Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:00
as soon as you have a strength conditioning
0:02
coach, who's a really good strength in conditioning coach,
0:04
having to also manage a
0:06
ton of data means you take away their time from
0:09
doing probably what you really hired them
0:11
to do and makes them less effective in that
0:13
space. So I would just say if we're not speaking
0:15
to someone in a physical preparation or the sports
0:17
medicine department, I'm speaking to an owner.
0:20
Sporting director or something, I would say, you want the
0:22
people that you hire as practitioners, being practitioners.
0:26
and you probably want people to do that to do
0:28
that.
0:40
Welcome
0:40
to the Paci Performance Podcast.
0:43
The podcast that dives into the philosophies
0:46
ideas and practices of
0:48
some of the best practitioners in high
0:50
performance spa.
0:52
Data science and data scientists
0:55
with no thought of their own, have got
0:57
mystique around them.
1:00
So members of staff sat in a darkened
1:02
room, not speaking to anyone, is
1:04
after a misconception of what data
1:06
science is. So what what is
1:08
data science? So we get today's episode,
1:10
we've got Johan Win. who
1:13
is head of Performance Data Science
1:15
at the Vancouver Whitecaps. So
1:17
where does data science fit into an organization? what
1:20
do they actually do? So we talk about
1:22
subject to evaluations, technology
1:25
evaluations, maximizing data
1:27
quality, And most importantly,
1:29
how data science influences various
1:32
aspects of the organization and
1:34
how it helps other
1:36
departments with their job and
1:38
makes their jobs easier by providing
1:41
good quality data. So
1:43
really interesting chat coming up. And one
1:45
thing that I think is the biggest takeaway
1:47
from this is how data science is
1:49
used to evaluate the effectiveness of
1:52
certain interventions from a spot science perspective
1:55
and also from a strength and condition perspective.
1:57
So a really interesting episode coming
1:59
up with you on.
2:02
This episode of the PACE of Performance podcast
2:05
is sponsored by Hydro. Have
2:07
you ever tried a blood flow restrict for
2:09
recovery. Hatrol has developed
2:11
the world's first BFR wearable
2:13
unlocking the recovery benefits of BFR
2:16
to spot athletes. BFR
2:18
is no longer just for one on one
2:20
physio rehab. Hydraul allows
2:22
teams to use this safe and scalable
2:25
sports BFR device, both exercise.
2:27
to dramatically enhance recovery.
2:30
So whether in the changing room post game
2:32
during a way game travel in the hotel
2:34
or at home, HATRO has
2:36
created a simple and effective tool to allow
2:38
BFR to be delivered to athletes
2:40
and squads simultaneously safely
2:43
and more convenient than ever before.
2:46
Check them out hitroome dot com or
2:48
email warren on warrenhydro
2:51
dot com to find out how
2:53
hydropie FR can give your athletes
2:55
a competitive edge. And
2:57
this episode is also sponsored by
2:59
Hocken dynamics. Hocking Dynamics
3:02
is the world's first wireless force
3:04
plate testing system. The Hocking
3:06
Dynamics system is built for coaches to
3:08
test in the real world, not just
3:10
in the lab. Capture reliable
3:12
data on all your athletes in matter
3:14
of minutes. and monitor their progress
3:16
in the cloud from anywhere in the world.
3:19
Dark
3:19
and Dynamics Forceplates are wireless,
3:21
portable, and trusted by teams
3:23
at every level of sport.
3:25
Integrating force plays into our athlete
3:27
monitoring program has never been
3:29
easier or more affordable. If
3:31
you want to see the Harking Dynamics BOSS
3:33
Play System in action, head over
3:35
to their website harking Dynamics dot
3:38
com. to schedule a demo or
3:40
for them on Twitter at harking
3:42
Dynamics. Also, sponsoring this episode
3:45
is Fusion Spark. Smartabase from
3:47
Fusion SPORT is the premier human
3:49
performance optimization platform
3:51
for elite sports teams and military
3:54
organizations. Built an
3:56
infinitely configurable framework,
3:58
Smartabase is the most flexible
3:59
software on the market. Create
4:02
an adaptable solution to support
4:04
your unique strategy, process
4:06
and culture for a fraction of the oftentimes
4:09
it takes to build your own. Centralize
4:11
your performance and health day by
4:13
easily integrating with other tech and
4:15
data systems using
4:17
smarter bases, robust
4:19
API and custom built connectors.
4:21
In proof performance and reduce injury
4:23
by enabling better communication and
4:25
decision making with role based access,
4:28
custom workflows, mobile
4:30
apps, and
4:30
personalized visual dashboards.
4:33
And with the Smartabase success
4:35
guarantee, you can be confident in
4:37
your human performance solution and
4:39
the people who stand behind it. Visit
4:42
fusionspot dot com forward
4:44
slash Smartabase to learn
4:46
how Smartabase can help you improve athlete
4:48
performance and service member
4:51
combat readiness.
4:52
So now further ado, over to the episode
4:55
with Johan. Johan
4:56
went Welcome to the pacing
4:58
performance podcast. It's a long time coming
5:00
that I'm delighted to have you. Hey.
5:02
Thanks for having me, Rob. It's a real pleasure.
5:05
No. Thank you for coming on. Thank you for coming on.
5:08
lots of back and forth on text.
5:10
Talk about
5:11
our journey through Father Hood.
5:13
You are far more experienced than me.
5:15
but
5:15
it's good to where it's good to get you on the podcast
5:17
and have a little chat about what you've
5:19
been doing at the at the football club,
5:21
what you've done in the past, and what kind of good
5:23
stuff. So and one that doesn't know who you are.
5:25
Would
5:25
you mind just giving us a brief bio?
5:28
Yeah. Happy to, Rob. I think first and
5:31
first and foremost, you've already flagged that probably
5:33
the most prominent parts of my life is
5:35
is right now husband and father. So I
5:37
I married my childhood sweetheart, and we're
5:39
now currently still trying to figure out this
5:41
parenthood thing with with a
5:43
couple of young ones. So if you let me know if
5:45
if you do figure out, please let me know. because
5:47
I mean, also, it's a trouble. I
5:49
can clearly describe it as a a very fast moving
5:51
target. So I don't if I if I
5:53
figure it out, I'll text you. And by the next day, it'll
5:55
be different. So I I don't know how much help
5:57
I can be.
5:58
But, yeah, professionally,
5:59
I think my
6:02
background would probably be described
6:04
as as
6:05
varied and then pretty consistent
6:07
with one foot in an academic space
6:09
in one foot and applied. So if we go,
6:11
like,
6:11
all the way back, moved around
6:14
a lot, grown up, like born in South
6:16
Africa, rooted
6:16
away when I was three, lived in
6:19
Canada for a year, moved to New Zealand,
6:21
lived in New Zealand for five years,
6:22
became a rugby and cricket fan. then
6:25
move to the middle of nowhere in Northern
6:27
Alberta
6:28
back in Canada and then spent
6:30
high school in the Vancouver kind of lower mainland
6:33
area.
6:34
And then underground went
6:37
to a school here in Langley,
6:39
did exercise science, physiology, with
6:41
my undergrad, like, a a ton of the the people
6:43
you've had on the pod. And then during that
6:45
whole time, tried to get as much
6:47
applied experience as I could. So I spent three
6:49
years as kind of a strike conditioning intern
6:51
during that time, trying to get on
6:53
the floor. Got my CSCS. Did
6:56
the the strength and conditioning route
6:58
and then graduated through
7:00
there,
7:01
tried to figure out what I was going to do next,
7:03
was either looking at academia, which
7:05
would have been the master's approach or in
7:07
an applied space of Physio.
7:09
An MD tried to go to
7:12
med school, potentially my dad was a doc, so
7:14
that was always one
7:15
of the discussions. Spent,
7:16
like, six months out of school,
7:19
trying to figure that out, and
7:21
then Professor Kieran Paun,
7:23
from Procter and Conn's Clinical Sports
7:25
Medicine, British Journal Sports
7:27
Medicine, all the things that he was doing at the time.
7:29
He gave
7:30
lectured through a bunch of physios and
7:32
gyros in in Avbotsford where I
7:34
was at the time. And I was like, I'm gonna
7:36
do MD, PHD. This
7:38
guy is literally both of those. He'll probably have
7:40
some wisdom to share about what's a good idea
7:42
in
7:43
that conversation.
7:44
sparked a kind of follow-up. He just said, look,
7:47
a
7:47
master's degree is not a bad choice if you're gonna
7:49
go either of those groups. So why don't you send
7:51
me some of your stuff? And and that led to a master's
7:53
degree with him. He said, look, if you got
7:55
a master's that sets you up for med school or
7:57
if you wanna do a PhD, do the master's first.
7:59
So that was
8:01
a really, really great conversation or
8:03
really, like, pivoting
8:04
point kind of in in my
8:07
career path. So I did the masters with them.
8:09
It was in
8:09
physical activity promotion. We work with
8:11
family practice physicians
8:13
trying to convince them that,
8:15
hey, one of the most
8:17
important things you could do in a fifteen
8:19
minute consultation is encourage your patients
8:21
to be physically active. It has, in
8:23
many instances, as much or more
8:25
benefit than many of the pharmaceutical interventions,
8:28
especially when there's nothing major acute
8:30
going on.
8:30
It's one of the major investments you
8:33
can do for physical activity promotion.
8:34
thought
8:35
it was super important as a topic.
8:37
I didn't necessarily love it. I think my
8:39
heart was still more in that sports space,
8:41
but that was
8:42
kind of the thesis and then
8:44
again,
8:44
tried to maintain that applied space. I was
8:46
a strength and conditioning coach at a local
8:48
college. I proposed
8:51
and then developed and taught a health
8:53
and fitness class at the college where I was teaching
8:55
as well, so I tried to do that.
8:57
And then
8:58
any random consulting
9:00
I could do in this kind of sports science, SNC
9:02
space. I just pick up and say, yes, to far too
9:04
many things.
9:05
And then during
9:07
that master's grad, this this interesting
9:09
hiatus. My my wife and I traveled to
9:11
South Africa. I got a research abroad scholarship
9:13
and we had, like,
9:14
five months living the life that
9:17
probably
9:17
would have had had we never left
9:19
South Africa in the first place. So I was
9:21
at the University of Cape Town where my dad
9:23
studied at the Sports Science Institute, the
9:25
Tim Nokes started My supervisor
9:28
there at the time was Martin Schwalnoz who
9:30
was just on his way out to to move
9:32
to Pretoria and Wayne German, the guys
9:34
that my dad studied with, but I've
9:36
spent time with all the other students. They have
9:38
noakes hour. So on, like, Friday afternoon,
9:40
Tim will sit down with all of the students
9:42
and staff that wanna go with the Sports Science
9:44
Institute and just kinda outline
9:45
his career journey and path for all the
9:48
students that are there for the year. So we'll talk about the
9:50
hyponatremia and the the
9:51
dehydration thing. We'll talk about central
9:54
governor theory and all of things that he's had there,
9:56
obviously, now with the low carbohydrate
9:58
diet approach and how he's
9:59
handling all of those things. So it's just really
10:02
interesting
10:02
stint
10:03
about, like, what life could have been, which
10:06
was really, really cool. And then, again,
10:07
we're
10:08
kinda at that discussion kind of at the end of
10:10
the undergrad. Like, what do I do next?
10:12
And we're trying to figure that out. And then while I was
10:14
in South Africa, I found out I had BHE
10:16
funding. And Karen
10:18
agreed to supervise PHD. So,
10:20
again, we moved back. We went to Vancouver.
