972 - Enabling Better Basketball Analytics Using Computer Vision w/ Bo Shaffer (Playswise)

All right, how's it going, everyone?

Welcome to another episode of Forward Thinking Founders, where we talk to founders about
their companies, their visions for the future, and how the two collide.

Today, I'm very excited to be talking to Beau Schaefer, who's the founder, CEO of
Playswise.

Welcome to the show, how's it going?

I'm doing well, how are you?

I'm good.

I am, you know, just here doing a bunch of calls, but we have this podcast, which I always
look forward to whenever I have podcasts scheduled in my day, because I get to meet

someone new, interesting, working on working on cool stuff.

um So let's kind of just dive right into what you are working on.

What is Playswise?

Yeah, so placewise is using computer vision to get in-depth basketball analytics from
broadcast video.

So NBA has a ton of cameras in their facilities to get some of these insights.

And right now, we're working on a way to use computer vision, AI, machine learning, that
type of stuff to get the same analytics just from broadcast video.

So let's say, once this is built, why is first, is this already built?

Are you building it?

And two, once you're able to get this data, uh who would be interested in it?

Is me as a spectator, am I interested?

Is a coach interested in evaluating talent to maybe get someone from another team?

Who cares about this data?

Yeah, so I think starting off, I played basketball at Duke, so this is what triggered my
curiosity in it.

And I saw AI being used in other applications and thought, I wonder if that's happening.

Or I knew it wasn't happening in the basketball world.

um But essentially, my thought was, can you use something like ChatGBT and ask it for
stats or questions and be able to get an output that's actually accurate and good?

um And current answer is I'd say no, especially to the depth we want it now.

um

I think right now, uh since I'm on the side of basketball, I'd say as like a player, my
thought is like, how can coaches use it or people on staff?

So I think that'll be like the initial target.

However, as a player, like there are definitely times where I think I would want, you
know, to record myself practicing shooting or games and be able to get just data from it,

like even just how many shots I made, uh how many ah like dribbles I took or what

kind of tendencies I had and stuff like that.

Yeah, that's awesome.

I am curious, know, were you in, I guess, what came first, your interest in basketball or
your interest in tech?

And if it was basketball, then tech.

ah How did you, how did you kind of discover the tech world?

And if it's the opposite, like, how'd you first get into tech?

Yeah, that's a good question.

think I really stumbled upon this kind of accidentally.

I think I always knew from for a while, I knew I wanted to get into sports somehow.

I didn't know when or how that would be.

And then I think I've been always pretty interested in, uh you know, AI tech and it's been
booming, uh obviously, with chat, tpt and LLMs and BLMs even.

And I kind of just saw a lot of it being developed in like a broad focus for users.

like your everyday person for random things or and whatnot.

And then it kind of started to dabble into other verticals like legal stuff or uh
medicine, stuff like that.

And then I just kind of thought of, you know, this would be really cool on the basketball
side.

think uh having a lot of friends or like scouts, coaches, analysts, that type of thing,
hearing their daily workflow and me thinking, like you guys have to do all this manually,

like seems seems kind of crazy to me, considering all the advancements

with technology right now and so I think that's probably what kind of sparked it.

So I don't know if there's probably I guess my initial interest was basketball but I
didn't mean for it to be basketball when I found the tech side of it.

No, totally.

That makes sense.

And, you know, we were talking, you know, earlier before we were recording about about
Twitter, that's how we met, right?

We met on Twitter.

And, you know, as I was saying, I probably interview 80 to 90 % of the guests who come on
the podcast, I meet on the internet.

How has like the internet been useful for you as you've been, you know, you know, building
this out, you know, getting into tech, you know, using Twitter, are there other sources or

websites that you use?

Like, how's the internet helped you?

Yeah, uh I think the tech world's very open.

They open source a lot, which is cool.

So you can kind of go on there, test things, see if it could work for your application, or
see how someone else is using.

uh

their LLM or VLM or whatever it is or their models and seeing how they're using it and
seeing if you can kind of apply it to yours.

