September 20, 2025 • 35 mins
LALIGA's Data Revolution: How Sportian Performance Uses AI and 3.5 Million Data Points to Win European Trophies
We offer a comprehensive view of how Artificial Intelligence (AI) and data analytics are transforming professional football, particularly within Spain's LaLiga. We focus on LaLiga's internal technological evolution, announcing that its core data platform, Mediacoach, is being rebranded as Sportian Performance through a partnership with Globant and Microsoft, emphasizing its role in generating millions of data points per match for all clubs. Parallel academic and industry analyses explore the benefits of AI in developing tactics, optimizing player performance through predictive analytics, and enhancing broadcasting and fan engagement. Critically, we also address the significant challenges and ethical considerations inherent in adopting AI in football, including concerns over data quality, implementation costs, real-time data limitations, and the fundamental need to balance complex AI insights with crucial human judgment and player privacy.
#TechneAfricaFutbol #TechneFutbol #TrainWithTechne #Playermaker #TraceFutbol #EPL #FootballTech #FootballInnovation #PlayerDevelopment #CoachTools #SmartFootballTraining #AfricanFootball #CAFOnline #FIFA #FIFAYouth #CAFDevelopment #YouthFootballAfrica #NextGenFootball #FootballScouting #TalentIdentification #DigitalScouting #EliteYouthDevelopment #FootballJourney #FootballAnalytics #FIFAForward #FutebolAfricano #TrainTrackCompete #FootballExcellence #FootballAfrica #TheFutureOfFootball #Laliga #MLS #Liverpool #ManUtd #SportsDirector #SoccerCoach #SoccerAcademy #Barcleona
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
Welcome everyone to a really exciting deep dive.
Today we're pulling back the curtain on something truly
fascinating. Forget the itch side drama for
justice a moment because honestly the biggest battles in
football right now they might just be happening inside server
rooms. That's so true.
We're talking about this incredible fusion, this mix of
(00:20):
high level science, massive amounts of data and you know,
the beautiful, unpredictable chaos.
It is football. Exactly.
And we're zooming in on Spanish football, specifically a Liga,
to figure out how they didn't just dabble in data, they built
an entire system around it. And the result?
Well, let's just say it involvesa lot of silverware.
A lot of silverware is right. It's quite a story of and
(00:42):
commitment that the platform they built has been a
cornerstone for Spanish teams, leading to this frankly
unprecedented era of success. We're talking 16 European titles
since its implementation. 16 That's not just luck, is it that
strategy, that's technology making a tangible difference on
the European stage. Absolutely, and that success
really points to the power of the engine behind it all.
(01:04):
For years, many of you will haveknown this system as media
coach. It's been the standard since
what, the 2010 eleven season, analyzing every single movement.
Right Media coach has been the name synonymous with allergy
data. But things are evolving, aren't
they? There's a big change coming.
There is. It's a significant transition
starting July 15th, 2025. Media coach is becoming Sportion
(01:26):
Performance. Now.
This isn't just slapping a new label on it.
It's much bigger than that. OK, tell us more.
What's behind the name change? It signifies a major strategic
alliance. Allegia has partnered with
Sportion, which is the dedicatedsports technology division of
the global giant Globin. So Sportion performance reflects
this new, deeper integration anda commitment to, well,
(01:48):
continuous innovation across theentire league.
So, Sportion Globent, this sounds like a powerhouse
partnership aimed at pushing theboundaries even further.
Precisely. The goal isn't just analysis
anymore. It's about driving this constant
technological evolution league wide.
It's about institutionalizing that competitive edge through
tech. And to really understand how
(02:09):
they do that, you've got to lookunder the hood right at the
technological foundation. Let's let's impact this portion
performance ecosystem. Yeah.
The ecosystem strength really comes from that three-way
partnership. La Liga, Band, Sportion and
crucially, Microsoft providing the cloud infrastructure and AI
capabilities it's set up designed for constant growth.
(02:30):
And who's steering the ship within La Liga?
That responsibility falls to La Liga's newly named Football
Intelligence and Performance Unit.
Think of them as the internal hub, the brain trust that
overseas the whole strategy. They support all 42 professional
clubs, Loviga, EA Sports and La Liga Hypermotion. 42 clubs
getting this level of support and they're doing training
(02:51):
workshops. Constantly.
It's about ensuring every club, regardless of size, can leverage
these tools effectively. They're integrating new features
all the times, keeping everyone at the cutting edge.
OK, You mentioned continuous evolution, and for that you need
fuel. Data, fuel.
The sheer volume of data processed here must be
astronomical. Oh, it is.
The scale is almost hard to comprehend sometimes.
(03:12):
Let's put some numbers on it because that really tells the
story. Hit us with it.
For every single match played inElega, the platform captures
over 3.5 million data points. Three and a half million per
game. Law, game and that raw data is
then processed, analyzed, crunched, eventually generating
more than 6400 unique metrics that are made available to the
(03:35):
clubs. 6400 different ways to look at a match.
Exactly. And if you zoom out across an
entire season, this system turnsout over 112,000 analytical
reports. It's just a fire hose of
information. 112,000 reports. How on earth do they capture
that initial 3.5 million data points with the speed and
accuracy needed for it to be useful?
(03:56):
That feels like the core challenge it.
Is, and the answer lies in some pretty sophisticated tech
installed in every stadium. It's primarily an optical
tracking system that use a combination of cameras.
OK. What kind of cameras?
So you've got three main tactical cameras sort of
following the general play like you might see on a broadcast
feed, but position higher, wider, but the real precision
(04:16):
comes from up to 16 high resolution fixed perimeter
cameras. 16 fixed cameras dottedaround the stadium.
That's right, these are the workhorses.
They're constantly tracking the position of every player on the
pitch, both teams, plus the referees and of course, the ball
itself. And how frequently are they
capturing this positional data? This is key. 25 times per
(04:38):
second. That's 25 Hertz.
That frequency gives you incredibly granular data.
You're not just seeing where a player was, you're seeing their
exact trajectory, their micro movements, acceleration,
deceleration. 25 times every second. 25 times a second.
OK, now I understand how you getto 3.5 million data points.
That level of detail lets you analyze things like body
(04:59):
orientation, angles of runs, theexact distance between players,
things you just can't see reliably with the naked eye.
Precisely, it turns simple positioning into a rich data set
for advanced analysis. Now historically, all this
incredibly rich raw data, it wasmostly kept within the Leaguers
own analysis teams, right? The club's got the finished
(05:21):
reports the output, but not necessarily the underlying raw
numbers. Exactly.
But that's undergone a massive shift recently with something
called the Media Coach Sandbox. The sandbox.
OK, what's revolutionary about that?
Ricardo Resta, the director of Media Coach, calls it a game
changer. He does, and for good reason,
think, feel like this. For years clubs were getting,
(05:41):
you know, the beautifully Bay take the final analysis report.
But increasingly they were saying, hey, we appreciate the
cake, but can we get the flour, the eggs, the sugar?
We want the raw ingredients. I like that analogy.
They wanted to do their own baking, test their own recipes
based on their specific coachingphilosophy.
You've got it. The sandbox provides exactly
that. It's an advanced analytics
(06:01):
environment running in a secure data bricks workspace where
clubs get direct access to the raw data.
And raw data means. We're talking the fundamental
stuff. Individual player coordinates,
XY 25 times a second. Collective positioning data,
specific angles of movement, theteam shape metrics like the
(06:21):
convex hull. That's the area the team
occupies, Distances covered at different speeds, all the
granular detail. OK.
So they have the raw ingredients.
Why is that such a big deal beyond just having more numbers?
Because it unlocks true customization and automation,
often using AI. Imagine a coach is obsessed with
a very specific type of defensive press.
(06:43):
Maybe they don't just care abouttotal pressures.
Right, they want something more specific to their system.
Exactly. Maybe they want to measure, say,
how quickly do my 2 central midfielders close the space
between them after we lose the ball in the opponent's half.