10:22
I started the PHD, moved out of
10:24
the physical activity promotion space, and kind
10:26
of moved into the injury
10:28
prevention space. So I was co supervised, Tim
10:30
Gabbett, and then Karen
10:31
Pond was the other one
10:34
there.
10:34
And then pretty soon after signing up, I got
10:36
embedded as a an intern with the
10:38
Vancouver Whitecap. So I spent the first two years in my
10:40
PhD with the club and I'm kind
10:42
of a sports science role on the ground each
10:44
day collecting wellness measures on
10:46
the pitch with with catapult and trying
10:48
to get
10:49
a handle on how can you collect this
10:52
information and kinda support the team. So it's getting the
10:54
applied experience
10:55
then then on
10:56
the academic side,
10:59
there's some huge coffees. Put
11:01
another one. I'm
11:03
I'm afraid I'm gonna run on. I think I drank too
11:06
much before the podcast. I
11:08
have
11:08
a backup water. So we'll see how this goes.
11:10
Six AM is six AM for you. How's
11:13
it's just a bit of context.
11:15
Sorry if I carry on. No. Of
11:17
course. It was it was interesting. Start of the PHD
11:19
was another one of those conversations again
11:21
with with Karen though is really helpful and kind of
11:23
pivotal for where my career ended up going. So
11:25
I definitely didn't see it going in in this
11:27
direction. But I think I was like a lot of
11:29
graduate students like you and that
11:31
conversation with them. A lot of students get
11:33
involved in academia and they learn
11:35
kind of the bare minimum enough
11:37
statistics.
11:39
to publish
11:40
a paper that they need. So in my
11:42
master, I did that. I learned how to do, like, a Macomart
11:44
test prepared comparison so that I published my paper,
11:46
and then I forgot about it. I learned r, I forgot
11:48
all of it,
11:49
and lost the statistics
11:51
stuff. And
11:52
then early in the PHE, it's like, okay. Well, I'm gonna
11:54
have to get back into this coding fast
11:56
thing
11:57
in talking with families. Like, you could
11:59
do that and a lot of
11:59
students do it. But look, academically, this can
12:02
become essentially one of your
12:03
superpowers that people often don't
12:06
consider
12:06
it. Like, you can dive
12:07
in, actually, learn, and develop
12:09
this as a skill, not as a, like,
12:12
a chore you have to do.
12:14
And that actually led into
12:16
me taking a sub specialization here at
12:18
UBC in the measurement evaluation research
12:20
methodology. substream.
12:22
So it's run by education and psychology. They
12:24
have a really strong psychometrics
12:26
department. It lets a conversation with
12:28
Professor Bruno Zumbow who's one of the world leaders
12:30
now in unified validity theory.
12:32
He's a psychometrician mathematician
12:35
really, really brilliant
12:36
You guys ended up being on my PhD
12:38
committee. So I learned about kind of the statistics
12:40
and research methods in a field
12:42
completely unrelated to support, which
12:44
actually I think it's come to pay dividends
12:46
for for how I think about different
12:48
sporting constructs because they're that's where
12:50
they live. In second metrics, you're trying
12:52
to quantify and
12:54
capture things like
12:56
emotional intelligence or even just general
12:59
intelligence or empathy.
13:01
It's like these things are constructs. We all think
13:03
and identify that they exist, but
13:06
you can't just strap
13:08
on a GPS and get someone's be on the other
13:10
end. You can't do that. So you have to develop things. And
13:12
then you have to say, how confident are
13:14
we that this survey we developed, captures
13:16
this thing that you can't have in that. I
13:18
think we end up doing that in support. We'll probably
13:20
get to that. But, like, learning about
13:22
that space was was really helpful during
13:24
that PhD. and then
13:26
trying to upscale myself thinking, hey, if the
13:28
statistics thing is helpful, this coding
13:30
thing might also be helpful to stop me from
13:32
doing all this manual I've done what
13:34
everyone does early in their careers in the
13:36
sports science space of, like, go
13:38
into software, download CSV,
13:40
all parameters. copy all of the
13:42
posts, paste it into the other spreadsheet,
13:45
manually clean the data, which is a
13:47
disaster, hope for the best, and
13:49
the publish
13:50
up the thing. Like, I I did that and I was
13:52
like, maybe there's a better more efficient way to
13:54
do this because this is a lot of time every
13:56
day doing the same thing.
13:58
So spent
13:59
a lot of
13:59
time just upscaling during the PHD,
14:02
trying to become more efficient and and helpful
14:04
in that space, and then wasn't
14:06
done two years into the PHD two years into the
14:08
Whitecap. So I was approached. I was at conference in Belfast.
14:10
Dr.
14:10
Dustin Knapp had at the time he was with the United
14:13
States Olympic and Paralympic committee. He said,
14:15
look, We
14:16
have a job opening in the sports medicine
14:19
department for a data analyst. We're trying to run
14:21
and generate kind of this athlete monitoring program
14:23
across and support national governing
14:25
bodies across the country. We think it'd be a
14:27
really good candidate. Would you be interested and
14:29
had that conversation? One thing led to
14:31
another, we packed up our bags, like,
14:33
a month later and moved to Colorado Springs.
14:35
So then I was
14:36
working in at
14:38
the Olympic training center there, trying to
14:40
develop and and build up these things, spent a
14:42
lot of time building Qualtrics surveys
14:44
for athlete wellness and athlete
14:46
load across a ton of different
14:48
sports have really good conversations
14:50
there also had the privilege of working with Dave
14:53
Taylor when I was, like, we've talked about him in
14:55
the past as well. So he's went on to Golden
14:57
State Warriors now. And then
14:59
Bryce Murphy, who went on to the
15:02
Orlando Magic now, IMG Academy.
15:05
it was, like, this this very condensed
15:07
period. We were only there. I was there
15:09
fourteen months in total. I was there crossing
15:11
paths with Dave for about a year and Bryce for
15:13
five months, but I is one of those we
15:15
work together for, like, a year, maybe less,
15:17
but we felt like we got three years of work
15:19
accomplished. It was just this really
15:21
fun office with
15:23
a lot
15:24
of Tableau dashboards and probably a little bit
15:26
too much espresso. But
15:28
it
15:28
was it was a really good time there.
15:31
And then tried to do the
15:33
the PHD right up the dissertation on
15:35
evenings and weekends, and we had a coffee shop that we'd
15:37
go through each weekend to try and get through
15:39
that.
15:39
And
15:40
were working there
15:42
less than a year end, and that kind of led to
15:44
this current role. I got a call
15:47
back from doctor Ben Spohr who had
15:49
taken over as VP of
15:51
performance here at the Vancouver Whitecaps. And
15:53
at
15:53
that time, they were having a bit of a
15:55
revamping in terms of the performance strategy. Some
15:57
staff had left to move to the
15:59
warriors and they were
15:59
reshaping and kind of planning what that
16:02
future for the performance side of the organization looked
16:04
like and, Vincent, but I've
16:05
had discussions with senior leadership. I've
16:07
pitched what I want this look like in one of the
16:09
pillars that I want within the performance
16:11
strategy and at the higher levels, I wanna make sure
16:13
that we have a data science to and there's
16:15
that we want to accomplish as an organization
16:18
strategically, whether
16:19
that be evaluate the effectiveness
16:21
of all the things that we're doing to have data
16:23
informed decision making to
16:25
look
16:26
at
16:27
these things in a in a long term
16:29
process perspective, looking at
16:31
what is the actual output just
16:33
the the outcome on a on a game to game basis.
16:35
And he said, look, to to do this, one of the things we're
16:37
gonna need is is to have
16:38
a data science
16:40
department or group dedicated to this. Would
16:42
you come back and kind of take it on
16:44
and and build it? And that's been
16:46
now since the start of twenty nineteen, three
16:48
and a half years of change my current
16:50
role and Yeah. Again, I've
16:52
managed to try and keep one foot in in that
16:54
academic space. I think now it's it's
16:56
largely applied. I teach a couple courses
16:58
here at UBC, I enjoy the teaching
17:00
side. We have an
17:02
embedded PhD students in partnerships with the
17:04
university to try and maintain this
17:06
kind of balance a daily training environment
17:08
here that we have to make sure that we deliver
17:10
on each and every day at a high level,
17:12
but also this research innovation space, which can
17:14
be really enjoyable as long
17:16
as it kinda can be used and supported
17:18
in in that the
17:20
actual daily operational environment if you can
17:22
bring it back. So long
17:24
winded answer, but that's the the story to
17:26
this point. No. It's
17:27
interesting. III like, especially
17:29
someone who has a
17:32
nonconventional route
17:33
into that position. So, yeah,
17:35
really interesting. One thing that I
17:38
that when I was listening, it
17:40
was interesting what you were saying in terms
17:42
of that.
17:42
foot in both camps. And that's something
17:45
that Robyn Thoth, I don't know if you've come across
17:47
Robyn, worked to Manchester United over in Red
17:49
Bull now. was very adamant
17:51
how important that is for a spot scientist
17:53
to have that
17:54
research one for in the
17:56
research. one foot in the applied so
17:58
that both can can aid each other.
17:59
So I think definitely echoed echoed
18:02
Robin starts there. But move moving on to
18:04
the data science piece.
18:06
data
18:06
science department
18:09
within the organization. Where
18:11
does that fit? Where do you
18:13
where does your influence
18:15
start and finish? And how is it kind
18:18
of sprawling to other
18:20
areas that may not be what people
18:22
imagine? Yeah.
18:23
No. A hundred percent. I think it's a it's
18:25
a great question. I think it's especially good
18:27
just because, like, that doesn't look the
18:29
same in every organization. And
18:31
I think one of the things before you can answer
18:34
where data science does or should I
18:36
think it's helpful that you just step
18:38
one level up to just even say, what do we mean
18:40
when we say data because it's weird. I
18:42
think a lot of people know what you mean when you
18:44
say like a deadlift or
18:46
a squat or a physiotherapist, you have a
18:48
pretty good sense. But it's
18:50
helpful to say, like, what do we mean when we say data
18:52
scientist? because, again, I think in support, it
18:54
probably looks slightly different than it does at
18:56
Google or Facebook just from the
18:58
nature of how many staff
19:00
can work in this space, what the bandwidth
19:02
is, what the scope of the organization
19:04
is. So when when I would refer
19:06
to data I I would refer
19:09
to
19:09
the application
19:10
of the data pipeline being
19:13
collection, analysis, and modeling, and
19:15
then communication of that data to inform decision.
19:17
And that's what I would do. So you you might
19:19
hear in a in a really technical space, you'd
19:21
probably differentiate, like data science is the modeling,
19:24
the machine learning model
19:27
generation process. And then, often, you'll have a
19:29
a data analyst or even a data visualization
19:31
expert for the communication side, and you
19:33
would have a data engineer and a database architect to
19:35
handle some of the collection and and
19:37
data basing of that information. So I think
19:39
you see more specialization in
19:41
other fields I think
19:43
in sport, at least
19:45
in a lot of the circles that I've had
19:47
discussions. To this point, it's the data
19:49
science department's kinda responsible for all
19:51
of those. Like, you have to figure out how to collect it. You have to figure out how to
19:53
manage it. You have to figure out how to analyze and model
19:55
when appropriate. And then communicate it to all of
19:57
the stakeholders. You have to do all
19:59
of it. And
19:59
in
20:00
that case, then I
20:02
think
20:02
i think yeah. I
20:04
think my answer and a lot of my thinking, we
20:06
we were talking about Patrick Ward before the podcast.
20:08
And some of my my own
20:10
thinking came out of a conversation. I think it was
20:12
twenty seventeen. I was in my PhD intern
20:15
here at the time. It
20:16
was the Saunders Sports Science Conference, and we were
20:18
just having this with this chat over drinks after
20:20
one of the sessions.