So it's definitely been uh inspiring in that sense.

And I honestly think if you aren't on Twitter or really keeping up with podcasts or using
something like Perplexity to get updates, think you're definitely falling behind.

um I think that is a battle, though.

Every time I go on Twitter, I feel like there's an update in some sort of new model or new
this, new that, new capabilities.

um But yeah, I think, yeah.

Yeah, sorry.

It's just, it's interesting because the standard advice is focus on your customers, you
know, turn all that stuff off and just, you know, talk to your users, build product,

right?

But in this age of AI and you know, look, there's been AI for decades, right?

But we're just in a moment.

We're in a moment where it's progressing at a very rapid rate.

I think you do have to be online.

Like I think you do have to be on Twitter, know what's dropping, just so you're in the
know and

Maybe if you're using a model today or using a tech technology today, in a year it could
be completely irrelevant.

And you need to know that and be on the ball.

But as you said, it's a balance.

You don't wanna switch models every five seconds or you don't wanna switch technologies
every day or every week, because that's just gonna be a pain in the butt.

ah But it is the battle.

How do you think about that?

Like you're building in this space.

How do you get, do you get distracted or do you let yourself get distracted by these new
models?

Like how do you handle that?

that's a good question.

think I tend to, if I get distracted, it's like, okay, I look at the post, I might go to
the link or whatever and check it out for a second, but I don't immediately switch.

uh I'd have to have a lot of conviction that this model's better, this is why we're gonna
use it.

ah But right now, I'm working on fine tuning our current models and more customizing them
to our needs.

A lot of the stuff on Twitter, from what I've found, is if they're open sourcing, it's
pretty,

It's too like a broader demographic, I'd say.

um But the biggest thing is, uh if I do see something new, I'll, I have a couple advisors
I talk to that are experts in like computer vision or machine learning that I trust, and I

know they've done a lot of good work, so I'll kind of consult with them and get some
advice if I do.

Or I'll send it to them and say, did you see this?

Check it out, and they'll kind of give me some feedback.

Yeah, that makes sense.

It does leave me to wonder, like, how are you spending your time right now?

It sounds like you're fine tuning your models.

Do you just be taking a lot of shots?

Are you dribbling?

Are you like giving a data to train?

And if it's not that specifically like how are you training without giving away your
special sauce?

ah I think yeah, right now a lot of my time is coding slash getting these models better
and more accurate.

um It's funny like on one hand, I think it's like the chicken versus the egg battle where,
okay, I need to make these models better and more accurate for teams, coaches to use and

trust, but.

the computational power it takes to get these models accurate.

For example, like, let's say I have a model that's 90 % accurate what it wants to do.

To get it from 90 to, or I should say, let me say 95 % to 96 % versus getting it 80 to 90
% is exponentially harder because I just like the amount you have to train it to get

better.

So yeah, right now that's why I'm kind of in the race

money pre-seed.

But yeah, sorry I don't remember the initial question.

fine.

It's fine.

I it's like, when you're as you're training or as you're like, you're building the tech, I
have no idea.

I'm not I'm not technical.

I'm like, you know, I'm a BC guy.

like a podcast guy.

I'm not like a tech guy.

Right.

So do you do know, feed your model like videos of like, like I'm a Phoenix Suns fan.

If you could you feed it a Phoenix Suns game and it gets better?

Is you actually feed it like live images or a video of you of you working is actually none
of that.

And you're just going off of like a

it already knows how the motions work because it's AI.

It's like, how do you know how to train it?

I don't know.

I think, I guess know how to train it based on, let's say there's a model, let's say
there's, yeah, there's a model of like object detection.

So just being able to detect like player hoop ref ball, for example, and you'll, put in a
video through an API, get the results back to see what it labels.

You'll go through and see, okay, it's misreading like people court side as players or it's
misreading refs as players.

So that's where you would go in basically uh redo the

and then train it again.

uh So yeah, a lot of my time is spent doing that.

I think a lot of it's going to a bunch of basketball events and games and things like that
and connecting with coaches, ah users, things like that.

uh That's pretty much it right now.

