provide personal performance analytics for basketball players. We intend to focus on player development and work with athletes at the non-pro level. We essentially gather regular box score data, analyze the data into players’ strengths and weaknesses, and in turn provide in-game tips / training recommendations. Since I’m a huge fan of our Fighting Illini, I decided to use what we’ve created to evaluate our beloved Fighting Illini players.

The strengths and weaknesses are assessed via hoopr’s Player Performance Index (PPI) algorithms. Currently, there are 4 major PPIs*, (1) scoring, (2) defense, (3) ball handling, and (4) basketball IQ. I’ve created a Google Sheet to show how our players have done in the 2015-2016 and the 2016-2017 seasons.

The first tab (Illinois 2015-2016) shows a game-by-game PPI on the left and a running average PPI on the right for all of the 34 games in the season. At the bottom of the PPI numbers, you’ll see 4 graphs showing players’ progression in the different PPIs. I have omitted some players for their minimal playing time. The second tab details the same things, but for the 2016-2017 season. The last tab provides a comparison of the same players in the 2 seasons.

I appreciate your taking the time to read all this, please feel free to offer any comments/recommendations/criticism on what we’re doing and/or the result for our players. GO ILLINI!!

*Each PPI is placed on a scale of 0 to 100, and is determined by multiplying basic (e.g., points per game, rebounds per game), advanced (e.g., true shooting percentage, defensive rebound %), ranking (e.g., number of games as the top scorer), and other misc (e.g., number of games with more than 10 points) stats with their respective weights as some stats are more telling than others.

Really cool! I like the visualizations. Kind of funny advice to Steph Curry though

Haha I appreciate it!

We hope to provide this platform to the non-pro basketball players so that they can have better resources in developing their games. Most of the kids today learn by playing a ton of games, but playing a ton of games without knowing what/how to improve isn't really going to improve your game (AAU, hint, hint), and that's where we want to help!

How do you get your data? Is it something that players will have to enter in manually, or can you get stats and such from some outside source?

I imagine that NBA and bigger college programs would have the data easily accessible, but I have no idea where you would find consistent stats on HS players.

Yea there is some data that we cannot obtain, but luckily some high schools enter their stats to sites like MaxPreps, and other more regionally-based websites. So we either grab them from sites like MaxPreps, or on our platform, we also allow for coaches to enter their stats directly into our website so that coaches can keep their stats somewhere online and players can see the PPIs and in turn receive any in-game tips / recommendations.

TIFWLIW, but boiling a player down to a number in a category doesn't tell me anything.

PPG? I understand that.
Usage? Yup
Steals, assists, rebounds? Yup
Qualitative (subjective) analysis? Yup to that too. If someone sees tendencies on where a player likes to shoot from on the floor, or sees that he can't seem to recognize and pass out of a double-team, I understand that.

A weighted scoring category of 6? Not sure what that actually means or how that helps me understand the game/players.

What if you were a basketball player and based on these evaluations, you're able to see how you stack up against your team in these abilities (assuming the evaluations are somewhat reflective of your skills)? Would you be interested in looking at this?

In part, the purpose of creating these evaluations is so that we can offer these athletes in-game tips and training recommendations.

hoopr said:

What if you were a basketball player and based on these evaluations, you're able to see how you stack up against your team in these abilities (assuming the evaluations are somewhat reflective of your skills)? Would you be interested in looking at this?

In part, the purpose of creating these evaluations is so that we can offer these athletes in-game tips and training recommendations.

Click to expand...

It wouldn't interest me, because it doesn't translate into actionable skill development.

Calvin said:

It wouldn't interest me, because it doesn't translate into actionable skill development.

Click to expand...

Care to elaborate?

There's a ton of skill development videos out on the web that address player development or skill deficiencies, should you care to look around. You need very detailed metrics to tell you what to work on, or drills that reveal an underlying metric. Usually it's a coach that sees what needs work because they will notice in the absence of a metric. For example, if a PG loses the ball 50% more often when they go low through traffic with their left hand, a coach will come up with drills to work on it, and perhaps in the short term, try and change his decision making. A player that's slow to recognize a double team in the post will likewise get drills to work on, etc.. Drills will in turn reveal problems. Then there's the issue of whether the drill translates to game situations (one of my personal pet peeves). You need a coach who can simulate game conditions in drills, or the player winds up good at a drill, and not good when the lights are on. We've certainly had a few of those under previous coaches. Metrics like the one you mention might be good for predicting scores when a player is injured, but I don't see how it points a player to what he has to do in the gym.

Interesting. Thanks for the feedback.