20:22
the And at the
20:24
time, I was musing over this conversation.
20:26
I I heard again, it was in
20:28
the S and C space. So Dan John, and it was
20:30
describing this conversation he had and said, what's
20:32
the of a strength coach, it's really easy to
20:34
quantify, help develop physical capacities like strength
20:37
and power, aerobic capacity, etcetera.
20:39
And then he said, but what's the impact
20:41
of a strength coach? and that might be harder.
20:43
Right? Like, you can influence the culture, you
20:45
can you can have these relationships
20:47
that change the dynamic of an organization, you
20:49
can help develop people and players,
20:51
all of those things. the impact's a different
20:54
thing than the role. And at the time, I was trying
20:56
to move more into this data science,
20:58
sports science space. Patrick was a
21:00
bit experienced and had been in it longer. I said, what do
21:02
you think the role and impact
21:04
is of someone in a sports science space?
21:06
Because, again, it's harder to quantify these things
21:08
at times. And
21:09
that conversation
21:10
actually then led to two papers. So
21:12
we Patrick and
21:14
I with Tom Canton published this one about
21:16
business intelligence saying, look, sports
21:18
science. We
21:18
in that one, we were talking about sports
21:21
science, is literally the application of, like,
21:23
data collection, data analysis, data
21:25
communication and then even using that as a decision
21:27
audit after the fact to help
21:29
inform decision making and support it. Same
21:31
thing right now, business intelligence does
21:33
in
21:33
that space.
21:35
That's kind of what I define
21:37
data science as well. But the other
21:39
part that came out is is when we're having that
21:42
discussion in terms of an an
21:44
IS team. I think all of us are familiar like, an athlete centered support
21:46
team. There's athlete, and then there's a sports
21:48
medicine team, and you're a physiotherapist,
21:50
and maybe you have physical preparation
21:52
coach or multiples. You have support
21:54
nutrition staff. You have technical coaching
21:56
staff. And you have video analysis team.
21:59
You have for
21:59
psychology, mental performance support. You have all
22:02
of these different
22:03
departments and supports and you put that
22:05
around an athlete. So I think the logical next
22:07
step in most people's mind is oh, it seems like
22:09
a lot of people are getting these sports science or
22:11
data science departments. Let's put them
22:13
in and then we have another cog within this
22:15
wheel of this athlete center support team.
22:17
And I Like, I I don't I think that happens
22:20
a lot. I think it can't function to start like that
22:22
will never work. I just don't think that's
22:24
optimally where data
22:26
science. If we think about that informing decision making pipeline, I
22:28
don't think that's the optimal fit. I I put
22:30
them as almost this glue
22:32
that
22:32
sits underneath
22:35
and
22:35
outside of all of those other practitioners
22:38
that that end up being the decision
22:40
makers when you have a physical preparation
22:43
they design the periodization, the plan loading,
22:45
the implementation of a lot of the
22:47
on pitch and in gym fitness work,
22:49
the sports medicine, team have to make
22:51
decisions about return to play, about what
22:54
type of prevention strategies are we going to
22:56
implement the coaching staff, the scouting and
22:58
recruitment, which players are we going Like, they're
23:00
the decision makers. And then our
23:02
job, and I think data science
23:04
optimally fits then, going to
23:06
all of those practitioners and saying, look, how can
23:08
we, as a data science department, make your
23:10
job easier or more efficient? Like,
23:12
what are the decisions you have to make?
23:14
What information can I come and give to
23:16
you? Either make it easier to make that decision.
23:18
Maybe it helps make it
23:19
faster, maybe it it challenges, or
23:22
maybe it encourages what you're already doing just.
23:24
And sometimes it's
23:24
literally like can I
23:26
set up
23:26
a folder structure on our shared
23:29
space just so your document
23:31
storage is a bit more
23:33
a
23:34
bit more organized and in line with the rest of what we're doing. And
23:36
somebody I need that. I'm
23:39
desperate
23:39
for that. But
23:42
I think when you're in a generalist space, right, like, it it isn't
23:44
always like, hey, can we build the most sophisticated
23:47
convolutional neural network and support?
23:49
Like, that's not where we
23:51
spend the majority of our time. And I think
23:53
that's that's a really healthy space. Just to be
23:55
one step below, we wrote a editorial in
23:57
BGSM, a few of us that have kind of
23:59
been in the space,
23:59
and that's I think organizationally where
24:02
I think it's optimal for data science and
24:04
support to sit. I
24:05
know people will have different perspectives. It's
24:07
how we sit
24:08
here and that's been
24:10
Yeah. I think it's been really nice
24:13
to transition into a role where that
24:15
is realized. And yeah.
24:16
Again, I I could go on. I'm gonna,
24:19
like, I
24:20
think if I can take a smaller
24:22
side, this this only works. So I think
24:24
it's optimal,
24:25
but the caveat is you you can't
24:28
do it if the
24:30
organization isn't set up for that to
24:32
work. Right? So as
24:34
a data science department, within a structure
24:37
like I'm not the primary decision
24:39
maker almost
24:40
in any realm. Right? So
24:42
what
24:42
that means is I can have
24:44
this really,
24:45
like, Let's say,
24:48
intellectual is
24:49
the wrong word. I can have this really broad
24:52
scoping role, which I really enjoy of being
24:54
almost like this glue between multiple departments to
24:56
have these discussions across the board and
24:58
understand what's going on, and try and
25:00
help as many people as I can. And
25:02
that extends again within each department,
25:04
but it if you have
25:05
an organization that has
25:07
a bunch of people that just think
25:09
they have all of the answers already,
25:11
and if they're instead of like,
25:14
encourage to look at other sources of
25:16
information or want to take information and
25:18
data onboard,
25:19
the like,
25:20
the data science department becomes
25:22
completely helpless.
25:23
It's useless to spend resources and
25:26
time. And then you have a frustrated group of
25:28
people in the data science space trying
25:30
to build dashboards or reports or something
25:32
and a bunch of decision makers who are like, why is
25:34
this guy in the room? I
25:35
already know what I'm doing. So
25:37
why is he here? And then that that
25:39
breaks down the perfect organizational structure if
25:41
you don't have a culture in
25:43
place and a structure in place and if you don't
25:45
hire the people that are willing to take that information
25:48
on board, then that it doesn't work. It'll
25:50
just implode, and I think that that can happen as
25:52
well. It's it's not just organizational
25:54
structure. I think it's it's culture. And I'm
25:56
I'm really fortunate here at the club. That's
25:58
when I talk back
25:59
to when they
25:59
revamped what the performance department
26:02
looked like, kind of when Ben took over
26:03
as
26:04
SVP And just within
26:07
the culture, I think I can list people
26:09
within every one of those departments
26:11
that have that have been brilliant
26:13
to work. like, in the physical
26:15
preparation space, John Foley and I have worked
26:17
together now since my first
26:19
joined. He's been here even before that. So he's been
26:21
brilliant as a physical preparation coach.
26:23
He's
26:24
he's brilliant to work with,
26:25
and that's been a really great experience. You'll know both
26:27
Steven Fletcher and Tom Ryan, the
26:30
some of UK invasion has begun. So they they've come
26:32
over and they've again, we we bring
26:34
on really quality people that are willing to take that
26:36
type of information on board. We can have these really
26:39
helpful discussions about how can I make their
26:42
jobs a little bit easier, and they've been
26:44
brilliant to have. So
26:46
that's
26:46
been great. Chris Branks James
26:48
Gartner has joined as
26:50
as an a t. Jose Yimin is on the sports
26:52
medicine side. Again, each of these departments have
26:54
these
26:54
practitioners that are great practitioners,
26:56
but also really good people that
26:58
are like you can have
26:59
these conversations with. Our
27:02
our director of outing, spent time working at what's
27:05
considered the one of the most data driven
27:07
football clubs in Brentford, F. C. Michelin,
27:09
owned by Matthew Benhama. a
27:11
a sports gambler turned owner. Obviously,
27:14
he takes a lot of data on board, and then he moved
27:16
from there into working directly with stats
27:18
on one of the a lead
27:20
data providers in sport. Right? Now he's come
27:22
on. So you can guess he's he's very
27:24
open and comfortable taking data onboard
27:26
to make decisions when we look at
27:28
look at players. We have a technical staff.
27:30
Not every coach is the type of coach that's
27:32
going to say, hey, what
27:34
you do the
27:35
data have to say about this performance,
27:38
about this player about these things and we have a group of technical staff
27:40
that are willing to have those discussions that we can be
27:42
in a room and just have
27:44
these
27:44
these chats about,
27:46
hey, where are we? What are we doing? How
27:48
can we improve? Where do we
27:50
see that
27:50
we've grown? And what can we say both
27:53
positively and negatively? And translate
27:55
that where necessary to other members of staff
27:57
or to the players. And those those
27:59
conversations don't
27:59
happen in every organization. We work
28:02
incredibly closely with, like, our video analysis
28:04
team with with Luke and Drew, and I
28:06
think from a date of his perspective, we'll
28:08
probably get there in this conversation. But,
28:10
like, if people
28:11
always say, like,
28:13
picture is
28:14
worth a thousand words kind of thing. And
28:16
and I was just, like, the other day thinking, like,
28:18
a a video in some cases is worth, like,
28:20
a thousand dashboards. Like, you
28:22
could take the most sophisticated model, you could build a
28:24
thousand dashboards to show up. But then when you
28:26
wanna translate that sometimes to a a player or
28:28
to a coach, you can
28:30
show one video example of saying, we want to do
28:33
this more. And that's
28:34
gonna be way
28:35
better than, like, my most sophisticated Tableau
28:38
visual.
28:39
in many of those cases.
28:40
But that again only happens if you work
28:43
with the other practitioners and
28:44
the staff that are one going to help
28:47
you create
28:48
that content and translate that across
28:50
to the other practitioners or to the
28:53
players. And
28:53
I think that's the other thing
28:55
from the structure. I know it's a bit of an aside here
28:57
on that space. But I think
29:00
in that
29:01
structure, if you go and you make
29:03
data science this supporting structure with
29:05
all of the other practitioners,
29:08
the it it actually
29:10
does remove us one step
29:12
from an athlete. Right? So
29:14
if
29:14
now I help make Jampoli's life
29:17
easier or loops
29:19
like easier or our coaching staff
29:21
their life easier.
29:23
And
29:23
by helping all of them,
29:25
what
29:25
it means is often, I think the information
29:28
that we provide is most effectively
29:31
translated by the person that usually works
29:33
in that avenue with an athlete. So John
29:35
Polley can communicate, hey, here is
29:37
where your loading is at. Here's why
29:39
we think you should come on for sixty
29:41
minutes this next game because we're in this
29:43
congested schedule. we think you'll play
29:45
sixty, then you'll be ready on the weekend. Hey, we're going
29:47
to implement this new Arova
29:51
capacity training top of what you're doing because, hey,
29:53
we've identified this on your screening or this
29:55
on your testing or we've looked at your loading over
29:57
time and here's where here's why we're making
29:59
this decision. So
30:00
I help build out the dashboard
30:02
or report that identify those things.