That's a bespoke metric. Something you wouldn't find in a
standard report. Probably not, but now in the
sandbox the club's own analyst can use the raw positioning
(07:04):
data, maybe apply some AI, and build an automated process to
track exactly that metric tailored 100% to their coaches
unique tactical ideas. It bridges the gap between
generic data and specific coaching needs.
That makes perfect sense. It turns the platform from a
provider of information into a toolkit for generating unique
insights. And that leads us perfectly into
(07:25):
the next phase. How do these tools actually land
with the coaches? Let's move into Section 2
products and, crucially, the language of the coach.
Yeah, this is so important because you can have the most
powerful analytics engine in theworld, but if the coach doesn't
trust it or doesn't understand it or doesn't use it, it's
(07:46):
basically worthless. It stays in the analyst's
office. Exactly.
And La Liga seems very aware of this.
The philosophy, particularly championed by Roberto Lopez, who
coordinates a Ligas sports research department, is crystal
clear. He stresses that the technology
is a supplement, never a replacement for the coach.
So it's about decision support, not decision making.
Precisely. Lopez's phrase is that the goal
(08:08):
is to help coach make fewer mistakes, to provide objective
information, to challenge assumptions or confirm
observations. But ultimately, the final word
always belongs to the coach. Data informs intuition, it
doesn't override it. That cultural sensitivity feels
key to why it's been adopted so widely.
OK, so let's breakdown the actual suit of products and
(08:29):
services they offer. How does it work in practice?
Maybe starting with the heat of the moment during the match
itself. Right.
So during the game the primary tools are the live NRT near real
time tools. These are typically delivered
pitch side, often on a tablet ormaybe a digital whiteboard in
the dugout area. Near real time, so not quite
instantaneous. There's a slight delay, yeah,
(08:50):
and you have to accept that the data quality and quantity in NRT
might be a little bit lower thanwhat you get post match.
Processing millions of data points instantly and validating
them is incredibly complex. OK, so if the NRT data isn't
perfect, what's its main job during a game?
It's about providing immediate critical alerts for decision
support. The source has got a great
(09:11):
example, high injury risk alerts.
How does that work? The system is constantly
tracking physical metrics like high intensity meters run Sprint
distance, number of accelerations for every player.
Let's see if a star player absolutely crucial.
It's the 70th minute an alert might pop up on the analyst
tablet. Really to the bench warning.
Player 10, has exceeded 90% of his typical high intensity
(09:33):
distance for this stage of the game based on his seasonal load.
Yeah. So it's not saying he's injured,
it's saying objectively he's entering a known risk zone for
potential muscle fatigue or injury.
Exactly that. It doesn't tell the coach sub
him off. It presents objective data.
Here's a calculated risk factor.Now the coach has to make a
judgement call, right? Is the game situation critical
(09:56):
enough to push him, or is it safer to protect him for the
next match? That's the dilemma.
The NRT tool provides the objective data point.
The coach provides the experience, the context, the
human judgement. It's a perfect blend.
That's a really smart application of NRT.
OK, so the final whistle blows, the immediate pressure is off.
What happens next? Well, almost immediately after
(10:17):
the match, often literally on the team bus or maybe on the way
at a press conference, the coaching staff might use the
mobile application. The mobile app what for quick
look UPS. Yeah, quick reviews.
It serves a couple of purposes. 1 is just immediate internal
debriefing, but another, perhapsmore tactical use is preparing
for the media scrutiny. Like a defensive tool against
(10:39):
criticism. You could definitely see it that
way. Imagine the team just lost 10
despite playing OK. The narrative might be they look
tired or they just didn't show up in the second-half.
Right, the easy narratives. So the coach or an assistant can
quickly pull up the mobile app and find objective data.
Actually, look, our total distance covered was above
average. Our number of penalty box
entries was our third highest this season.
(11:02):
It allows them to immediately counter subjective criticism
with objective facts. Provides a data shield.
I like that. OK so NRT for alerts, mobile for
quick review and press conference prep.
But for the really deep post match dissection, where does the
series analysis happen? For that, Roberto Lopez points
to one tool as his personal favorite, and it seems to be the
(11:23):
core of the analytical workflow,the desktop visualizer, which is
called vision. Vision Why is this one the gold
standard for the analysts? Because Vision is where the
complex data finally starts speaking the coach's native
language. Yeah.
And what is that language? Video.
Video. Absolutely.
Coaches understand the game through watching it.
Vision's power is that it embedsall that granular tracking data,
(11:46):
the 3.5 million points per game,directly onto the corresponding
video footage. So you're not looking at
spreadsheets or abstract charts,you're watching the game, but
with layers of data visualized on top.
Exactly. You might see player movement
trails, distances between players displayed in real time,
shaded zones indicating controlled space, maybe even
probability metrics popping up during key moments, all
(12:08):
synchronized perfectly with the video.
I can see why that's powerful. It bridges the abstract data
world with the concrete visual world of football.
It's critical you can show a coach the most elegant Power BI
dashboard full of complex metrics, and honestly, their
eyes might just glaze over. It's not their language.
Show them three curated video clips using vision, highlighting
a specific tactical pattern withthe data overlaid.
(12:31):
Instant understanding, instant engagement.
Video is the universal translator, and Vision
apparently takes this further with deep customization.
They talk about achieving the quadrature of the circle.
Yeah, that's the Alleghay analyst term for it.
It basically means hitting the bullseye, finding the perfect
solution. Vision achieves this by
(12:51):
connecting three things, the rawdata, the analytical visualizers
or widgets, and the video itself.
And the crucial part is that it's highly customizable.
How custom are we talking? Extremely Lugay doesn't just
hand over the software. Their analysts provide what they
call 360° Consulting. They actually go to the club,
sit down with the coaching staffand the club's own analysts.
(13:14):
Right partnership, not just vendor relationship.
Exactly. And together they design and
build custom vision dashboards tailored specifically to that
club's football philosophy and the coach's specific needs and
points of focus. So a team focused on counter
attacks would have a completely different vision setup than a
team focused on high possession.Completely different.
The counter attacking teams dashboard might automatically
(13:37):
flag moments of rapid transition, measure the speed of
the forward runs, show spaces exploited behind the opposition
defense. The possession team's dashboard
might focus on passing networks,player spacing in the build up
phase, metrics around breaking defensive lines through
intricate passing. That level of personalization
making the tool fit the coach's brain rather than forcing the
(13:59):
coach to adapt to a generic toolthat feels like a huge
competitive advantage. Elygia certainly sees it that
way. They view this deep
customization and consultancy asa key competitive
differentiator. It ensures the powerful data is
actually used effectively because it speaks the specific
dialect of each individual coach.
OK, this is fascinating. We've got the data engine, we've
(14:20):
got the tools. We understand the importance of
tailoring it to the coach. Now let's push into the really
cutting edge stuff. Section 3 AI driven strategic
application and predictive analytics.
How is artificial intelligence changing the game here?
AI is really the catalyst, taking this whole operation to
the next level. It represents A fundamental
paradigm shift, moving beyond just descriptive analytics
(14:43):
telling you what happened towards genuine predictive and
even prescriptive capabilities. And the key advantage of AI is
speed and scale. Absolutely.
Think about those 3.5 million data points per game.
No human team could possibly sift through all that, connect
all the dots, identify all the subtle patterns in real time, or
even post match efficiently. AI algorithms excel at
(15:05):
processing these vast data sets incredibly quickly and
accurately. So it's supercharges the whole
decision making loop. Where are we seeing the biggest
impact of AI and football strategy right now?
Let's break down the core functions.
OK, first up, and arguably the most exciting, is predictive
analytics. Now, this isn't about redicting
the final score like 2-1, it's more granular.
