Quote:

---

Kelly's lightbulb moment led him to professor Shlomo Argamon and students Denis Bajic and Larry Layne in Argamon's master of data science program who dedicated their summer practicum to basketball analytics.

The result? The Scarlet Hawks have a 17-5 record.

It doesn't take a mathematician to calculate that's a major improvement over last season's 4-21 mark or the 2-23 record they posted in 2014-15, Kelly's first season. The season before that, IIT went 0-25.

---

At Illinois Tech, data analytics play big part in basketball team's success

Does anyone have knowledge of any of our sports programs using analytics involving students or outsourcing?

My son is pursuing a Data Science career and will be starting his 2nd year at A&M. Are there any projects he can help with?

OpenMouthGaze.jpg

Yes

GE said:

---

Yes

---

Can you expand on this?

Are there any students involved? Graduate assistants? Is it outsourced?

This is where we make the final four and someone does an ESPN story talking about how despite the protests of fans, A&M's proprietary algorithms actually explain Kennedy's bizarre substitution patterns.

Between roster turnover, quality of competition and lack of games I can't imagine there's enough useful data to help much in college basketball.

Tobias Funke said:

---

How'd you get your job working in sports analytics?

---

He is actually more interested in bioinformatics than sports.
But I was curious to hear how we were using analytics and if we could make better use of it.

The football team is, at least as far as I know, light years ahead of any of our other teams in using any sort of advanced data for anything and they mostly use it for analyzing how much players are working in practice to try and prevent injuries and keep guys as fresh as possible.

Yes, I remember reading an article (or was that "The Pulse" video) about that monitoring.
I would guess that real time analysis during games would be limited to the coaching and support staffs. If not, you could have 20 students breaking down the plays and formations providing input.

My understanding is that the head of analytics of the Houston Rockets (ie maybe the most analytics focused team in the NBA) is the father of Braden Mann, the kicker on the football team.

Doesn't have anything to do with A&M using analytics, but just a random semi relevant interesting fact (at least I think I remember it being true) for the day...

As far as I can tell there are really two major takeaway from basketball analytics in recent years, none of which we have been particularly good at incorporating.

1. Shoot 3's and layups/ dunks and nothing in between.
2. Give your good players some rest or they wear down and are less effective at the end of the year.

There's some other things at the NBA level you can use that I don't think would work at the college level. Lineup analysis and that sort of thing.

By the time you really neutralize the key components in lineup variance, you'd be working with such a small dataset that it would be hard to to actually make a strong case for anything that the basic stats didn't already tell you (play your good players as much as you can).

And a lot of the problem with data-driven strategy when it comes to a game like basketball is you risk interrupting the flow of your players. So really all you get is very broad, almost "common sense" strategy like mdancy03 mentioned -- shots farther away from the basket are harder to make, so if you are going to shoot something that isn't close to the hoop it should be a 3-pointer.

It's naturally most useful in a stat-heavy, almost "turn-based" game like baseball where we are learning such fun things as steals really aren't worth it, the order of your lineup really doesn't matter, and you can have a sub-200 batting average if you take a lot of walks and hit a lot of homeruns.

Yeah, even taking out the lack of games between good teams and how many garbage teams out there would foul up the data, a lot of advanced stat stuff works because most of the measurable events in baseball are independent of one another and there are a lot of them.

Neither of those are true in basketball.

Yup... and really the most likely thing we'd actually get out of more data-driven basketball strategy is a lot more unintentional intentional fouling or even worse -- mediocre teams realizing that once they are down double-digits in a game (at any point) that their win probability compared to stamina saved for the next night's game is at a point that they pull all of their starters by the time the 2Q rolls around because statistically it makes sense to just go ahead and rest them for a winnable game.

It may take more game video breakdown to get the data, but it seems like it would be useful to know your opponent's key player and coaching tendencies. Coaches would be the main beneficiary, but also the defenders of the key players.

Yeah, but that isn't really advanced analytics, it's just having data that we as fans don't have access to. "The point guard goes right 80% of the time" is basic analysis and I would assume that we're already doing that type of stuff.

Advanced stats (hypothetically) would be able to tell you something more like "DJ Hogg shoots 10% better when Tonny Trocha is on the floor." And then you could go to the tape and see why that is and maybe work it into your strategy or decide it's just an anomaly.

But that's what we're saying is there's nowhere near enough data in college basketball to know with any reliability if something like that is really true or not.

There's not even enough good data in college basketball for the basic stats a lot of the time.

Like here's a basic stat question we don't have enough data on: "Is DJ Hogg a good three point shooter?"

People who say no can point to his career percentage of 34.7% and say he's average, others would bring up that last year he was shooting at 40% prior to his injury which is definitely good, but you could also debate whether basically two performances (St. Francis and South Carolina) throw off the rest of the numbers. Do you let a guy just let it fly when most nights he shoots under 30% but once about every seven games he'll go nuts?

You'd have some valid arguments on all sides, and that's a very basic statistic. But it speaks to the quality of competition problem, lack of individual event problem and small overall sample size problem.

Quote:

---

Advanced stats (hypothetically) would be able to tell you something more like "DJ Hogg shoots 10% better when Tonny Trocha is on the floor." And then you could go to the tape and see why that is and maybe work it into your strategy or decide it's just an anomaly.