30:04
But I also need
30:05
to be comfortable that I'm it
30:07
probably means
30:07
more to the athlete coming from
30:10
John Pope. And when the coaches say, hey,
30:12
we want you to do this behavior more. I might
30:14
have some information that helps
30:16
guide the decision to say, hey, we want to
30:18
encourage this action or this type
30:20
of play, it means more when
30:22
a coach, like Ricardo Clark, who's
30:24
played it in the world. Cup goes to the player and
30:26
says, hey, you're doing this. Maybe you
30:28
wanna do this more than
30:30
having a data science background and be like, you know
30:32
what I found? Let me pull up my dashboard and talk
30:34
about this. It's it's it's a different
30:36
role. And sometimes, like, having come out of the
30:38
strength conditioning realm where you're on the floor, you have these
30:41
discussions every single day with the athlete. I'll
30:42
have less interaction with athletes
30:44
in structure than I have in the past. And
30:47
that's also just
30:47
personally some of the things that you need
30:49
to to work
30:50
through and kinda kind
30:52
of work with. But is is that sometimes a criticism
30:55
of this
30:56
type of structure? I think
30:58
I think it
30:58
can be because I think Ryan Curtis
31:01
had that discussion with you as
31:03
well because the more embedded you are, the more
31:05
you're there with the athletes, like, the more you're on the
31:07
floor, the more likely when
31:09
you do have something you want to communicate. Look, they're more
31:11
likely they are to list. Right? That
31:13
that that's a
31:14
that the that's going to
31:15
happen. You like Ryan mentioned, you you're
31:18
more likely see some of the
31:20
challenges, the hiccups, the pain points,
31:22
especially on, let's say, the physical preparation
31:24
sports medicine side if you're in that room and on
31:26
the floor every single day. I
31:29
think the way we try and balance
31:32
that is like we have a a sport
31:34
scientist we've hired, Luke petty, He's
31:36
in his PhD here at UBC. He has extensive
31:39
experience, both as, like, a national
31:41
level swimmer, but also now helping
31:43
in kind of the sports science supports
31:45
SaaS based and and swimming. Now he's moved over
31:47
into the the football realm. So he he's
31:49
within our data science department,
31:52
but he's he's our sports scientist and he's on the ground with the
31:54
athletes every single day. So what we do
31:56
then if within that structure is
31:59
loop is primarily our conduit with our sports
32:01
medicine staff and our physical preparation
32:03
staff. So we as a department
32:04
spread all the way through, but to
32:06
make that as efficient
32:07
as we can,
32:09
Like, Luke then is on the floor communicating every
32:11
single day. So one of us as a
32:13
data science department needs to help
32:16
with force plates or warm ups. Like, he he's kinda
32:18
first line there. So he's gonna spend the most time
32:20
with the athletes. So he'll rep the data
32:22
science department. It's not always loop. So we just this
32:24
past week, we've rolled into academy testing. So
32:26
there's whatever eighty
32:27
academy athletes that have arrived. We need to run them
32:29
through a kind of testing. So all three of the
32:31
data science department alongside Tom
32:33
Ryan or they're doing whatever we need to to make
32:35
the testing function. We're willing to do that, but
32:37
I think we try and
32:40
have representation in each department. So Luke's there in the
32:42
sports medicine space and in the physical preparation
32:45
space. He's on pitch. He's in the way of room. He's
32:47
doing all of those things. But
32:49
we also, again, in that structure, don't just live in that space. We also
32:51
have to help with the coaching staff and with
32:54
opposition analysis, with scouting and
32:56
recruitment. So In that case, we also have
32:58
Alexander Henson who's a
33:00
more brilliant and better data scientist than
33:02
I will ever be. We hired him after he finished
33:04
the masters in economics and he finished the
33:07
master and data science. During that time, he got to do a Capstone
33:09
project with us. It was a brilliant
33:11
eight week time. He was a member of a group
33:13
that worked with us. And he was one of those
33:15
as you spend eight weeks with him. Like, if we can get him
33:17
as part of the organization, there'll be a big win for
33:20
us. So we we we talked and we
33:22
told him we got him here, and he's been he's been a
33:24
great addition. But there, as our senior
33:26
data scientist. Now he spends more of his time
33:27
really closely with our coaching staff, with our
33:30
video analysis team on
33:32
opposition
33:32
scouting. spends more
33:34
time building some of our models off
33:36
events or tracking data or doing some of
33:38
the database architecture. He spends more time doing those
33:40
things. and being our primary conduit down
33:42
other avenues. And then I'll have the
33:44
same, like, communicate in certain avenues
33:47
more so than others. to
33:49
try and identify. But I think back to your
33:51
point is a criticism. I think if you were
33:53
completely removed, so we do sit in a different
33:55
office a lot of the time when we plug
33:57
away if we only sat in this room and we never got out, we
33:59
never spoke to the other practitioners,
34:01
it's really
34:01
hard to have positive relationships and
34:04
ask the questions about how we
34:06
can make their life easier if they don't know who we are and
34:08
we don't experience any
34:09
of those things. So I think it's it's a
34:12
balance point that
34:12
you have to hit. We've tried our best to
34:15
do that with our structure, but I think some moving
34:17
target kinda like like parenthood
34:19
as well. No. No.
34:22
I understand. So we're just gonna
34:24
take a very quick break in the chat with
34:26
Johan Hope in Joint Power one. So we
34:28
have a little chat around
34:30
data visualization in Power two. some
34:32
principles that Johan and the team at
34:34
the Whitecaps live by, which is not
34:36
ugly and efficient. What that
34:38
actually means how they live by it,
34:40
an example how you can
34:42
improve your dates visualization as
34:44
well. So really interesting part too
34:46
coming on.
34:48
This
34:49
episode of
34:51
the Pasty Performance podcast
34:53
is sponsored by Omega Wave.
34:55
Omega Wave is the only non
34:57
invasive at rest technology on the
34:59
market that analyzes readiness to train via both brain and
35:02
cardiac analysis. Using DC
35:04
potential and HRV to understand
35:08
your brain's energy level and autonomic nervous system
35:10
balance allows you to use objective
35:12
data on recovery and
35:14
readiness that
35:16
in turn helps you to
35:18
truly individualize your training and thus optimize performance.
35:20
A megawave also measures
35:22
ECG from the v six
35:24
position. this
35:26
data can be used by the medical profession to cardiac health
35:28
on a frequent basis. The
35:30
measurement takes only four
35:32
minutes to perform
35:34
and results are visualized in an intuitive way,
35:36
thanks to our windows of trainability
35:39
concept. A megawave
35:40
used by hundreds of
35:42
elite sports, military, and law
35:45
enforcement organizations. A
35:46
Mega Wave, also the official partner
35:48
of the UFC performance in YouTube.
35:51
Learn more about A Mega Wave
35:51
by visiting their website at megawave dot
35:54
com and their social media
35:56
channels. And this episode is also
35:58
sponsored by
36:00
Tanta College. Tanta College, led by doctor Liam Hennessy,
36:03
provides international recognized qualifications in
36:05
strength in conditioning and
36:08
performance science, to
36:10
master's level. Castle designed
36:13
by industry leaders such as Des
36:15
Ryan and Professor Iain Jefferies,
36:17
ensuring students and
36:19
graduate at the cutting edge of technology and learning
36:21
the most current methodologies from
36:24
world renowned
36:26
practitioners. Instacology's
36:26
unique blended learning approach allows you to take
36:29
the next step in your career, in your own
36:31
time and in your own pace.
36:34
Lectures are delivered in an online classroom,
36:36
while residential workshops provide
36:39
the perfect opportunity for
36:41
practical application of your studies with
36:44
guidance from experts in the field
36:46
of sports science and performance
36:48
coaching. With
36:50
campus location across Ireland, the UK,
36:52
USA, India and
36:53
South Africa. Applications are now
36:55
open for courses in
36:58
including the BSC in strength and conditioning,
37:00
MSC in
37:01
performance
37:01
coaching, and MSC in applied
37:04
spa and
37:06
exercise physiology. visit
37:08
santa college dot com for more
37:10
information on how to
37:12
apply. And now back to the episode
37:14
with Yoam.
37:16
So there are many people out there.
37:18
There are probably more people out there who are one man bands
37:20
who are one man band
37:21
plus an
37:24
intern. versus the data science department. So just to
37:26
finish off this little area that we're
37:28
we're we're chatting in regarding data science,
37:31
what would be
37:34
the
37:34
what would be your
37:36
y'all
37:38
recommendations
37:38
for those people
37:40
to upscale
37:42
not only for now, but to future proof
37:44
themselves with
37:45
new technology, the amount of
37:47
data that's been processed, generated,
37:51
and has to be analyzed, visualized, cleaned, all
37:53
that kind of stuff. What would be
37:55
your recommendations for those
37:58
people to take on as a I suppose
37:59
the
38:00
in situ default data
38:03
scientist because there's there's only
38:06
one there. will be our recommendations. I think it's a I think
38:08
it's
38:08
it's a really important thing to consider because
38:10
I think that is the case in many
38:13
many organizations that The real
38:15
quick caveat I think is organizationally. I don't think it's
38:17
optimal because as soon as
38:19
you have
38:19
a strength conditioning coach who's a
38:21
really good strength initiative
38:24
coach. Having to also
38:25
manage a ton of data means you take
38:27
away their time from doing probably what you
38:29
really hired them do and it makes
38:31
them less effective in that space. So I would just say
38:34
if we're not speaking to someone in a
38:36
physical preparation or the sports medicine department,
38:38
if I'm speaking to
38:40
an owner, sporting director
38:40
or something, I would say you want the people that
38:42
you hire as practitioners, being practitioners, and
38:45
you
38:45
probably want people
38:46
to do that to do that.
38:49
that's an organizational thing. It's not the reality. Sometimes it's budget.
38:52
Sometimes it's just organizational
38:54
constraints. So you you do run
38:55
into this this avenue. And
38:57
then I think to speak
38:59
to the practitioners, they have to do what
39:01
they're great at on the floor with
39:03
the athletes on the pitch, whatever that might be, and
39:05
then also have to have some data
39:08
skills. Or let's say if almost like
39:10
myself, if you're in a practitioner role
39:12
and you want to gain those skills and say, I
39:14
actually want to do more of this the
39:16
future. I think
39:16
what I would
39:17
what I would say is I've come back to
39:19
that description of data
39:20
science being that full pipeline.
39:23
So I think if you Google, like, what is
39:25
data science or, like, upskilling in data science, you're
39:27
really quickly gonna get into, like, a coding course
39:29
and pipeline about
39:32
machine learning. which is fine. Like, you can learn that and understand what a
39:34
random forest does or it's, like, decision
39:36
tree. Oh, obviously, you can
39:38
do that. But I I think in
39:40
support, your day to day isn't just sitting back
39:42
and and fitting models. It's actually managing
39:44
data from its onsets to
39:46
its conclusion where you're trying to inform
39:48
some decision making often the
39:50
modeling doesn't necessarily have to be as
39:52
sophisticated as what it might be.
39:54
There's gonna be a time and space for that. So I
39:56
would say, you wanna in
39:58
that space, I think the most
40:00
helpful thing that I have found and the thing
40:02
that I've often recommended when people ask that
40:04
question, is
40:05
find projects. It can either be at your work or
40:07
it could even be personal projects. Like, if you
40:10
have a a
40:12
Fitbit or some wearable
40:14
technology, even though you have on your own, think
40:16
about what do I want to
40:17
know about whatever my own training or what do
40:19
I wanna know about the athletes that I currently
40:21
work with and,
40:23
like, build out something. Like, imagine what it
40:25
would look like at
40:28
the end and then try and figure out a way to collect the data. Maybe
40:30
for the first time that is going to be a
40:32
download CSV. But when you do it, you might find that
40:34
the spreadsheet's
40:36
a mess. that
40:37
you have six header rows that things
40:39
are out of place, and you have to
40:41
change a bunch of stuff in order to even
40:43
get it into another platform. And then
40:45
when you do that, maybe you're like, oh, I need
40:47
to create a new variable ranker to talk about
40:49
that. Like,
40:50
i 0II
40:52
my
40:54
speed or velocity. Like, maybe I wanna do
40:56
force velocity profile. I need to create some other
40:58
variable. Maybe I need to estimate B02
41:01
max from a thirty fifteen result have
41:03
these thirty fifteens or whatever, and I wanna get someone view
41:06
it to max that you
41:06
have to create the equation to get there. Maybe you have
41:08
to do some sort of analysis, create new
41:11
variables, describe what that is and some basic summary
41:13
statistics, and then you have to build
41:15
the visualization of dashboard or
41:18
report or whatever that is to
41:20
communicate that at the end. And I think you
41:22
if you break apart your learning into,
41:24
like, trying to solve real world problems,
41:27
whether they
41:28
be for yourself or for the
41:30
organization, then I think you encounter and learn problems
41:32
of what you actually do in sport. because
41:34
it's it can be one less motivating
41:36
to just
41:36
open up this online coding course
41:40
from track and just learn how
41:41
to type print hello world into
41:42
the coding platform and then go
41:44
through all of the syntax.