(15:25):
Redicting what then? Redicting opponent behaviors,
tactical adjustments, even likely player actions based on
context. For instance, AI can analyze
historical data and real time patterns to predict with a
pretty significant degree of accuracy when a specific
opposition midfielder is likely to drop deeper to cover space
based on the ball's position andhis teammates movements.
(15:48):
OK. So if you can predict that micro
adjustment. You can preempt it.
A coach might get an alert or see a pattern suggesting this
will happen and instruct their forward to make a run into that
predicted space, maybe seconds before the opponent even moves.
It's about anticipating and exploiting tactical shifts
before they fully materialize. That's next level chess.
(16:09):
What about focusing on your own players?
That brings us to the second core function, player
performance optimization. AI models are brilliant at
analyzing individual player dataover long periods.
Not just match data, but training load, biometric data
where available, movement efficiency.
And what does it do with that? Create personalized plans.
Exactly. It can identify patterns
(16:30):
indicating fatigue risk, suggestmodifications to training
intensity, or even pinpoint specific biomechanical
inefficiencies that might lead to injury.
If player A always shows a drop off in Sprint speed after
playing 390 minute games in a row, the AI flags this and
suggest a tailored recovery or training plan for the following
week. So it's about maximizing
(16:52):
potential while minimizing risk player by player, yes.
Which leads to better team performance overall, fewer non
contact injuries, and actually higher player satisfaction
because they feel the training is truly individualized.
Makes sense, and obviously AI must be transforming how teams
find these players in the first place.
Absolutely. That's the third area scouting
in a recruitment. Traditional scouting, you know,
(17:14):
relied heavily on the eye test on subjective human judgement.
Watching hours and hours of video, going to games, all.
Right now, AI platforms can analyze performance data from
thousands of players across dozens of leagues using far more
metrics than just goals and assists.
They look at efficiency, decision making under pressure,
spatial awareness metrics derived from tracking data,
(17:35):
things that are hard to quantifyvisually.
So we can find hidden gems players undervalued by
traditional methods. Precisely.
It can identify talent much moreefficiently and objectively.
And perhaps even more powerfully, AI models can
attempt to project long term player development paths.
They might identify a 17 year old whose specific movement
signature and decision making profile strongly correlate with
(17:59):
success in the coaches desired tactical system five years down
the line. Wow, turning recruitment into a
much more data-driven predictivescience.
OK 4th area. How does AI impact the
day-to-day grind? That would be optimizing
training. AI can simulate various game
scenarios based on the upcoming opponents likely tactics pulled
from the predictive models we discussed.
(18:19):
Virtual practice. Sort of.
It calculates the probabilities of different outcomes within
those simulated scenarios. This allows coaches to design
training sessions that aren't just generic drills, but highly
specific scenario based exercises targeting the most
likely situations and challengesthey'll face in the next match,
according to the data. So practice becomes much more
(18:40):
focused and efficient. It sounds like AI is becoming
this indispensable analytical copilot.
That's a great way to put it, especially with real time
decision support. We talked about the NRT alerts,
but AI enhances this significantly.
During the game, AI systems are constantly processing the live
data stream, looking for emerging patterns or deviations.
(19:01):
And offering suggestions. Yeah, offering tactical
suggestions, not commands, but options.
Maybe suggesting a specific substitution based on fatigue
data and the tactical situation,or highlighting a sudden
weakness in the opponent's formation that's just appeared
and allows teams to react more quickly and effectively to the
dynamic, unfolding chaos of a match.
(19:21):
That analytical power is immense, but as we hinted
earlier, it's not just about themachine.
Let's transition to Section 4, which is crucial advanced use
cases applying expert criteria. This is where human intelligence
directs the AI, right? Absolutely, the LL analysts are
very clear on this. AI is a tool, but it's value
comes from the quality of the questions you ask it.
(19:43):
It needs expert human criteria to guide it towards meaningful
insights, not just noise. So the human analyst defines
what the AI should look for. Let's look at case one
optimizing space using the Boronoi spaces model.
Can you briefly explain what that model does?
Sure. In basic terms, a Boronoi
diagram divides the pitch into zones of control.
(20:03):
For any point on the field, the model tells you which player is
closest to that point, so it maps out who owns which area of
the pitch at any given instant. Sounds useful for understanding
space, but what were the problems with the traditional
versions of this model in football?
The traditional Boronoi models were too simplistic for
football's specific rules in physics.
They had two major flaws they completely ignored offside and
(20:25):
they assumed all players moved at the same speed or had the
same acceleration. Right, a space isn't useful if
it's offside. Exactly.
If the model highlights a huge open space, but receiving a pass
there would trigger an offside flag, that space is tactically
worthless. The advanced Allega model
automatically annulls basically ignores any calculated space
(20:45):
that is in an offside position relative to the ball and the
defense. It makes the spatial map
instantly more realistic. OK, that fixes the offsite
problem. How did they address the speed
and acceleration issue? They moved away from generic
assumptions. Instead, they assign unique
maximum speed and crucially, maximum acceleration values to
each individual player's ID based on their actual track
(21:07):
performance data overtime. So it's like giving each player
their own custom physics engine in the model, reflecting how
they actually move, like in a FIFA game, but based on real
data. That's a perfect analogy, and
interestingly, the analyst foundthat for controlling these
Bornois spaces, especially in tight areas, a player's
acceleration, their ability to burst into a space quickly, is
(21:28):
often more critical than their absolute top speed.
Football is often one in those first few yards.
That's a really sharp insight derived from applying football
knowledge to the model. Now.
There was also fascinating advice for analysts building
these models. Prioritize false positives over
false negatives. Why is it sometimes better for
the model to be slightly wrong but helpful?
(21:50):
This comes down to human psychology and maintaining trust
with the coaching staff. Imagine you're the analyst and
you present the coach with an 8 clip video package showing
defensive errors identified by your model.
OK, now if the coach vividly remembers one absolutely
disastrous defensive breakdown from the game, A really obvious
error, and your fancy model missed it, a false negative,
(22:11):
what happens? The coach loses faith.
Your system missed the most important mistake.
Instantly, credibility shattered.
So the pragmatic advice is it's actually better to show an 8
clip package where maybe two of the clips are borderline calls.
Technically false positives. The model flagged something that
wasn't a huge error but still relevant.
Rather than missing that one glaring mistake, the coach
(22:33):
remembers. So over deliver slightly on
potential issues rather than under deliver and miss something
obvious. Precisely.
You maintain the coach's trust that the system is seeing the
game realistically, even if it'soccasionally a bit
oversensitive. It's about building that human
computer relationship. Very insightful.
Let's look at case 2, the effectIman model, the magnet effect.
(22:55):
What's that measuring? The Effect Iman model quantifies
how strongly defending players are drawn towards the ball
carrier. Like a magnet, it measures
defensive compactness or sometimes over compactness
around the ball. Why is that useful?
It's incredibly useful for scouting opponents.
If your analysis shows that a rival teams midfield tends to
have a high Iman score, meaning they get sucked towards the
(23:17):
ball, you know where the space will likely open up.
On the opposite side or behind them?
Often, yes, on the weak side or in the gaps they vacate.
So you can prepare specific tactics like quick diagonal
passes or runs into those predicted spaces to exploit that
known tendency of your opponent.The model can also measure the
opposite how well A-Team organizes around the ball when
(23:39):
they win it back. Super specific tactical
preparation based on quantifiable opponent
tendencies. OK.
And case 3 involved intersectinglines using convex hull
analysis. What's the focus there?
Right. The convex hull, as we
mentioned, is basically the shape outlining your team's
overall position on the field. This specific analysis, though,
doesn't just look at the whole shape.
(24:01):
It focuses on the crucial intersections, the zones where
key lines of the team overlap. Which lines are we talking
about? Specifically, the intersection
between the defensive line and the central midfield line, and
also the intersection between the central midfield line and
the attacking line, particularlyas the team gets closer to the
opponent's box. And why is that intersectional
space so important? The analysis revealed that the
(24:22):
most dangerous attacking plays, the ones most likely to
resulting goals, often involve combinations and movements
within these specific intersectional zones near the
penalty area. It's the area where lines are
being broken, where defenders and midfielders are forced into
difficult decisions. So it pinpoints the highest
value real estate on the pitch for creating scoring chances,
(24:43):
moving beyond just saying attackthe final third.