Yeah, that makes sense.

bet when coaches or folks in the industry or in like, I guess the sport, hear about what
you're doing, they must be jazzed.

Like there's probably not that much out there that's quite like this, right?

Is that right?

Yeah, so there's a couple things that are out there that are going or do something similar
to what I want it to do.

uh Like the base product of being able to...

uh get possessions clipped up, right, from a full game.

But that's done manually, uh for the most part right now.

I know of a couple other companies trying to do some computer vision stuff.

I don't think like they have all the features that I'm thinking of.

It's my knowledge anyways.

uh yeah, like everybody I've talked to in the basketball world and even in general have
been super supportive, like think it's a great idea, like it would be sick.

ah They tell me I'm the person to do it.

So.

Now I'm just trying to have that come into fruition.

no, 100%.

That's great.

Do you spend any time thinking about other parts of tech or like is there any is there any
developments in the startup world or tech tech world that are interesting to you that may

not be directly related to what you're currently building?

Yeah, I think something interesting, it's kind of tied into what I'm building, but just
like the whole AI agent thing, like how much you can really automate.

uh And I honestly think AI, I don't know if it'll replace entire jobs.

It probably is already starting to a little bit, but like you're higher, more skilled
jobs.

I think if you don't know how to use AI, like you're definitely gonna fall behind.

uh And so if you know how to, know, prompt LLMs, build agents, things like

that can make you more efficient.

think that's where AI is the most helpful.

Like my thing even, I don't want this to replace video coordinators or coaches or
analysts.

I think it'll make their job much more efficient because they'll be able to go through
more film, at more data, things like that, and just have more informed decision making.

The thing with agents though is I think that no one really

has a defined definition of what an agent is and everyone just slaps agent on their
product and they're like, hey, we have agents and then us as a excited industry is like,

oh, know, XYZ has an agent and we get excited.

have you seen uh an agent that you think is the future?

Like, guess, let me actually, let me ask a different question.

Have you seen technologies that you think will be around in 15 years that like power
agents or an agent that you think in its current form will exist in 15 years or are we

just really early?

in that development of like a technology in your opinion.

yeah, I think a lot of it is really early.

think a lot of people are kind of throwing and then we're gonna see as time progresses
what sticks.

uh I think as far as AI agents, I think things that are good are being able to...

use it as almost like an external brain.

So if you can connect it to your email, Otter, Otter bots or whatever, note taking apps,
things like that, and it's able to recall information for you that you might have

forgotten, I think that's where it could be really cool or helpful.

And I kind of have some stuff set up for that.

But as far as like an agent doing something completely for me,

You can do it.

I don't know if I trust it yet, but you can obviously do it.

Yeah, I-I-

I haven't leaned out on it, but I'm kind of waiting for just the industry or at least some
folks to be like, oh, this, this is the thing to try.

Cause I feel like I've been not burned, but kind of disappointed in being an early adopter
through a lot of this tech specifically in the agent world.

It's just never quite, at least from my experience, never quite as marketed.

So I'm like, I'm expecting to do blanks and then we can get you maybe 80 % of the way
there, but you can't get the full, the full hundred.

And it's kind of a letdown cause the whole idea of a gut.

an agent is it can work a ton.

You tell it what to do and it does it, then it tells you once it's done it, right?

Or you just know.

So I think it's probably out there, just like the phrase, what the future is here, but
it's not evenly distributed, like something like that, right?

It's probably similar.

So I have a couple questions around your time in college.

did you, like you play basketball, you you play basketball.

Were you, are you art?

Well, I'm gonna ask, are you still in college or have you graduated?

Cool.

So when you were in college, were you, were a lot of your peers?

uh You said you wanted to do, was there a lot of tech like in college?

Were you influenced by the scene at your college or not necessarily?

Um, I think when I was in college, like part of my college years was COVID.

Um, so that's when we were all on zoom.

So that's, guess, tech, right?