41:47
that's
41:48
not really how we learn on the ground either. It's a lot of time. Like,
41:50
I need to solve this problem. So how do
41:52
I do? Right?
41:53
So this one won't
41:54
be solved by Excel.
41:56
maybe
41:56
a survey that has a is a powerful enough platform will
41:58
be a great way to collect
41:59
this data. What we encountered here is some of
42:02
the stuff we wanted to collect is a bit too
42:04
sophisticated. So
42:06
a SurveyMonkey and a Qualtrics doesn't have what we need. So
42:08
then
42:08
we have to learn how to build apps
42:10
because then we can build it more to
42:12
what we need it to be. So then you
42:15
learn how
42:15
to do it. You learn it because you're solving a problem, not
42:17
because you're like, oh, I need to learn how to build
42:19
a a web application, and I'm
42:21
gonna
42:21
spend ten hours doing it. So I I think when I
42:24
think about learning. If
42:26
you think about those actual project oriented, what it'll
42:28
look this look like in the real world, I think you
42:30
start to learn and develop skill sets along
42:32
the way that will serve you well in the
42:34
future. That's kind of the big
42:36
things. And, again, it's across that pipeline. Ryan Curtis
42:38
is a lot of people, the
42:39
Jose Fernandez, to meet the
42:42
minimum required Well, sure. If you want the list Scott referenced that again. I just
42:44
he comes with every I wanna get in every week.
42:46
I want the first of the list. There's
42:48
a list. You want the list.
42:50
I I often go to, like, look,
42:52
find a problem, try
42:53
and solve it
42:55
with data.
42:56
the And
42:58
then, look look, last thing. There's just a last point
43:00
on this. If you're a one
43:02
man show and you're in charge of, like,
43:05
developing all the programming, end
43:08
loading, and
43:09
gym work, all
43:10
of those things, and then data's laid on
43:12
top. Like, I think it's just a really
43:14
important thing to come back to. It's a similar question
43:16
I asked with technology. Like, let's pretend you did
43:18
have all of the data and you you
43:20
do all of the work to put
43:24
it into and then you build the whatever report. what
43:26
in your practice as a physical
43:28
prep or a strength conditioning coach is going to
43:30
change if you have all of that data? Like, it's nice
43:32
to have
43:34
sometimes if you at the end of the day are,
43:36
like, I could dump
43:36
ten hours a week into trying to collect all of this and you'd
43:39
come to the conclusion
43:40
the the conclusion that like that, like,
43:42
I'll
43:42
still program this way. Like, this is still gonna be a
43:44
maintenance block. This will still be a build
43:47
block. I'm still gonna emphasize power and speed
43:49
here, and then I'm gonna emphasize
43:52
size the robot capacity over here. If you say, you're unlikely
43:54
to change most of your operational
43:56
things regardless of that ten
43:58
hours of
43:59
data spent. I
44:01
think it's worth would great to have all of the
44:03
data, but maybe it's not that important that
44:05
I need to spend ten hours away from
44:07
other things I could be either
44:10
at work or personally to say,
44:12
like, it's
44:12
not worth it at this point. And
44:14
then to have that conversation,
44:17
with the
44:17
leaders of the organization. Like, I
44:19
think
44:20
it would be helpful to have someone to manage
44:22
this, but I can do it because I've been hired
44:24
to do this. and I'm willing to
44:26
do this. But I
44:27
can't do all of the other job as
44:29
well. And I think that's the space
44:31
people have to consider and ask. before
44:34
they just jump headlonging it. Of
44:36
course.
44:36
One thing that I wrote
44:38
down pretty early when we when we started
44:40
talking was about
44:41
Ben when when you had the initial
44:43
chat with Ben was for
44:46
for this department and for you to
44:48
evaluate the effectiveness of
44:50
what we do, And I
44:52
think that is a really interesting
44:54
point across many conversations that I've
44:56
had, not justifying
44:58
what we do, but trying to understand
45:01
how we do
45:03
exactly that. Like, look at
45:05
the effectiveness of Australian conditioning coach or a
45:07
a spot scientist and that feeds
45:09
many other conversations around salary, around working
45:11
conditions, around staffing, all that
45:13
kind of stuff. So
45:16
just bringing that back into the frame from that that initial comment.
45:19
How do you
45:22
monitor, manage
45:22
the effectiveness
45:24
of the interventions that are going on
45:26
all over the ecosystem
45:28
within the performance department. I
45:31
think
45:32
that's yeah.
45:33
If if there's a quote about
45:35
our performance strategy, I think that that integrates with probably two or
45:37
three to come up, but, like, our ability
45:40
or our desire to do
45:42
that drive so much of
45:44
what we try and do as a department to
45:46
say, like, Yeah.
45:48
Why why do we build these apps
45:50
to collect information about, like, what we're
45:52
doing in sessions about the time spent
45:54
with athletes doing different modalities? or
45:57
why do we have
45:58
technology that that
45:59
captures this stuff? Or why do we use in match
46:02
data the way that we
46:04
do it? and often it
46:06
comes back to this idea. Right?
46:07
Like, there's so
46:08
many things that we do in sport.
46:10
If you think about that IST, there's things that
46:13
we're going to do to try encourage the the right
46:15
mentality of our athletes. There's things
46:17
that we do
46:18
from a physical preparation and and a
46:21
prevention standpoint. So for
46:23
every one of
46:25
the
46:25
interventions that we think about. Right? Like,
46:27
there's there's some outcome we're trying to change.
46:29
So within within the way that we frame this
46:31
at our our performer strategy. We consider the
46:33
the ninety minutes on
46:34
the pitch. How we play is
46:36
is ultimately what we mean by performer.
46:39
Right? It it doesn't mean what someone gets on any
46:41
test score, some
46:44
isolated
46:44
individual players. And
46:45
it's like, how
46:47
does our team perform for ninety minutes on the pitch over the course of
46:49
the year. That's what we mean, and that's what we're trying
46:51
to optimize. We
46:54
understand
46:54
that that is essentially
46:55
this coming together of
46:58
a cohesive eleven players on the pitch that all
47:00
interact. I want to it can be really challenging to
47:02
do. We can talk about how that team
47:04
performs.
47:05
and then And then we
47:06
can think about each athlete that has this kind of profile
47:08
that we think about into
47:10
like a physical, technical, tax
47:14
and mental domains of how this
47:16
app
47:16
is coming into that team environment. And
47:19
we think within each of
47:21
those domains, there's characteristics constructs
47:23
that we we value. Right?
47:24
So if you think about the physical bucket,
47:27
like an athlete's gonna come into that
47:29
with a given aerobic capacity
47:31
and anaerobic power, a top
47:33
speed that they can run an acceleration,
47:35
change of direction, deceleration ability.
47:37
There's all these things, their constructs within
47:39
a within a physical domain and
47:41
we think that we
47:42
have interventions, that we can
47:45
change those things. Again,
47:46
it doesn't help to change just this
47:49
isolated physical construct for the sake
47:50
of the physical construct. We have to identify that that physical construct
47:53
is important for this player because
47:55
what they're expected to do within
47:57
our overall team strategy is
47:59
to do
47:59
x. During the ninety minutes, our team
48:02
strategy, our game model demands
48:04
this player do this
48:05
in this
48:07
position. So
48:08
therefore, this player is
48:10
unable to do it because maybe they are
48:12
simply not fast enough to recover in
48:14
defensive transition or this player simply isn't fitting
48:16
up like, ninety minutes in this position that that demands within our game model
48:18
are are really demand. So
48:20
then you identify
48:21
what the construct is that you
48:22
want to change, and then you develop the
48:26
intervention. So in some cases, that might be a a top speed. In some
48:28
cases, it might be an aerobic capacity. In some cases, it might
48:30
not be a physical thing at all. It might be
48:34
that this player needs additional technical training
48:36
because this action is
48:37
missing from their game. So we might they might
48:39
be in the right position. They might do the right
48:41
thing. They might be fit enough to
48:43
do that for a hundred and eighty minutes consecutive.
48:45
But, hey, when we get to this moment, they
48:47
are unable to
48:49
consistently perform this action. So what they
48:51
need is a dedicated technical
48:53
work again, the reason they need
48:55
that is because in our game model, at
48:57
the team level, that's
48:58
what it
48:59
requires. So that's really what the
49:01
performance strategy is is about is trying to understand
49:03
all the way from saying, are we adhering to the principles
49:06
of play? What does
49:09
that principle with blade demand from
49:11
each of the positions, then what are the different constructs
49:13
that we think in each bucket that
49:15
relate to the athlete's ability to
49:17
do that on pitch? And then
49:19
we have to think one step
49:21
lowercase,
49:21
can we intervene and change these
49:23
things? Like,
49:24
an aerobic capacity is a more
49:26
malleable thing. than a max velocity, especially in a
49:29
short term thing. It's just easier to
49:31
change. And
49:31
they're same thing on the technical or
49:33
on the practical. components
49:35
like these things might be really easy to change
49:37
and they might take a little bit of time. They might be
49:39
harder to change and take
49:41
longer to get to. But
49:43
if you're a data science department, you
49:45
have to be able to
49:47
say, did
49:48
we intervene? How
49:49
much did we intervene? And what
49:52
measure, if any, do we have, to latch
49:54
onto this construct
49:56
to say, did this change? Like, did this
49:58
get better? Because first, you have to
49:59
say, like, Did
50:00
they get better? Did they get they get at crossing? Did they get better
50:03
at this thing? And then you can ask, okay,
50:05
now are they
50:05
doing that in
50:08
a game? at speed, with defensive pressure, and is that
50:10
translating to our team being
50:12
better? And and that's a fascinating process. It's
50:13
really fun to talk to you radically, but
50:16
obviously that
50:17
it a long time
50:18
to build the ability to do that both
50:20
at a really discreet level
50:22
and at a macro scale. macro
50:24
scale so that So that's it's
50:27
a challenge and it's everything from
50:29
the small component parts all the way to
50:31
the the integrated piece. I don't think we're there yet.
50:33
I think that we're we have nice incidents and
50:36
nice examples that we can we
50:37
can highlight
50:38
along the way and we're getting in that direction.
50:40
But a lot of, as you can imagine, the
50:42
first couple years in setting up
50:44
a data science department is like, Do we
50:47
have data from the matches that we can use to say whether
50:49
a player is or isn't doing
50:52
this? What's
50:54
good? for this and what's necessary. And, oh, we don't have information
50:56
about this. So what
50:58
do we need to do to get? And then once
51:00
you start
51:01
building up components that
51:02
kind of get to both the intervention
51:04
and the performance, you can start to answer those
51:06
questions down the line to say, hey, we've
51:09
really consistently been successful
51:11
here and hey, this seems to be a bit more
51:13
challenging to to
51:14
get after and change.