Exactly, it gives a much more precise geographical target for
attacking strategies. These cases really highlight how
human expertise guides the AI tofind genuinely actionable
football insights. They really do.
But we need to be balanced. This whole system isn't perfect.
(25:04):
There are significant challenges.
Let's move into Section 5, the obstacles and the philosophical
debate of AI. What are the major hurdles?
Well, the first and perhaps mostfundamental is the issue of
imperfections of data. The old saying applies Garbage
in, garbage out. AI models are utterly dependent
on the quality of the data they receive.
And getting clean, reliable datafrom a chaotic football match
(25:25):
isn't always easy. Not at all.
Think about poor weather. Heavy rain, snow, even dense fog
can significantly impact the accuracy of the optical tracking
cameras, potentially corrupting the data feed.
And infrastructure differences. That too.
While Allega aims for standardization, there might
still be subtle differences or perhaps older camera systems in
(25:46):
some stadiums compared to others.
Ensuring consistent, high fidelity data capture across
every single venue under all conditions is a massive, ongoing
challenge. OK, so data quality is one
thing. What about actually making sense
of it? The interpretation challenge.
This is huge. As Roberta Lopez pointed out,
data needs context. Knowing a player sprinted 50
(26:07):
meters is meaningless unless youknow why.
Was it a brilliant attacking runor was he caught out of position
and desperately tracking back? AI struggles without why?
It struggles immensely with nuance and context, especially
those split second decisions made under pressure, influenced
by factors like opponent positioning, game state,
fatigue, maybe even just a gut feeling.
(26:27):
AI can measure the what, but understanding the why behind
complex human actions in a dynamic environment is still
very difficult. Which leads directly into the
cultural issues like the black box and trust issues.
Coaches aren't always comfortable with answers they
don't understand. Exactly.
Many advanced AI algorithms, especially deep learning models,
can function like a black box. They produce an output, a
(26:50):
suggestion, a prediction. But the exact step by step
reasoning how they reach that conclusion can be opaque even to
the experts who built them. And that's a problem for
adoption. It's a major barrier.
Football coaching relies heavilyon experience, intuition, and
clear cause and effect reasoning.
If a coach gets a tactical suggestion from an AI they don't
fully understand and it goes against their gut feeling
(27:12):
developed over decades, well, they're unlikely to trust it,
especially in high stakes moments.
It creates resistance and there's a financial barrier too,
isn't there? This tech isn't cheap.
Not at all. Developing and maintaining this
level of AI capability, plus therequired infrastructure, the
high res cameras, powerful servers, sophisticated sensors,
(27:33):
requires a substantial ongoing financial commitment.
This can definitely create a gapbetween the wealthiest clubs and
smaller clubs, even with La Liga's efforts to democratize
access. Plus you need highly skilled
staff to run it all, which adds to the cost.
And all this feeds into culturalresistance.
Football is a traditional sport in many ways.
Deeply traditional, it's historyis built on the genius of
(27:56):
coaches, the intuition of players, the feel of the game.
Embracing data-driven AI assisted decision making
requires a significant shift in mindset for everyone involved.
Roberto Lopez joked that coacheslike Ancelotti or Simeoni aren't
likely to take a Power BI course.
Exactly. The onus is on the analysts to
bridge that gap. They need to learn to speak the
coach's language. The source called it learning
(28:16):
German. If the analysts speak Chinese,
they need to translate complex data insights into simple,
visual, actionable formats, usually video clips that
resonate with the coach's existing understanding of
tactics and players. So analysts need tactical
understanding as much as data science skills.
OK, perhaps the biggest philosophical hurdle is the game
(28:36):
itself. The chaos of the game, it's
inherent unpredictability. This is fundamental.
Football is fluid, dynamic, chaotic.
So many critical factors are incredibly hard, maybe
impossible to quantify reliably.Things like player creativity,
mental resilience, team morale, the emotional state of
(28:57):
individuals on any given day. How do you measure confidence or
the impact of a player having a bad night's sleep?
You really can't, not systematically anyway.
There's another layer. Players operate in a hyper
competitive world. They often actively hide
weaknesses. What do you mean?
Well, if a player is struggling with a personal issue, stress at
home may be grieving a loss. They're often reluctant to share
(29:17):
that with the club hierarchy forfear it might be seen as
weakness or cost them their place in the team.
So crucial data about their truemental and emotional state might
simply be unavailable. Even if you had a way to measure
it, this prevents AI from havinga truly complete picture.
All of this builds towards what you called the paradox of
prediction. Can you explain that?
Yeah, it's a fascinating thoughtexperiment.
(29:38):
The analysts involved acknowledge that actually
prescribing exactly what to do to guarantee a win is basically
science fiction. Why?
Because if you could. Because if a perfect predictive
model existed 1 they could say with 100% certainty if you do XY
and Z you will win 20 what wouldhappen?
The other team would get that information or figure it out and
(30:00):
change their strategy to counterXY and Z.
Precisely the moment a perfect prediction becomes known, it
influences behavior in a way that invalidates the prediction
itself. Knowledge of the future would
change the future. So perfect prediction in a
competitive environment is logically impossible.
It seems to be. And the deeper philosophical
point is, if football did somehow become 100% predictable,
(30:23):
if the outcome was known in advance, would anyone watch the
chaos, the uncertainty, the possibility of upsets?
That's the magic. Analytics aims to reduce
uncertainty, but eliminating it entirely would kill the sport.
A powerful reminder of the limits of data in a human game.
OK, mindful of those limits, let's look ahead.
Section 6. Ethical concerns and the future
(30:45):
horizon. What are the key ethical
considerations with this level of data collection?
This is becoming increasingly important as data collection
gets more granular, touching on biometric and physiological
data. Player privacy is paramount.
Clubs need clear consent policies.
Where do you draw the line? Is tracking sleep quality or
stress levels fair game or is that confidential medical
(31:06):
information? Clear boundaries are essential.
That's a big one. What?
Else algorithmic bias and fairness.
If only wealthy clubs can affordthe absolute cutting edge of AI
analysis, does that create an unbridgeable gap, undermining
competitive balance and fairness?
Well, your standardized platformhelps mitigate this, but it's an
ongoing concern across sports tech.
(31:26):
And the risk of over reliance devaluing human expertise.
Yes, the human judgement factor.There's a risk that over
reliance on AI could sideline the invaluable experience and
instincts of seasoned coaches and scouts.
Find the right balance. The optimal fusion of talent and
technology is the goal. It's not AI versus human, it's
(31:46):
AI enhancing human capability. So the winning formula is talent
plus technology. Looking ahead, what emerging AI
applications are they exploring,particularly with that Globeint
and Microsoft partnership using Azure Open AI?
They're actively piloting several exciting generative AI
solutions using models like GPT 3.5 and GPT 4.
(32:08):
One key area is enhancing accessibility and engagement.
Like the NRT multi language subtitles.
Exactly, using AI like Whisper for transcription and GPT models
for translation, they're workingon generating automatic
subtitles in multiple languages for live match broadcasts almost
instantly. This is huge for global
audiences and particularly for inclusivity like helping people
with hearing impairments. Follow the commentary.
(32:29):
That's a fantastic use case. What about other content?
Automatic sports content translations.
Think about the sheer volume of articles, social media posts,
video descriptions Allajet produces.
Using Gen. AI to translate this content
quickly and accurately into manydifferent languages massively
increases their global reach andfan engagement worldwide.
(32:49):
And making content more relevantto individual fans.
Yes, that's the third pilot area.