---

With something like that, by the time conference play rolls around, for instance, and of course, more so when the post-season comes, there is perhaps enough there so that it's worth acting on in order to see if a pattern is confirmed.

At least you have something. I think that's part of the idea. In other words, at a minimum, it provides a hypothesis, of sorts, for which you test and put under scrutiny. You'd try it in practice, you'd focus more on it in the tape room. Maybe you try it in a game.

And besides, statistics are funny. sample sizes smaller than your intuition would suggest often are statistically significant.

Quote:

---

But it speaks to the quality of competition problem, lack of individual event problem and small overall sample size problem.

---

With 120 possessions a game, its not outside the realm of statistics to draw outlines to create isolated data sets, create outputs, and detect patterns in those outputs. You'd do that multiple times and detect patterns within your patterns. Its multi-layered, but it seems doable to me.

Isolate Trocha and Hogg playing together in a game, isolate them not playing together in a game, do this over 10 games, is there a pattern within individual games? Over the aggregate 10? Then add another layer with another variable, the pattern still hold? Add in more layers. The pattern still hold? Etc.

I just don't think there's enough reliable data for it to be a valuable resource in college basketball at this point, but I'd readily admit I'm wrong if there's a team out there using it and having success with it at the major college level.

For what it's worth I do think we usually tend to scout our opponents really well. I don't know how much of that is data-driven, but it could be an indicator that we do some of that. We just don't usually adjust to their adjustments particularly well in my opinion.

You can apparently coordinate time stamps with shot locations, misses/makes, player substitutions, rebounds, steals, fouls and do a lot of advanced analysis that will reveal patterns just from what is available to fans:

KY vs. A&M 3/4/16 Stats

Overhead video can be converted into useful data by identifying player positions at the start of the play and tracking positions at set intervals. This can probably automated with software available today.

It's not the creation of the data that's the issue, it's the value of it. But if your son can DISRUPT college basketball, tell him to go for it.

I imagine any sort of college basketball revolution is going to start at either a mid-major school or a school that completely sucks at basketball.

I always thought the one big takeaway from NBA analytics is that corner 3s are a good thing.

I don't think it takes a statistics degree to know that a shot worth the same amount of points from a closer distance is a good shot.

Long time listener... I love this thread. My son will be a freshman in college in the fall, and interested in statistics and data analysis, and coaching, and plays basketball - so this has been a great topic.

But... It it so intuitive that a corner 3 is better? I would think that the lack of depth perception available from the corner would even out the percentages. Do the stats show that he corner 3 is the best shot?

JDindallas said:

---

Long time listener... I love this thread. My son will be a freshman in college in the fall, and interested in statistics and data analysis, and coaching, and plays basketball - so this has been a great topic.

But... It it so intuitive that a corner 3 is better? I would think that the lack of depth perception available from the corner would even out the percentages. Do the stats show that he corner 3 is the best shot?

---

Couple years old, but:

http://www.basketballanalyticsbook.com/2014/03/29/the-importance-of-the-corner-3/

mdanyc03 said:

---

As far as I can tell there are really two major takeaway from basketball analytics in recent years, none of which we have been particularly good at incorporating.

1. Shoot 3's and layups/ dunks and nothing in between.
2. Give your good players some rest or they wear down and are less effective at the end of the year.

---

1-b: Defend against 3s and layup/dunks. Don't do the first part.

JDindallas said:

---

Long time listener... I love this thread. My son will be a freshman in college in the fall, and interested in statistics and data analysis, and coaching, and plays basketball - so this has been a great topic.

But... It it so intuitive that a corner 3 is better? I would think that the lack of depth perception available from the corner would even out the percentages. Do the stats show that he corner 3 is the best shot?

---

As mentioned in the article, but I think the spacing component is the biggest asset of being able to hit corner 3s consistently. Creates more 1-on-1 matchups in space.

Its partly the spacing, threes in the corner tend to be wide open, but also partly because in the NBA, which the article is about, it's a closer three than anywhere else on the court.

A foot makes quite a bit of difference.

Thanks for the replies and the link to the article. I always thought that the fact that there was no backboard behind the basket for depth perception on a corner 3 would have more effect on percentages, but I guess that's only for me! Also, you obviously can't bank one in from the corner like you can from out front, but I guess there are few enough of those to make it statistically irrelevant. (?)

Yeah I imagine somewhere in the neighborhood of 95% of three point shots don't hit anything other than the net or rim.

Even if the 3 point line is closer at the corners, it is a difficult shot to make because it's a set shot not necessarily a jump shot. Muscle memory is much more important than anything else with this shot because of the angle.

There's only 3 feet between the 3 point shot line and the out-of-bounce line. I'd be curious to see the number of turnovers that happen because a guy accidentally steps out-of-bounce while trying to put up a 3 point shot.

While actually trying to put up a shot I'd say it's extremely rare. Most sideline out-of-bounds are because someone overthrew a pass and they have to jump or move to catch it, or they're not watching their feet while handling the ball. I'd think it's pretty rare for someone to step out of bounds in the act of shooting.

But it isn't a difficult shot. Other than an open layup or dunk it's the most efficient shot on the floor.

Also, it's not closer in college, just in the NBA. Just to make sure nobody is confused.