51:16
Really interesting point.
51:17
I'm gonna use deceleration
51:19
as an example. So we we identify
51:21
that deceleration is an issue because
51:24
it's linked to this
51:28
imaginary athletes change
51:30
direction ability that we've identified in
51:32
game. That a code does come from a
51:34
coach. No. That probably would,
51:36
but it has this time. So
51:38
we go, okay, we need to measure it.
51:39
So we measure
51:41
it with a
51:44
suitable declaration
51:44
oregon test. We isolate it. We
51:46
go, okay. This person is here on the
51:49
scale compared to the rest of
51:51
the team. That's
51:52
fine. But can number on that.
51:54
But the difficulty then comes to
51:56
go, have we improved it
51:59
in game? and
52:01
we can count them. And we can say, are they doing
52:03
more than they have done? The deceleration intensity is
52:05
higher based on this
52:08
number and and whatever. We can
52:10
link that back to
52:11
video. But then how
52:13
do we still bring these
52:16
physical constructs into the
52:18
game and go that's
52:19
got better. We can always look at the training
52:21
and and the testing and say it's got better.
52:23
But how are you going about answering
52:25
that question? Okay. that's
52:27
got better. But did it does it actually
52:29
now happen in game? And
52:31
I don't wanna
52:32
take you down the path that you don't wanna go in terms of
52:34
saying things that you do in club that you don't wanna
52:36
talk about, but why is that
52:38
process to get there? And
52:43
Yeah. How did you go how did
52:45
you go about doing that? Did you come
52:47
back to subjective? because it come back
52:49
all the way to a subjective Yeah.
52:51
I think
52:52
that's a it's a really nice segue into something
52:54
that I know we're probably gonna get into on that
52:56
subjective space.
52:57
So I think for deceleration. Let's stay
53:00
on your example. Right? So when
53:02
you say, like, we've
53:04
changed
53:05
objectively this this
53:08
like, set environment deceleration
53:10
capacity of this athlete. We've
53:12
we've identified that, hey, actually,
53:14
they're middle of the road, they're deceleration really. Maybe we've
53:16
even worked on it and brought it to average. So they're
53:18
they're an average player within our environment
53:21
for for deceleration
53:22
the salaries the capacity
53:24
capacity. The question
53:25
becomes in
53:28
a match. So
53:28
there's two ways that you consider and we'd probably
53:30
want to look at both. One is
53:33
that deceleration
53:34
test might measure their max deceleration ability
53:36
when they're fresh, fit, whatever in
53:39
a close environment. It could be
53:41
that in a match,
53:42
the their fatigue
53:43
levels actually the limiting factor. So you'd actually want
53:46
to combine that with some other measures of
53:48
of different fitness or different capacity
53:50
to say, actually,
53:51
when we use the tracking or the GPS data, seems like their deceleration
53:53
intensity does decrease and drop throughout a
53:55
match. Like, that is one theoretical
53:57
scenario that
53:59
maybe it they don't decelerate properly, but the reason they don't decelerate
54:02
properly is actually some other physical
54:04
limitation. That being fitness is
54:05
is one
54:06
key marker. But
54:09
The other question could be, the coach
54:11
is saying
54:12
this athlete isn't decelerating properly,
54:14
and it's not actually their deceleration ability.
54:16
It's something around their tactical awareness
54:19
of knowing when to decelerate.
54:21
Or in this moment, they're just they're getting
54:24
distracted by the fact that someone else is making a
54:26
run-in behind them and there's a player with
54:28
the ball they're closing down and not
54:30
a physical deficiency.
54:31
It's a tactical deficiency. In
54:33
which case,
54:35
that gets
54:35
harder. So some of these constructs are really nice
54:38
and easy, the max velocity one minus
54:40
the noisy tracking or
54:41
GPS data. Is this a
54:43
nice everyone understands what you
54:44
mean when you say, hey, this player is
54:46
really fast at the top end speed and this
54:48
player slow. Like, we get that.
54:52
Is this player effective pressing?
54:54
Like, closing down opponents when they're
54:56
pressing to guide the shape of play? Or
54:58
something like that as a construct is
55:02
much more challenging answer. And that type of construct
55:04
is where I think the subjective
55:06
measures become brilliant because
55:08
you you you think
55:10
and I think we all intuitively know that a
55:12
really experienced wise coach and a practitioner
55:14
when they see something. They can say, hey,
55:17
that was a really good well executed performance, and
55:19
that was that was worse. And then from a
55:21
data science department, the
55:22
fascinating question that I have is,
55:24
like, That's awesome.
55:26
I believe that there's absolute gold
55:28
in that brain when you evaluate
55:30
and you assess a performance.
55:32
Now I want that. Like, that needs
55:35
to be in our database so that
55:37
I can use
55:38
that for something. Like, I don't just want it
55:40
to be comment that's made in a locker room, I want that to be something
55:42
that we can use as information.
55:44
And that's
55:45
the big discussion, the
55:47
road, about subjective measures,
55:50
which Yeah. If
55:50
you want to that, but that's fascinating topic that
55:53
I think is is still
55:54
wide open. Yeah.
55:56
And
55:57
I think it isn't it links into
55:59
so many things. I mean, never chat last
55:59
night with with three
56:02
guys on a round table with the guys
56:04
to agility. Exactly
56:06
that. Kinda educated in the in the
56:08
science with the science head. Okay. We
56:10
wanna change something. So
56:12
first, we have to test it. But
56:14
then when it comes back to the training, it
56:16
becomes so chaotic and and try and encourage coaches
56:19
to be happy in
56:21
this chaotic space. So
56:23
then we go, well, how
56:25
how do we
56:26
link that testing them testing numbers into
56:28
this chaos that we're seeing in
56:30
the training environment
56:31
because we've made it that
56:33
way. because it links
56:34
to the game because the game is also
56:36
chaotic. So this, like,
56:39
uncomfortableness.
56:39
And I I was actually last night
56:41
when we were talking fighting in my
56:43
brain to go, loosen up. It doesn't have to be
56:46
this stiff environment
56:48
that we test and we
56:51
intervention
56:51
and then retail and and and the
56:54
kind of comb to comb
56:56
drill versus the chaotic
56:57
activity. So I
57:00
find this this
57:00
messy space really interesting.
57:03
And I
57:03
think as as pro code just get and like
57:05
people like yourself get more
57:08
experienced, they become more comfortable in
57:10
this messy zone.
57:12
Yeah. I
57:13
I think that
57:15
It's in
57:16
between. It's it's kind of back to that or
57:18
you kind of alluded to it earlier. So what's the
57:20
breaking point for that athlete when you get to this
57:22
acceleration change of direction? one like, hey. They're not doing this
57:25
in a game. The first question,
57:26
this kinda gets
57:28
back to, like, how we evaluate these
57:31
interventions. You you have to figure out what
57:33
the intervention point is. So if they if
57:35
they can't decelerate or change
57:38
direction
57:38
around a cone, that they know
57:40
the root. It's just to
57:42
be that much harder to do it
57:44
dynamically with an uncontrolled
57:47
environment with pressure with some contact is gonna be that much
57:49
harder. So first, you have
57:50
to say, hey, they're are they capable of doing
57:52
it in this? And then are they capable of
57:55
doing it in training or with some constraint where you introduce an
57:57
opponent or you add some uncontrolled thing. And then
57:59
you get into, hey,
58:01
are they doing this effectively game,
58:03
but then you run into the, hey, that's a lot harder
58:05
to quantify than
58:07
when they were
58:08
just with the cones and you had the time.
58:11
So you
58:11
have the spectrum from easy to quantify
58:13
to, like,
58:14
actually the relevant point when you
58:16
get to the match. So that's yeah. there's
58:18
some gray in there, but I think it's important to go through those steps because it I
58:21
think if you only evaluate let's say
58:23
we have the perfect subjective measure,
58:26
you know back, we like, oh, this
58:28
player is our worst player at decelerating when they close out opponents
58:30
in a match. It's helpful.
58:32
But if
58:33
you wanna intervene
58:35
to know that
58:36
is. Like, why are they not
58:38
able to close down opponents?
58:40
And
58:41
your intervention point
58:43
could probably be more
58:45
efficient if you identify that limiting
58:47
step and
58:47
then attack that rather than just
58:49
having them practice closing out a bunch of opponents every time
58:51
of the session. Like, there's a way to
58:53
do it. But if it's if it's overall capacity
58:54
fitness thing, just having them do
58:56
close ups might not be the most efficient
58:58
way to change their fitness level.
59:01
If it's a if it's an awareness issue, maybe you
59:03
have to introduce those game moments. If it's
59:05
a, again, a deceleration capacity, maybe you want to
59:07
do some other dedicated. Maybe it's literally that they just
59:09
need to get stronger. and
59:10
then the time opponents could actually
59:13
be better spent
59:13
developing a bit more of a
59:16
robust
59:18
physical base in a wav room set
59:20
that then they can transfer. But if you only assess
59:22
the in mass performance, it's hard to get
59:24
to what the right limiting step is that then
59:26
would
59:26
inform what your intervention could be. So it's
59:28
trust Like,
59:29
I I challenge being assessing the game performance. Yeah. I mean, I
59:31
wanna I wanna continue down
59:32
this track because I think it's an interesting
59:36
one in I'm just thinking of examples using the deceleration. So I'm
59:38
a defender. I'm a tech is coming
59:40
towards me. I wanna go and
59:42
meet them.
59:44
but a coach
59:45
identifies that I'm not
59:47
able to slow down quick and get that quick
59:49
enough and slow down quick
59:52
enough to being a relevant or optimal position to be
59:54
able to defend that
59:55
attacker. But
59:56
for me to do that, that may be nothing
59:58
to do with my ability to
59:59
get there quick and slow
1:00:02
down. That may
1:00:03
be something
1:00:05
mentally
1:00:05
from my point of view to think
1:00:07
I don't wanna
1:00:09
go as as possible
1:00:10
because I think that person's quicker than
1:00:12
me.
1:00:12
Therefore, I'm
1:00:14
going to hold back
1:00:17
to give myself more time to prepare
1:00:19
body position, etcetera, so that person doesn't
1:00:21
whip past me. But how
1:00:22
do we get into
1:00:23
a position where we
1:00:25
are subjectively
1:00:28
evaluating
1:00:28
something like that, so we can go, yes,
1:00:30
it's improving on the pitch. No, it's not
1:00:32
improving on the pitch. Yes. And
1:00:33
and to even follow-up on that, it might
1:00:36
be that literally within the game model,
1:00:38
your coach has communicated that in
1:00:40
this scenario, you do not
1:00:42
go and press. the
1:00:43
attack. In those moments.
1:00:45
Right? So you just like
1:00:47
that is a different that's
1:00:50
a different structure trying to get to.