Match briefings and personalizedcontent instead of generic match
reports. Imagine getting a summary
automatically generated by AI that focuses specifically on
your favorite team, maybe highlighting stats for players
you follow. Compare them to rivals, all
tailored to your known interests.
This deep personalization aims to build stronger fan loyalty
(33:12):
and reduce churn. Very cool.
Beyond Gen. AI, what other technologies are
on the horizon for a training and simulation?
The big ones mentioned are augmented reality AR and virtual
reality VR. The potential here for training
is immense. Imagine players using VR
headsets to run through specifictactical scenarios, defending a
(33:32):
particular set piece, practicingpenalty shootouts under
simulated crowd noise and pressure without the physical
fatigue of doing it repeatedly on the training pitch.
Arkansas could overlay tactical instructions or ghost players
onto the real training ground. It promises much more immersive
and efficient tactical learning.Wow.
From tracking dots 25 times a second to VR training
(33:54):
simulations, it's an incredible journey.
So wrapping this up, the move from media coach to sports and
performance, this deep commitment from Eliga, it seems
aimed at securing the entire league's future competitiveness.
That's the core strategy. By providing this advanced,
standardized platform and consultancy to all 42 clubs,
Eliga acts as a crucial democratizing factor.
(34:16):
It helps level the playing fieldsomewhat, ensure that
competitive advantages aren't soley based on which club has
the biggest budget for Tech. It lives the analytical
capability of the entire league.Ricardo Resta really emphasized
this. R&D, the university
partnerships, it all flows back to making the teams better on
the pitch. Exactly.
It translates directly, they believe, into increased
competitiveness, more sophisticated tactics league
(34:38):
wide and ultimately improve teamperformance across the board.
It's an investment in the quality and appeal of La Liga
football itself. A profound transformation using
data not to replace humans, but to empower them to refine
decisions and to push the boundaries of the beautiful
game. Beautifully put.
It's about making the game smarter without losing its soul.
(35:00):
OK, so you've heard how data is dissecting tactics, optimizing
players, even predicting microbehaviors.
But if this powerful sportion performance platform becomes the
standard, the baseline for everyone, it leaves us with a
fascinating question for you to ponder.
In this new era of data-driven strategy, what single,
unquantifiable human element, something the algorithms can't
(35:21):
truly capture, will ultimately become the most valuable
competitive edge in the next decade of football.
Think about that.
Welcome everyone to a really exciting deep dive.
Today we're pulling back the curtain on something truly
fascinating. Forget the itch side drama for
justice a moment because honestly the biggest battles in
football right now they might just be happening inside server
rooms. That's so true.
We're talking about this incredible fusion, this mix of
(00:20):
high level science, massive amounts of data and you know,
the beautiful, unpredictable chaos.
It is football. Exactly.
And we're zooming in on Spanish football, specifically a Liga,
to figure out how they didn't just dabble in data, they built
an entire system around it. And the result?
Well, let's just say it involvesa lot of silverware.
A lot of silverware is right. It's quite a story of and
(00:42):
commitment that the platform they built has been a
cornerstone for Spanish teams, leading to this frankly
unprecedented era of success. We're talking 16 European titles
since its implementation. 16 That's not just luck, is it that
strategy, that's technology making a tangible difference on
the European stage. Absolutely, and that success
really points to the power of the engine behind it all.
(01:04):
For years, many of you will haveknown this system as media
coach. It's been the standard since
what, the 2010 eleven season, analyzing every single movement.
Right Media coach has been the name synonymous with allergy
data. But things are evolving, aren't
they? There's a big change coming.
There is. It's a significant transition
starting July 15th, 2025. Media coach is becoming Sportion
(01:26):
Performance. Now.
This isn't just slapping a new label on it.
It's much bigger than that. OK, tell us more.
What's behind the name change? It signifies a major strategic
alliance. Allegia has partnered with
Sportion, which is the dedicatedsports technology division of
the global giant Globin. So Sportion performance reflects
this new, deeper integration anda commitment to, well,
(01:48):
continuous innovation across theentire league.
So, Sportion Globent, this sounds like a powerhouse
partnership aimed at pushing theboundaries even further.
Precisely. The goal isn't just analysis
anymore. It's about driving this constant
technological evolution league wide.
It's about institutionalizing that competitive edge through
tech. And to really understand how
(02:09):
they do that, you've got to lookunder the hood right at the
technological foundation. Let's let's impact this portion
performance ecosystem. Yeah.
The ecosystem strength really comes from that three-way
partnership. La Liga, Band, Sportion and
crucially, Microsoft providing the cloud infrastructure and AI
capabilities it's set up designed for constant growth.
(02:30):
And who's steering the ship within La Liga?
That responsibility falls to La Liga's newly named Football
Intelligence and Performance Unit.
Think of them as the internal hub, the brain trust that
overseas the whole strategy. They support all 42 professional
clubs, Loviga, EA Sports and La Liga Hypermotion. 42 clubs
getting this level of support and they're doing training
(02:51):
workshops. Constantly.
It's about ensuring every club, regardless of size, can leverage
these tools effectively. They're integrating new features
all the times, keeping everyone at the cutting edge.
OK, You mentioned continuous evolution, and for that you need
fuel. Data, fuel.
The sheer volume of data processed here must be
astronomical. Oh, it is.
The scale is almost hard to comprehend sometimes.
(03:12):
Let's put some numbers on it because that really tells the
story. Hit us with it.
For every single match played inElega, the platform captures
over 3.5 million data points. Three and a half million per
game. Law, game and that raw data is
then processed, analyzed, crunched, eventually generating
more than 6400 unique metrics that are made available to the
(03:35):
clubs. 6400 different ways to look at a match.
Exactly. And if you zoom out across an
entire season, this system turnsout over 112,000 analytical
reports. It's just a fire hose of
information. 112,000 reports. How on earth do they capture
that initial 3.5 million data points with the speed and
accuracy needed for it to be useful?
(03:56):
That feels like the core challenge it.
Is, and the answer lies in some pretty sophisticated tech
installed in every stadium. It's primarily an optical
tracking system that use a combination of cameras.
OK. What kind of cameras?
So you've got three main tactical cameras sort of
following the general play like you might see on a broadcast
feed, but position higher, wider, but the real precision
(04:16):
comes from up to 16 high resolution fixed perimeter
cameras. 16 fixed cameras dottedaround the stadium.
That's right, these are the workhorses.
They're constantly tracking the position of every player on the
pitch, both teams, plus the referees and of course, the ball
itself. And how frequently are they
capturing this positional data? This is key. 25 times per
(04:38):
second. That's 25 Hertz.
That frequency gives you incredibly granular data.
You're not just seeing where a player was, you're seeing their
exact trajectory, their micro movements, acceleration,
deceleration. 25 times every second. 25 times a second.
OK, now I understand how you getto 3.5 million data points.
That level of detail lets you analyze things like body
(04:59):
orientation, angles of runs, theexact distance between players,
things you just can't see reliably with the naked eye.
Precisely, it turns simple positioning into a rich data set
for advanced analysis. Now historically, all this
incredibly rich raw data, it wasmostly kept within the Leaguers
own analysis teams, right? The club's got the finished
(05:21):
reports the output, but not necessarily the underlying raw
numbers. Exactly.
But that's undergone a massive shift recently with something
called the Media Coach Sandbox. The sandbox.
OK, what's revolutionary about that?
Ricardo Resta, the director of Media Coach, calls it a game
changer. He does, and for good reason,
think, feel like this. For years clubs were getting,
(05:41):
you know, the beautifully Bay take the final analysis report.
But increasingly they were saying, hey, we appreciate the
cake, but can we get the flour, the eggs, the sugar?
We want the raw ingredients. I like that analogy.
They wanted to do their own baking, test their own recipes
based on their specific coachingphilosophy.
You've got it. The sandbox provides exactly
that. It's an advanced analytics
(06:01):
environment running in a secure data bricks workspace where
clubs get direct access to the raw data.