Posted the previous comment without refreshing the page :tsk:

Agreed that there are a ton of resources out there for ball players to explore, but I think players have to first identify their weaknesses before they can figure out which drills to search for. They can either do so by realizing themselves or through their coaches. Even when the players know what drills (say, dribbling) to look for, there are many different difficulty levels of the same, or closely similar, drills. If they aren't able to figure out what's wrong with their game, for lack of playing time or otherwise, players either assume they'll get better by playing more games or rely on their coaches to point out mistakes. Unfortunately, coaches don't have the time to individually develop their players as they have an entire team to care for. They are more likely to identify the wrongs as a group, rather than on an individual basis.

In addition to providing drills to the players based on their strengths and weaknesses, we want to offer in-game tips as well. So if we notice a player shoots a lot of 3s but averages on the lower end of the spectrum for made 3s, then we may prompt a conversation with the player on our platform like "hey we notice you shoot a lot of 3s, where are you shooting most of your 3s from?" with the following answers "top of the key," "45 degree," and "corner 3s" and if the player picks the first two answers, we'll say something along the lines of "try to take your 3s in the corner because it is the most efficient location on the court."

Hi all, I wanted to update you on our progress since my original post. Since then, we have been able to plot the data you saw in the Google Sheet onto our platform for a better visualization. You can see the variations of graphs we have prepared for our players. As you can see in here, we're using Malcolm Hill as our example, but we have the same "player dashboard" for all of our players.

The data that was used to plot these (pretty but still a work in progress) graphs is presented in our league hosting feature here. We use league hosting feature as a way for league / tournament hosts to host their data online and we will use the data they input to provide evaluation to the players.

As always, we welcome any comments / questions / recommendations you may have. Thanks!!

What is your work developed in? R? Python? Are you sharing or working on github? I'm way into this stuff.

The evaluation algorithm we have today was originally developed in Excel because we weren't sure what data set to begin with, i.e., what stats we should use to evaluate a player's scoring ability. The algorithm currently works to a good degree of accuracy in terms of evaluation, after getting feedback from players and coaches from all 3 divisions, but we will be going into Python for fine-tuning the evaluations in the future.

P.S. We have opened up our platform for private beta, so if you are hosting a tournament in the near future, we are offering our platform for you to use for free. This way, you will be able to manage your teams, schedules, and stats online, and your players will get their individual player analytics when the stats are inputted. If you are interested in working with us, please feel free to PM me and I will get you started. Thanks!

hoopr said:

The evaluation algorithm we have today was originally developed in Excel because we weren't sure what data set to begin with, i.e., what stats we should use to evaluate a player's scoring ability. The algorithm currently works to a good degree of accuracy in terms of evaluation, after getting feedback from players and coaches from all 3 divisions, but we will be going into Python for fine-tuning the evaluations in the future.

P.S. We have opened up our platform for private beta, so if you are hosting a tournament in the near future, we are offering our platform for you to use for free. This way, you will be able to manage your teams, schedules, and stats online, and your players will get their individual player analytics when the stats are inputted. If you are interested in working with us, please feel free to PM me and I will get you started. Thanks!

Click to expand...

Man, I gotta say this looks like GIGO to me. Analylizing ballhandling/defense with box score data? What sort of box score data reflects basketball IQ? Seems like a fools errand. And the scoring analysis...he can look at the box score and figure that out.

Thanks for your comment. Totally get it. Again, we are offering the view with the graphs to the players, for their benefit. What you are seeing in terms of the spider graph and the trends for each category are summaries of the player's ability in that category. We are currently creating the in-depth analyses with actionable bullet points for each category, so as a player you would be able to see what contributes to the rise or fall in the trend.

For basketball IQ, we look at turnovers, fouls, free throw attempts, free throw %, etc. We do understand that box score data alone does NOT fully capture the stories on these categories and we are looking at more things to extrapolate from box score. That said, we did get positive response on these from current D1-D3 players at a tournament last weekend.

The problem is that you're conflating ability with production. These graphs don't tell a player anything other than what they would already know from reading a box score. All you're doing is recontextualizing them into a number which doesn't have any intrinsic meaning.

Thanks for your feedback. We understand the presentation is not the most effective at this point. Once we have our in-depth analysis developed, we'd love to have your feedback again!

Hi guys, we at hoopr have developed a new tool that allows you to do real-time stat tracking on your phone (here are some screenshots of how it's done). This is a great way to track your players', or kids', performance as you watch them play on the court, and begin building their basketball history.

As summer ball is underway, you can sign up now at www.hoopr.com, and track performance when you, your kids, or your players play.

As always, we welcome your feedback. Thank you!

[YOUTUBE]66ko_cWSHBU[/YOUTUBE]