1:00:52
And I think now we're getting into the space where I think having some subjective
1:00:54
ability to evaluate and
1:00:57
capture becomes absolutely vital
1:01:00
because it it's it's a lot easier to get to some contracts. But if I say,
1:01:02
hey, is this defender adhering
1:01:04
to our game model more
1:01:06
than they
1:01:08
were before? Right now, I would venture against as much as we
1:01:10
try with event or tracking or etcetera data to
1:01:12
try and get to these things. So
1:01:14
probably the most
1:01:15
reliable is to ask
1:01:17
the manager, ask some of the coaching staff
1:01:19
in this
1:01:20
game. How well did
1:01:21
this player adhere to principles or
1:01:23
use video to say? In this example, did they did
1:01:25
they not do what they were post and
1:01:28
use
1:01:28
that. So so
1:01:30
from a data science provider to entrepreneur
1:01:32
and surgeon drug, from a
1:01:34
data science perspective, what are you doing to
1:01:37
help get that information
1:01:38
out of the coach's head and like you
1:01:40
say the analogy that you you
1:01:42
described, put it into our
1:01:45
database. Yeah. Hundred
1:01:46
percent. So that
1:01:48
that's really
1:01:48
the
1:01:49
impetus behind. We we just published we
1:01:52
have a a PhD soon right
1:01:54
now at Keith Hamilton. He's a PhD student with doctor David
1:01:56
Cox, who's an
1:01:58
unbelievable
1:01:58
sports
1:01:59
psychologist. before psychologists He's
1:02:02
had
1:02:02
a very prolific career. We we're very fortunate to have him as
1:02:04
as part of our performance staff
1:02:07
here. Chuck, he's
1:02:08
the applied brilliant
1:02:10
academic grade. He he
1:02:12
has a student, Keith Hamilton, and his PHE is
1:02:14
really trying to get into this space. And we started
1:02:17
with this paper
1:02:18
we just released to say, look, subjective evaluations. I
1:02:21
think one
1:02:22
are everywhere
1:02:22
in sport. We do
1:02:25
this. when select the starting lineup. We do this when
1:02:28
we announce an MVP
1:02:30
award. We do it at the end of every academy
1:02:32
season when we say, okay, this player
1:02:34
is gonna come back next year and this player
1:02:36
isn't those
1:02:37
decisions are based
1:02:40
on the
1:02:42
Like,
1:02:42
the collection
1:02:43
of information in the brains of
1:02:45
our practitioners as they make decisions. And we believe
1:02:47
that there's the coach's eye. We believe that someone that's
1:02:50
really experienced has the
1:02:52
capability of doing
1:02:53
it. And they're just
1:02:54
done all the time. It's on every radio or
1:02:57
every talk
1:02:57
show. It's like this player's awesome and
1:02:59
this player's terrible. what that is is
1:03:02
it's a subjective evaluation based on what you've seen as you assign
1:03:04
this categorical rating
1:03:05
of terrible or excellent
1:03:07
or average
1:03:08
to a player. Now
1:03:10
I would liken
1:03:12
this back to, like, my
1:03:14
time during my PHD was sitting in
1:03:17
like, psychometrics
1:03:18
classes as they talk about how do we develop
1:03:20
stuff to capture empathy,
1:03:22
emotional intelligence, general intelligence, etcetera.
1:03:26
And
1:03:26
how do
1:03:27
we know
1:03:28
what we're collecting? Is getting to this contract
1:03:30
that you can't really measure? And like, tactical
1:03:33
adherence is really hard or
1:03:35
you get into this these spaces of,
1:03:37
like, hey, is this player a
1:03:39
good teammate? Or is this like, you
1:03:41
get to
1:03:41
these things that are harder to measure, and that's what I
1:03:43
would liken it to. And I would say, like, I think
1:03:45
the jury is up. So when
1:03:46
I say there's there's moments where a
1:03:48
survey doesn't do it or an Excel workbook doesn't do
1:03:50
it. And we we've tried to develop things
1:03:54
in house that can capture. I
1:03:55
don't I don't think we
1:03:57
know yet what's best
1:03:58
practice. What I would say
1:03:59
is say is
1:04:01
I've seen
1:04:02
I've seen
1:04:04
surveys ruin subjective evaluation.
1:04:07
I we
1:04:08
don't yet in sport have
1:04:10
what a gold standard. This is what it should
1:04:13
look like. So, like, for
1:04:13
example, I've seen before,
1:04:16
hey, can you rate this
1:04:18
player's performance in this given match
1:04:20
or in this
1:04:22
given competition? Did they
1:04:22
perform better than expectation, at expectation,
1:04:25
worse than expectation. And this was
1:04:27
collected for a long
1:04:27
period
1:04:30
of time. in
1:04:30
a given environment.
1:04:31
And essentially, then what you would have
1:04:33
is you'd have some
1:04:34
of the best players
1:04:37
would
1:04:37
end up over the course of the year
1:04:39
with a degrading performance.
1:04:40
And they
1:04:41
would be, like, kinda, like,
1:04:43
your bottom third of performers. And then
1:04:45
you go in and be like, I thought this was your best player. They would
1:04:47
say, well, he he was. But, like, as because they're so
1:04:49
good, the expectations rise,
1:04:52
and they continue rising over the course of So a given period of
1:04:54
time or given match, their performance is
1:04:57
further below the moving
1:04:58
bar that is the expectations placed
1:05:02
on. and you
1:05:03
might have an average player that performs above expectation. But if you were
1:05:05
using that at the end of the year to say,
1:05:07
okay, which players stay, which players go,
1:05:09
which player should start,
1:05:12
you're
1:05:12
not capturing the construct you're
1:05:14
interested in, which is which of
1:05:16
the players perform the That's
1:05:20
it.
1:05:20
That's what you're trying to get to, but the
1:05:22
survey or the question is
1:05:24
not
1:05:24
capturing. It's capturing something
1:05:26
else which might
1:05:27
be relevant, but it's this
1:05:29
construed ratio of
1:05:32
coaches' evaluation
1:05:32
with player performance that
1:05:35
gets you something else. You
1:05:36
see this this is in
1:05:37
the in the athlete's self
1:05:39
report space as
1:05:40
well. Right? So Aaron Foods has been on. You've
1:05:42
had guys pop well, like, use the
1:05:45
the scales
1:05:45
that have some validity related evidence versus
1:05:48
scales you create on your own. And I have you've seen,
1:05:50
like, sometimes one is good and
1:05:52
five is good, bad, and then sometimes
1:05:54
five is and one is good or
1:05:56
whatever. And vice versa, sometimes it's a one to ten anchor, sometimes it's a zero to
1:05:58
seven, sometimes it's this, sometimes
1:06:00
it's
1:06:02
that And
1:06:02
I've seen data sets before where all of a sudden an athlete goes from, like,
1:06:05
being fantastic to being terrible.
1:06:07
And then terrible
1:06:09
you're, like, what's going on on that day
1:06:11
and you're on top of them box and they're like, oh, no. I've
1:06:13
been misinterpreting this scale completely since I
1:06:16
started recording
1:06:18
data.
1:06:19
They're like, well, that's
1:06:20
it's gone now. So I I've
1:06:23
seen many examples of, like, hey, here's subjective
1:06:25
evaluations. And because of the way
1:06:27
that the survey
1:06:27
or the scale or the question was
1:06:30
designed, you've lost the
1:06:31
ability to capture what you're
1:06:33
trying
1:06:33
to get to.
1:06:35
The question
1:06:35
that we're after is how do you develop
1:06:38
the survey to scale in a
1:06:40
way that
1:06:42
does capture That's what we're working on at the club. I think the jury's out. think can
1:06:44
learn a lot from the psychometric space. We're
1:06:46
partnering with Bruno. It
1:06:48
was one of
1:06:50
the authors on that paper that we now
1:06:52
to try and at least get a framework to
1:06:54
think about it because there there's actually
1:06:56
a lot of similarities with
1:06:59
objective data when you think about how data
1:07:01
is generated and how it can actually
1:07:04
help
1:07:04
inform an
1:07:05
inference. There's a lot of
1:07:07
crossover. But what we have and I think we
1:07:09
said this in the paper was in
1:07:12
the
1:07:12
objective space. You have this data collection process.
1:07:14
Right? The data is generated.
1:07:16
We
1:07:16
run a sprint. We do a fitness test. Right? That athlete
1:07:18
runs on a treadmill. Right?
1:07:20
To
1:07:21
then capture
1:07:22
that
1:07:24
information. It's the calibration side. Which case is, like, is the is
1:07:26
the treadmill calibrated? Does the mask
1:07:28
fit on the mouth? What is are
1:07:31
we doing the
1:07:31
Bruce protocol or something
1:07:33
else. And we've spent so much time in exercise
1:07:36
science talking about the
1:07:38
calibration of
1:07:39
objective tests. Then there's
1:07:40
the inference, which I think can't
1:07:43
forget, and that's a a lot of the unified validity theory, psychometrics talking,
1:07:45
which is another giant conversation.
1:07:47
We'll leave that. then
1:07:49
you have to make an inference.
1:07:50
This VO2 max or this player
1:07:52
is from
1:07:53
this very calibrated
1:07:56
bike, whatever fifty two, then
1:07:57
you have to make an difference. This player
1:07:58
is, isn't fit enough, needs
1:07:59
work, we're gonna use that to do something.
1:08:02
It's not just about the score, it's about what you're
1:08:04
trying to do
1:08:06
with it. and then you
1:08:07
make an inference. The subjective thing is the same thing. Right?
1:08:08
So you you watch a game back to
1:08:10
your example. The defender does
1:08:12
or does not close
1:08:14
out. on the attacking plan.
1:08:16
Then maybe you have this,
1:08:17
whatever, the
1:08:18
survey that asked the coach. In
1:08:20
this environment, how
1:08:23
effective was this defender? at
1:08:25
doing what you've asked.
1:08:27
Data is then captured in that
1:08:28
survey, in that
1:08:30
question, in that whatever, That's
1:08:34
the calibration. And then there's an inference. We're gonna compare that with all the other times. The defender did or didn't or in
1:08:36
the
1:08:36
situation that say they are getting
1:08:38
better or are not getting better.
1:08:43
In the sports
1:08:43
science space, we've spent like
1:08:46
so much time calibrating bikes
1:08:47
and
1:08:48
building protocols, developing field
1:08:50
tests, and lab tests, and
1:08:52
saying, what are all the other things
1:08:55
that go
1:08:55
calibrating so can trust test? And in the subjective
1:08:57
and then in the subject of space space,
1:09:00
we skip
1:09:00
that for the most part. It's like stuff is stuff
1:09:02
happens. People play games. We look at their performance, and then
1:09:04
there's a lot of hot takes. Like,
1:09:07
this player is terrible. They're getting
1:09:10
worse every game or this player is great.
1:09:12
That was a brilliant performance.
1:09:14
This was above
1:09:15
average. And often,
1:09:16
we don't capture when we capture it,
1:09:19
even in the academic literature, there's
1:09:20
almost no scales
1:09:23
that have had dedicated effort
1:09:25
to saying, does this scale itself have
1:09:27
any evidence related to its ability to capture
1:09:29
someone's technical or tactical performance. And that's
1:09:32
the space we're living in
1:09:34
now. It's like, we're trying to borrow best practices from
1:09:36
psychometrics, having these discussions, and
1:09:38
saying, what do we know
1:09:40
from survey design and scale
1:09:41
design and many things that we
1:09:43
can apply here But I
1:09:45
think
1:09:45
best practices is up for discussion, up for debate, which makes it a really
1:09:47
exciting space, but
1:09:48
for discussion upper debate which makes it really
1:09:50
exciting space for that are we that's how
1:09:52
we at
1:09:52
least how we try and answer
1:09:54
the question as we build tools. I think it's
1:09:55
fascinating. I think it's
1:09:57
a fascinating area
1:09:59
and
1:09:59
something that just the more
1:10:02
you talk about, the more you kinda get level down, level down, level down, then I go
1:10:05
my mind just
1:10:07
goes, oh, no. Yeah.