And raw data means. We're talking the fundamental
stuff. Individual player coordinates,
XY 25 times a second. Collective positioning data,
specific angles of movement, theteam shape metrics like the
(06:21):
convex hull. That's the area the team
occupies, Distances covered at different speeds, all the
granular detail. OK.
So they have the raw ingredients.
Why is that such a big deal beyond just having more numbers?
Because it unlocks true customization and automation,
often using AI. Imagine a coach is obsessed with
a very specific type of defensive press.
(06:43):
Maybe they don't just care abouttotal pressures.
Right, they want something more specific to their system.
Exactly. Maybe they want to measure, say,
how quickly do my 2 central midfielders close the space
between them after we lose the ball in the opponent's half.
That's a bespoke metric. Something you wouldn't find in a
standard report. Probably not, but now in the
sandbox the club's own analyst can use the raw positioning
(07:04):
data, maybe apply some AI, and build an automated process to
track exactly that metric tailored 100% to their coaches
unique tactical ideas. It bridges the gap between
generic data and specific coaching needs.
That makes perfect sense. It turns the platform from a
provider of information into a toolkit for generating unique
insights. And that leads us perfectly into
(07:25):
the next phase. How do these tools actually land
with the coaches? Let's move into Section 2
products and, crucially, the language of the coach.
Yeah, this is so important because you can have the most
powerful analytics engine in theworld, but if the coach doesn't
trust it or doesn't understand it or doesn't use it, it's
(07:46):
basically worthless. It stays in the analyst's
office. Exactly.
And La Liga seems very aware of this.
The philosophy, particularly championed by Roberto Lopez, who
coordinates a Ligas sports research department, is crystal
clear. He stresses that the technology
is a supplement, never a replacement for the coach.
So it's about decision support, not decision making.
Precisely. Lopez's phrase is that the goal
(08:08):
is to help coach make fewer mistakes, to provide objective
information, to challenge assumptions or confirm
observations. But ultimately, the final word
always belongs to the coach. Data informs intuition, it
doesn't override it. That cultural sensitivity feels
key to why it's been adopted so widely.
OK, so let's breakdown the actual suit of products and
(08:29):
services they offer. How does it work in practice?
Maybe starting with the heat of the moment during the match
itself. Right.
So during the game the primary tools are the live NRT near real
time tools. These are typically delivered
pitch side, often on a tablet ormaybe a digital whiteboard in
the dugout area. Near real time, so not quite
instantaneous. There's a slight delay, yeah,
(08:50):
and you have to accept that the data quality and quantity in NRT
might be a little bit lower thanwhat you get post match.
Processing millions of data points instantly and validating
them is incredibly complex. OK, so if the NRT data isn't
perfect, what's its main job during a game?
It's about providing immediate critical alerts for decision
support. The source has got a great
(09:11):
example, high injury risk alerts.
How does that work? The system is constantly
tracking physical metrics like high intensity meters run Sprint
distance, number of accelerations for every player.
Let's see if a star player absolutely crucial.
It's the 70th minute an alert might pop up on the analyst
tablet. Really to the bench warning.
Player 10, has exceeded 90% of his typical high intensity
(09:33):
distance for this stage of the game based on his seasonal load.
Yeah. So it's not saying he's injured,
it's saying objectively he's entering a known risk zone for
potential muscle fatigue or injury.
Exactly that. It doesn't tell the coach sub
him off. It presents objective data.
Here's a calculated risk factor.Now the coach has to make a
judgement call, right? Is the game situation critical
(09:56):
enough to push him, or is it safer to protect him for the
next match? That's the dilemma.
The NRT tool provides the objective data point.
The coach provides the experience, the context, the
human judgement. It's a perfect blend.
That's a really smart application of NRT.
OK, so the final whistle blows, the immediate pressure is off.
What happens next? Well, almost immediately after
(10:17):
the match, often literally on the team bus or maybe on the way
at a press conference, the coaching staff might use the
mobile application. The mobile app what for quick
look UPS. Yeah, quick reviews.
It serves a couple of purposes. 1 is just immediate internal
debriefing, but another, perhapsmore tactical use is preparing
for the media scrutiny. Like a defensive tool against
(10:39):
criticism. You could definitely see it that
way. Imagine the team just lost 10
despite playing OK. The narrative might be they look
tired or they just didn't show up in the second-half.
Right, the easy narratives. So the coach or an assistant can
quickly pull up the mobile app and find objective data.
Actually, look, our total distance covered was above
average. Our number of penalty box
entries was our third highest this season.
(11:02):
It allows them to immediately counter subjective criticism
with objective facts. Provides a data shield.
I like that. OK so NRT for alerts, mobile for
quick review and press conference prep.
But for the really deep post match dissection, where does the
series analysis happen? For that, Roberto Lopez points
to one tool as his personal favorite, and it seems to be the
(11:23):
core of the analytical workflow,the desktop visualizer, which is
called vision. Vision Why is this one the gold
standard for the analysts? Because Vision is where the
complex data finally starts speaking the coach's native
language. Yeah.
And what is that language? Video.
Video. Absolutely.
Coaches understand the game through watching it.
Vision's power is that it embedsall that granular tracking data,
(11:46):
the 3.5 million points per game,directly onto the corresponding
video footage. So you're not looking at
spreadsheets or abstract charts,you're watching the game, but
with layers of data visualized on top.
Exactly. You might see player movement
trails, distances between players displayed in real time,
shaded zones indicating controlled space, maybe even
probability metrics popping up during key moments, all
(12:08):
synchronized perfectly with the video.
I can see why that's powerful. It bridges the abstract data
world with the concrete visual world of football.
It's critical you can show a coach the most elegant Power BI
dashboard full of complex metrics, and honestly, their
eyes might just glaze over. It's not their language.
Show them three curated video clips using vision, highlighting
a specific tactical pattern withthe data overlaid.
(12:31):
Instant understanding, instant engagement.
Video is the universal translator, and Vision
apparently takes this further with deep customization.
They talk about achieving the quadrature of the circle.
Yeah, that's the Alleghay analyst term for it.
It basically means hitting the bullseye, finding the perfect
solution. Vision achieves this by
(12:51):
connecting three things, the rawdata, the analytical visualizers
or widgets, and the video itself.
And the crucial part is that it's highly customizable.
How custom are we talking? Extremely Lugay doesn't just
hand over the software. Their analysts provide what they
call 360° Consulting. They actually go to the club,
sit down with the coaching staffand the club's own analysts.
(13:14):
Right partnership, not just vendor relationship.
Exactly. And together they design and
build custom vision dashboards tailored specifically to that
club's football philosophy and the coach's specific needs and
points of focus. So a team focused on counter
attacks would have a completely different vision setup than a
team focused on high possession.Completely different.
The counter attacking teams dashboard might automatically
(13:37):
flag moments of rapid transition, measure the speed of
the forward runs, show spaces exploited behind the opposition
defense. The possession team's dashboard
might focus on passing networks,player spacing in the build up
phase, metrics around breaking defensive lines through
intricate passing. That level of personalization
making the tool fit the coach's brain rather than forcing the
(13:59):
coach to adapt to a generic toolthat feels like a huge
competitive advantage. Elygia certainly sees it that
way. They view this deep
customization and consultancy asa key competitive
differentiator. It ensures the powerful data is
actually used effectively because it speaks the specific
dialect of each individual coach.
OK, this is fascinating. We've got the data engine, we've
(14:20):
got the tools. We understand the importance of
tailoring it to the coach. Now let's push into the really
cutting edge stuff. Section 3 AI driven strategic
application and predictive analytics.
How is artificial intelligence changing the game here?
AI is really the catalyst, taking this whole operation to
the next level. It represents A fundamental
paradigm shift, moving beyond just descriptive analytics
(14:43):
telling you what happened towards genuine predictive and
even prescriptive capabilities. And the key advantage of AI is
speed and scale. Absolutely.
Think about those 3.5 million data points per game.