As far as AI, like no one was really worried when I was there worried about people using
chat to BT for essays or anything like that, because it wasn't that good.

Like, I don't think it was until this past year, maybe where it became really good and you
can give it context.

can give it like templates for it to kind of follow, um, an instruction, but, um, I wasn't
using any, any of like chat to BT.

or entropic or Gemini or any of that when I was in college.

But I think we definitely knew it was kind of on the come up.

um And it was more innovative than real life.

I think we still thought it was, as far as I know and what I use, speaking personally, I
think we thought it was more innovative.

This will be in the near future, but it wasn't implement now.

Totally.

I honestly can't imagine having access to this tech while I was in school.

I graduated Arizona State 2016 before all of this.

Just like a year after opening, I was founded, I guess.

And it's just kind of crazy to think that, yes, students just today, any age, not just
college age, but high school, middle school, they've access to this tech.

And then some schools are gonna...

lean in and they're gonna help the students use it to the best of their ability, kind of
like a calculator, and some are gonna say no AI for anything and it'll be interesting to

see how things just, know, things go in the future.

So what do things look like for you, you know, over the next couple of years?

Like what's the focus on for plays-wise?

Like you said, you're raising a pre-seed.

What's like some of your goals that you'd like to have accomplished within a couple of
years?

I think a couple years, I mean, I'm kind of delusional, but like if I could say every like
Power 4 school was using it, I think that would be the goal in two to three, four years.

Like I want it to be kind of league wide.

want it to be accessible for teams and players.

I want as many people and users as possible.

I think that's with any given software, right?

But I think something to be able to say, oh, every Power 4 school,

college uses this software, um I think would be incredibly cool.

And I think just being able to like right now the industry or like the the products that
I'm I guess competing with have completely monopolized the industry and kind of everybody

in in basketball has told me that like, huddle and synergy are their main video platforms
to get game footage uh kind of reviewed and stuff.

And I I kind of want to disrupt that.

Because I think I think they're lacking with some things and I don't think they've been uh

um as present for their users.

I kind of wonder if they have a disadvantage in that they were built well before this AI
boom.

So they're not AI native versus you're building this now you're able to build kind an AI
native company from the ground up.

It probably gives you an advantage.

Mm-hmm.

Yeah, would think so.

I think I can be more efficient at things um than somebody else could, you know, manually
doing a lot and they're still doing stuff manually.

Built, you know, 20, 30, 40 years ago.

Like, you can even tell by the UI of HODL and Synergy, like, it looks like it's from the
1990s or whatever.

um

That's crazy.

I'm familiar with Huddle.

There was a period I had a company a few years ago called Seed Scout.

We were helping connect at that time founders, investors and job seekers together.

But then we started to expand for a period of time to athletes.

So we had an ambition to connect athletes to coaches, to schools, to brands.

And we did it for six months or something.

But I found Huddle around this time and it's fine.

It does the job, but it's definitely not a magical experience by any means.

yeah, I would agree.

And then I don't know what I guess package you had, but seeing some of the stuff on what
they give to teams on like how they label plays or actions like half that like there was

one time I was on it and it labeled the complete wrong player like not even close.

So it's just like things like that.

Like when we're talking like winning games and

the threshold to what's gonna push you over to win a game.

We just can't have silly mistakes like that.

And that's why we obviously have coaches, analysts, a lot of people on staff to do it
manually.

But when you're paying a large subscription, I think it should be more accurate or better
product, better user interface, stuff like that.

100 % no it makes sense.

Well if someone wanted to kind of learn more about what you're doing online, connect with
you maybe on Twitter or X, I still call it Twitter, I probably always will.

Where can someone learn more about Playswise, you know, your online presence, where can
someone connect?

Yep, mean Twitter's good.

check my Twitter and Instagram DMs pretty regularly, LinkedIn, all the basic social media
platforms and then uh can email me at bo at playswise.com or check out the website and

connect with me there.

Cool.

Well, I appreciate you coming on the podcast.

It's really interesting space that you're building in and I'll be following along.

Thanks for coming on.

Thanks, Matt.