1:10:09
So this is so deep, but no, it's tackling it. So good work. Good work. I've
1:10:11
kept you for an hour already, but we
1:10:14
still have one more I
1:10:18
think important point to cover. And
1:10:20
I think it's I'll I'll reference
1:10:22
your article that you kindly wrote
1:10:24
for Sportsmith. which has done very well. So
1:10:27
clearly, this is a point that people want more information on and it's data visualization. And I think
1:10:29
if we can keep it to a
1:10:31
couple of minutes, that great
1:10:34
and just get the I hate
1:10:37
to say it, the
1:10:39
hot takes. But your
1:10:41
principles, when it
1:10:42
comes to debt visualization. I know you some a lot
1:10:45
in the in the
1:10:46
article itself, but what what
1:10:48
do you
1:10:49
live by when it comes to visualizations?
1:10:51
And how do you evaluate whether that's been
1:10:53
successful or not?
1:10:54
Few minutes. Mhmm. Few minutes.
1:10:57
I'll try to keep it as
1:10:59
brief as I can. I
1:11:01
think taking
1:11:01
care in what you're doing by taking
1:11:03
private is is one
1:11:06
of the the overarching things.
1:11:09
you're using this again as one
1:11:11
end of the data pipeline. So your ability
1:11:13
to communicate all the back end
1:11:15
work in terms of, like, collecting
1:11:19
aggregating,
1:11:19
analyzing, building the thing,
1:11:21
all ends up coming down to your
1:11:23
ability to communicate that.
1:11:26
So
1:11:26
taking pride in that and just realizing you're
1:11:29
hoping that
1:11:30
you can
1:11:32
convey a
1:11:32
message effectively to the end user.
1:11:35
They're more likely to look if
1:11:37
it looks nice. That's just the reality. Like, you it
1:11:39
it's just they're more likely to look longer, and
1:11:40
the longer they left, but
1:11:42
hopefully, more likely they are to
1:11:45
take
1:11:46
on the information that you're trying to
1:11:48
convey. So there's this
1:11:50
base level where you have
1:11:52
to
1:11:52
reach before someone goes, I
1:11:54
think I said
1:11:55
in the article, if not, it's been in
1:11:57
hard conversations. Like, it's
1:11:58
it's efficiently
1:11:59
beautiful. So you
1:12:02
you have to We would describe our
1:12:03
structure. So the number of dashboards or reports are
1:12:06
things we have to deliver is is pretty high.
1:12:08
So
1:12:09
if we spend
1:12:11
every last minute kinda fine
1:12:13
tuning to the nth degree, like, we will run out of
1:12:15
time in the day to do that. So
1:12:18
the that gets into
1:12:19
just having templates in place. So having color schemes, having
1:12:22
general font recommendations, having these general
1:12:24
templates for
1:12:26
how
1:12:26
we build. gets you eighty percent of the way there. It takes no extra time to
1:12:29
use the default g g plot, whatever, or
1:12:31
Tableau out of the box. It takes
1:12:33
a little
1:12:33
bit of upfront work to
1:12:36
build templates you can build on that gets you eighty percent
1:12:38
of the way. And
1:12:38
then it gets to that, like, at the right time, hey, this
1:12:40
visual should
1:12:41
be fine too, and then you can take
1:12:43
that any expert twenty percent. And
1:12:45
then I think the other
1:12:46
principle that I I hope we all continue going
1:12:49
with, and this
1:12:52
is like when I write an
1:12:54
article like that, the impostor syndrome is real. The note at the end of the article, like, I think the biggest principle I try and live by constantly
1:12:56
learning. I think
1:12:59
John Polley would love
1:13:00
hollywood love to
1:13:02
say, and the joke around here is I'm
1:13:04
not even the best date of this person
1:13:06
in the room I'm
1:13:07
in. So, like, Luc PD arrived and
1:13:09
produced a better looking foreplay dashboard on, like,
1:13:11
week number two of MVP. No. So I hope that I can
1:13:13
at least continue to learn and grow and
1:13:15
get better. So it's
1:13:19
when it's done brilliantly, like, I think we all know what
1:13:21
that looks like. You know? Like, when you look at
1:13:23
Cedric here, create
1:13:26
a
1:13:26
plot in g two plot two? Or
1:13:28
you even see, like, I
1:13:30
think, right now of the inspirational
1:13:32
view
1:13:33
of, like, how data can
1:13:35
inform and, like, in this quote, inspire. But,
1:13:36
like, capture people. If you
1:13:39
think about John Murdoch's, like,
1:13:41
like plots
1:13:42
on COVID incidence rates across
1:13:44
the world during the pandemic. Like,
1:13:46
you've never seen so many people interested
1:13:49
in exponential growth rates.
1:13:51
Just
1:13:51
able to turn What's this
1:13:53
global pandemic and a ton of
1:13:55
information into a really digestible form to talk about whatever it'd be,
1:13:57
like cases.
1:13:59
But that's or increasing that, like, all those things, and
1:14:00
it's done in a way that
1:14:02
just, like, captures people and
1:14:04
helps them to
1:14:07
understand what's going on. And
1:14:08
then from, like, the
1:14:09
beautiful side, we we published
1:14:10
a a preprint on this this abstract thing. I
1:14:12
called it. A lot of conversations that Dave
1:14:14
and I had at USOC, like, Well
1:14:18
done data
1:14:19
visualization can be such a great communication medium, and
1:14:22
and
1:14:23
science often offers from, like, a knowledge
1:14:26
translation problem in some instances. So can we capture that? So we tried to take a dataset
1:14:28
and communicate it in a
1:14:30
in
1:14:30
a visualization. So I had
1:14:34
phone
1:14:34
call with Lisa Crescribe. She won
1:14:36
IronBiz, which is like
1:14:38
Tableau's
1:14:39
data visualization contest, which is
1:14:41
of
1:14:41
sight to behold. I was at a
1:14:43
New Orleans, so there's, like, ten thousand data science, data analyst, data biz
1:14:45
people in a room and
1:14:46
people get twenty minutes to build
1:14:50
their
1:14:50
best looking visual. And the stuff that people
1:14:52
can build in front of me is absolutely
1:14:54
fascinating. So Lisa, actually, won, I am Biz.
1:14:57
I reached out and said, I'm trying to
1:14:59
developing. Can we
1:14:59
turn a research, abstract research paper into a really
1:15:01
nice news? If we published this using
1:15:04
player
1:15:04
maker data that
1:15:06
was publicly available, Steve Barrett,
1:15:08
Chris
1:15:09
Allison, they had this publicly available data. So we took it and we created a Tableau vision. You can see you
1:15:11
take one look at it and you'll probably say there's no way
1:15:13
Johann actually produced
1:15:14
this. So I had to as
1:15:19
I did a very small part. Lisa did a lot of the heavy lifting. You'd see
1:15:21
how beautiful it is. So it's like, as a
1:15:23
principal, I hope
1:15:24
I, our team and people that
1:15:25
are in this space can be liberated
1:15:27
to say, look, It's another
1:15:29
one
1:15:29
of those moving partners. You're never gonna be arrived and say, like, I am now, like, the
1:15:32
perfected date of biz artists. Like, there's
1:15:34
so many examples of people doing it
1:15:36
well. so
1:15:39
many different ways to do it well that you can continue to learn and
1:15:41
evolve and then think what looks what's
1:15:43
perfect in one instance
1:15:45
for one user is gonna be different than another.
1:15:47
I kinda end on that.
1:15:47
Like, it goes back to idea of, like, a video
1:15:50
can be more effective than
1:15:52
the
1:15:54
best pristine Tableau dashboard that's as
1:15:56
interactive and as sophisticated or as simple
1:15:59
as you want.
1:15:59
Because it comes down to can
1:16:01
you take the information on board and
1:16:03
change the decision you're making.
1:16:05
For a player, for a coach, for different practitioners, sometimes it's it's
1:16:07
literally a a video
1:16:09
and say,
1:16:11
do this more.
1:16:13
do this less, do this differently or a picture of
1:16:15
a pitch and say, we
1:16:15
want you to be here rather
1:16:18
than, hey,
1:16:19
here's this trend
1:16:21
line over time that's contextualized to the other assets. So I'll
1:16:23
end there. I'm missing principles,
1:16:24
contextualize
1:16:28
the data. be
1:16:30
honest with your reporting. There's a lot there, but
1:16:32
you told me to keep it brief and I'm already over,
1:16:34
so I'm gonna stop with with
1:16:35
that. That's fine. That's that's loads
1:16:38
of good loads of good stuff there. And I think
1:16:40
people can people can read the article, and I'll link it
1:16:41
in the show notes and on iTunes, it's WiFi, and YouTube, all that
1:16:43
kind of stuff. where
1:16:46
people are listening so people can have a little look and and dive into it. But Johan, we've
1:16:48
been on the
1:16:49
phone for an
1:16:50
hour and twenty five minutes. an
1:16:55
hour
1:16:55
and ten has been recorded. So I really appreciate
1:16:57
your time. I really appreciate you getting
1:16:59
up early. And as
1:17:01
the Is that big coffee gum? That big coffee
1:17:03
gum has a big time to go
1:17:06
to the water. Nice. Nice. Well, thank
1:17:08
you very much. I really do appreciate it.
1:17:10
But last but not least, where can people get know
1:17:12
more about you, where can people keep up to
1:17:14
date with projects you got going on, research
1:17:18
gay, all that kind of stuff? You'll probably
1:17:20
be sadly
1:17:20
disappointed if you follow me on social
1:17:22
media because something will come on
1:17:24
once every six months or
1:17:27
something along those lines.
1:17:29
Yeah.
1:17:29
I guess there's probably
1:17:30
the question on the on the academic space, like the research gate
1:17:34
or the whatever the
1:17:37
the
1:17:37
Google scholars or whatever the people wanna see. Like, if I if I'm
1:17:39
fortunate enough to push out a publication, then those are
1:17:41
gonna be the avenues
1:17:43
to find them.
1:17:45
in
1:17:45
the applied space. If I say
1:17:48
anything publicly, it's probably on Twitter. But again, it's
1:17:50
probably
1:17:50
with disappointing frequency. When I say anything, but also,
1:17:52
like,
1:17:55
if you if people really wanna connect,
1:17:56
like, I'm I'm happy to have a chat, always happy
1:17:58
to to hop on a call or
1:17:59
send
1:17:59
some some messages.
1:18:02
So that'd be direct message on Twitter, do the LinkedIn
1:18:04
thing or whatever. Like, it's it's always fun
1:18:06
to connect
1:18:06
with people in this space. So those
1:18:09
are all options
1:18:11
in avenues.
1:18:12
Sweet. Right? I'm gonna
1:18:13
find and let you go and get on with some work.
1:18:15
But I really appreciate your time. It's great to have finally the chat and great to get
1:18:17
you on the podcast and, yeah, really appreciate
1:18:19
your help and speak soon.
1:18:22
Awesome. Thanks
1:18:22
for having me, Rob. Yeah. Take your time. Thank
1:18:25
you. Yeah. Thanks
1:18:27
to your ninja episode
1:18:29
four hundred and fifteen. of the Pacer performance
1:18:31
podcast. Really appreciate you tuning in. Hope you got
1:18:33
as much out of this episode as I
1:18:36
did. Big thanks to
1:18:38
Johan for giving up his time. very busy guy getting to the office early,
1:18:40
so it fits in with with
1:18:42
everyone's kids, bedtimes, and all that
1:18:45
kind of things. I really appreciate his time. Big
1:18:47
thanks to Hocking Dynamics, Fusion Sport, Omega
1:18:50
Wave, to Tensor College,
1:18:52
and Hydro
1:18:54
Fast Response and happy so today. The podcast could not run its
1:18:56
code following out these guys, so I really
1:18:58
do appreciate all their support. Big thanks to
1:19:01
you've tuning in and look forward to
1:19:03
chatting to you next time.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More