No human team could possibly sift through all that, connect
all the dots, identify all the subtle patterns in real time, or
even post match efficiently. AI algorithms excel at
(15:05):
processing these vast data sets incredibly quickly and
accurately. So it's supercharges the whole
decision making loop. Where are we seeing the biggest
impact of AI and football strategy right now?
Let's break down the core functions.
OK, first up, and arguably the most exciting, is predictive
analytics. Now, this isn't about redicting
the final score like 2-1, it's more granular.
(15:25):
Redicting what then? Redicting opponent behaviors,
tactical adjustments, even likely player actions based on
context. For instance, AI can analyze
historical data and real time patterns to predict with a
pretty significant degree of accuracy when a specific
opposition midfielder is likely to drop deeper to cover space
based on the ball's position andhis teammates movements.
(15:48):
OK. So if you can predict that micro
adjustment. You can preempt it.
A coach might get an alert or see a pattern suggesting this
will happen and instruct their forward to make a run into that
predicted space, maybe seconds before the opponent even moves.
It's about anticipating and exploiting tactical shifts
before they fully materialize. That's next level chess.
(16:09):
What about focusing on your own players?
That brings us to the second core function, player
performance optimization. AI models are brilliant at
analyzing individual player dataover long periods.
Not just match data, but training load, biometric data
where available, movement efficiency.
And what does it do with that? Create personalized plans.
Exactly. It can identify patterns
(16:30):
indicating fatigue risk, suggestmodifications to training
intensity, or even pinpoint specific biomechanical
inefficiencies that might lead to injury.
If player A always shows a drop off in Sprint speed after
playing 390 minute games in a row, the AI flags this and
suggest a tailored recovery or training plan for the following
week. So it's about maximizing
(16:52):
potential while minimizing risk player by player, yes.
Which leads to better team performance overall, fewer non
contact injuries, and actually higher player satisfaction
because they feel the training is truly individualized.
Makes sense, and obviously AI must be transforming how teams
find these players in the first place.
Absolutely. That's the third area scouting
in a recruitment. Traditional scouting, you know,
(17:14):
relied heavily on the eye test on subjective human judgement.
Watching hours and hours of video, going to games, all.
Right now, AI platforms can analyze performance data from
thousands of players across dozens of leagues using far more
metrics than just goals and assists.
They look at efficiency, decision making under pressure,
spatial awareness metrics derived from tracking data,
(17:35):
things that are hard to quantifyvisually.
So we can find hidden gems players undervalued by
traditional methods. Precisely.
It can identify talent much moreefficiently and objectively.
And perhaps even more powerfully, AI models can
attempt to project long term player development paths.
They might identify a 17 year old whose specific movement
signature and decision making profile strongly correlate with
(17:59):
success in the coaches desired tactical system five years down
the line. Wow, turning recruitment into a
much more data-driven predictivescience.
OK 4th area. How does AI impact the
day-to-day grind? That would be optimizing
training. AI can simulate various game
scenarios based on the upcoming opponents likely tactics pulled
from the predictive models we discussed.
(18:19):
Virtual practice. Sort of.
It calculates the probabilities of different outcomes within
those simulated scenarios. This allows coaches to design
training sessions that aren't just generic drills, but highly
specific scenario based exercises targeting the most
likely situations and challengesthey'll face in the next match,
according to the data. So practice becomes much more
(18:40):
focused and efficient. It sounds like AI is becoming
this indispensable analytical copilot.
That's a great way to put it, especially with real time
decision support. We talked about the NRT alerts,
but AI enhances this significantly.
During the game, AI systems are constantly processing the live
data stream, looking for emerging patterns or deviations.
(19:01):
And offering suggestions. Yeah, offering tactical
suggestions, not commands, but options.
Maybe suggesting a specific substitution based on fatigue
data and the tactical situation,or highlighting a sudden
weakness in the opponent's formation that's just appeared
and allows teams to react more quickly and effectively to the
dynamic, unfolding chaos of a match.
(19:21):
That analytical power is immense, but as we hinted
earlier, it's not just about themachine.
Let's transition to Section 4, which is crucial advanced use
cases applying expert criteria. This is where human intelligence
directs the AI, right? Absolutely, the LL analysts are
very clear on this. AI is a tool, but it's value
comes from the quality of the questions you ask it.
(19:43):
It needs expert human criteria to guide it towards meaningful
insights, not just noise. So the human analyst defines
what the AI should look for. Let's look at case one
optimizing space using the Boronoi spaces model.
Can you briefly explain what that model does?
Sure. In basic terms, a Boronoi
diagram divides the pitch into zones of control.
(20:03):
For any point on the field, the model tells you which player is
closest to that point, so it maps out who owns which area of
the pitch at any given instant. Sounds useful for understanding
space, but what were the problems with the traditional
versions of this model in football?
The traditional Boronoi models were too simplistic for
football's specific rules in physics.
They had two major flaws they completely ignored offside and
(20:25):
they assumed all players moved at the same speed or had the
same acceleration. Right, a space isn't useful if
it's offside. Exactly.
If the model highlights a huge open space, but receiving a pass
there would trigger an offside flag, that space is tactically
worthless. The advanced Allega model
automatically annulls basically ignores any calculated space
(20:45):
that is in an offside position relative to the ball and the
defense. It makes the spatial map
instantly more realistic. OK, that fixes the offsite
problem. How did they address the speed
and acceleration issue? They moved away from generic
assumptions. Instead, they assign unique
maximum speed and crucially, maximum acceleration values to
each individual player's ID based on their actual track
(21:07):
performance data overtime. So it's like giving each player
their own custom physics engine in the model, reflecting how
they actually move, like in a FIFA game, but based on real
data. That's a perfect analogy, and
interestingly, the analyst foundthat for controlling these
Bornois spaces, especially in tight areas, a player's
acceleration, their ability to burst into a space quickly, is
(21:28):
often more critical than their absolute top speed.
Football is often one in those first few yards.
That's a really sharp insight derived from applying football
knowledge to the model. Now.
There was also fascinating advice for analysts building
these models. Prioritize false positives over
false negatives. Why is it sometimes better for
the model to be slightly wrong but helpful?
(21:50):
This comes down to human psychology and maintaining trust
with the coaching staff. Imagine you're the analyst and
you present the coach with an 8 clip video package showing
defensive errors identified by your model.
OK, now if the coach vividly remembers one absolutely
disastrous defensive breakdown from the game, A really obvious
error, and your fancy model missed it, a false negative,
(22:11):
what happens? The coach loses faith.
Your system missed the most important mistake.
Instantly, credibility shattered.
So the pragmatic advice is it's actually better to show an 8
clip package where maybe two of the clips are borderline calls.
Technically false positives. The model flagged something that
wasn't a huge error but still relevant.
Rather than missing that one glaring mistake, the coach
(22:33):
remembers. So over deliver slightly on
potential issues rather than under deliver and miss something
obvious. Precisely.
You maintain the coach's trust that the system is seeing the
game realistically, even if it'soccasionally a bit
oversensitive. It's about building that human
computer relationship. Very insightful.
Let's look at case 2, the effectIman model, the magnet effect.
(22:55):
What's that measuring? The Effect Iman model quantifies
how strongly defending players are drawn towards the ball
carrier. Like a magnet, it measures
defensive compactness or sometimes over compactness
around the ball. Why is that useful?
It's incredibly useful for scouting opponents.
If your analysis shows that a rival teams midfield tends to
have a high Iman score, meaning they get sucked towards the
(23:17):
ball, you know where the space will likely open up.
On the opposite side or behind them?
Often, yes, on the weak side or in the gaps they vacate.
So you can prepare specific tactics like quick diagonal
passes or runs into those predicted spaces to exploit that
known tendency of your opponent.The model can also measure the
opposite how well A-Team organizes around the ball when
(23:39):
they win it back. Super specific tactical
preparation based on quantifiable opponent
tendencies. OK.
And case 3 involved intersectinglines using convex hull
analysis. What's the focus there?
Right. The convex hull, as we
mentioned, is basically the shape outlining your team's
overall position on the field. This specific analysis, though,
doesn't just look at the whole shape.
(24:01):
It focuses on the crucial intersections, the zones where
key lines of the team overlap. Which lines are we talking
about? Specifically, the intersection
between the defensive line and the central midfield line, and
also the intersection between the central midfield line and
the attacking line, particularlyas the team gets closer to the
opponent's box. And why is that intersectional
space so important? The analysis revealed that the
(24:22):
most dangerous attacking plays, the ones most likely to
resulting goals, often involve combinations and movements
within these specific intersectional zones near the
penalty area. It's the area where lines are
being broken, where defenders and midfielders are forced into
difficult decisions. So it pinpoints the highest
value real estate on the pitch for creating scoring chances,
(24:43):
moving beyond just saying attackthe final third.
Exactly, it gives a much more precise geographical target for
attacking strategies. These cases really highlight how
human expertise guides the AI tofind genuinely actionable
football insights. They really do.
But we need to be balanced. This whole system isn't perfect.
(25:04):
There are significant challenges.
Let's move into Section 5, the obstacles and the philosophical
debate of AI. What are the major hurdles?
Well, the first and perhaps mostfundamental is the issue of
imperfections of data. The old saying applies Garbage
in, garbage out. AI models are utterly dependent
on the quality of the data they receive.
And getting clean, reliable datafrom a chaotic football match
(25:25):
isn't always easy. Not at all.
Think about poor weather. Heavy rain, snow, even dense fog
can significantly impact the accuracy of the optical tracking
cameras, potentially corrupting the data feed.
And infrastructure differences. That too.
While Allega aims for standardization, there might
still be subtle differences or perhaps older camera systems in
(25:46):
some stadiums compared to others.
Ensuring consistent, high fidelity data capture across
every single venue under all conditions is a massive, ongoing
challenge. OK, so data quality is one
thing. What about actually making sense
of it? The interpretation challenge.
This is huge. As Roberta Lopez pointed out,
data needs context. Knowing a player sprinted 50
(26:07):
meters is meaningless unless youknow why.
Was it a brilliant attacking runor was he caught out of position
and desperately tracking back? AI struggles without why?
It struggles immensely with nuance and context, especially
those split second decisions made under pressure, influenced
by factors like opponent positioning, game state,
fatigue, maybe even just a gut feeling.
(26:27):
AI can measure the what, but understanding the why behind
complex human actions in a dynamic environment is still
very difficult. Which leads directly into the
cultural issues like the black box and trust issues.
Coaches aren't always comfortable with answers they
don't understand. Exactly.
Many advanced AI algorithms, especially deep learning models,
can function like a black box. They produce an output, a
(26:50):
suggestion, a prediction. But the exact step by step
reasoning how they reach that conclusion can be opaque even to
the experts who built them. And that's a problem for
adoption. It's a major barrier.
Football coaching relies heavilyon experience, intuition, and
clear cause and effect reasoning.
If a coach gets a tactical suggestion from an AI they don't
fully understand and it goes against their gut feeling
(27:12):
developed over decades, well, they're unlikely to trust it,
especially in high stakes moments.
It creates resistance and there's a financial barrier too,
isn't there? This tech isn't cheap.
Not at all. Developing and maintaining this
level of AI capability, plus therequired infrastructure, the
high res cameras, powerful servers, sophisticated sensors,
(27:33):
requires a substantial ongoing financial commitment.
This can definitely create a gapbetween the wealthiest clubs and
smaller clubs, even with La Liga's efforts to democratize
access. Plus you need highly skilled
staff to run it all, which adds to the cost.
And all this feeds into culturalresistance.
Football is a traditional sport in many ways.
Deeply traditional, it's historyis built on the genius of
(27:56):
coaches, the intuition of players, the feel of the game.
Embracing data-driven AI assisted decision making
requires a significant shift in mindset for everyone involved.
Roberto Lopez joked that coacheslike Ancelotti or Simeoni aren't
likely to take a Power BI course.
Exactly. The onus is on the analysts to
bridge that gap. They need to learn to speak the
coach's language. The source called it learning
(28:16):
German. If the analysts speak Chinese,
they need to translate complex data insights into simple,
visual, actionable formats, usually video clips that
resonate with the coach's existing understanding of
tactics and players. So analysts need tactical
understanding as much as data science skills.
OK, perhaps the biggest philosophical hurdle is the game
(28:36):
itself. The chaos of the game, it's
inherent unpredictability. This is fundamental.
Football is fluid, dynamic, chaotic.
So many critical factors are incredibly hard, maybe
impossible to quantify reliably.Things like player creativity,
mental resilience, team morale, the emotional state of
(28:57):
individuals on any given day. How do you measure confidence or
the impact of a player having a bad night's sleep?
You really can't, not systematically anyway.
There's another layer. Players operate in a hyper
competitive world. They often actively hide
weaknesses. What do you mean?
Well, if a player is struggling with a personal issue, stress at
home may be grieving a loss. They're often reluctant to share
(29:17):
that with the club hierarchy forfear it might be seen as
weakness or cost them their place in the team.
So crucial data about their truemental and emotional state might
simply be unavailable. Even if you had a way to measure
it, this prevents AI from havinga truly complete picture.
All of this builds towards what you called the paradox of
prediction. Can you explain that?
Yeah, it's a fascinating thoughtexperiment.
(29:38):
The analysts involved acknowledge that actually
prescribing exactly what to do to guarantee a win is basically
science fiction. Why?
Because if you could. Because if a perfect predictive
model existed 1 they could say with 100% certainty if you do XY
and Z you will win 20 what wouldhappen?
The other team would get that information or figure it out and
(30:00):
change their strategy to counterXY and Z.
Precisely the moment a perfect prediction becomes known, it
influences behavior in a way that invalidates the prediction
itself. Knowledge of the future would
change the future. So perfect prediction in a
competitive environment is logically impossible.
It seems to be. And the deeper philosophical
point is, if football did somehow become 100% predictable,
(30:23):
if the outcome was known in advance, would anyone watch the
chaos, the uncertainty, the possibility of upsets?
That's the magic. Analytics aims to reduce
uncertainty, but eliminating it entirely would kill the sport.
A powerful reminder of the limits of data in a human game.
OK, mindful of those limits, let's look ahead.
Section 6. Ethical concerns and the future
(30:45):
horizon. What are the key ethical
considerations with this level of data collection?
This is becoming increasingly important as data collection
gets more granular, touching on biometric and physiological
data. Player privacy is paramount.
Clubs need clear consent policies.
Where do you draw the line? Is tracking sleep quality or
stress levels fair game or is that confidential medical
(31:06):
information? Clear boundaries are essential.
That's a big one. What?
Else algorithmic bias and fairness.
If only wealthy clubs can affordthe absolute cutting edge of AI
analysis, does that create an unbridgeable gap, undermining
competitive balance and fairness?
Well, your standardized platformhelps mitigate this, but it's an
ongoing concern across sports tech.
(31:26):
And the risk of over reliance devaluing human expertise.
Yes, the human judgement factor.There's a risk that over
reliance on AI could sideline the invaluable experience and
instincts of seasoned coaches and scouts.
Find the right balance. The optimal fusion of talent and
technology is the goal. It's not AI versus human, it's
(31:46):
AI enhancing human capability. So the winning formula is talent
plus technology. Looking ahead, what emerging AI
applications are they exploring,particularly with that Globeint
and Microsoft partnership using Azure Open AI?
They're actively piloting several exciting generative AI
solutions using models like GPT 3.5 and GPT 4.
(32:08):
One key area is enhancing accessibility and engagement.
Like the NRT multi language subtitles.
Exactly, using AI like Whisper for transcription and GPT models
for translation, they're workingon generating automatic
subtitles in multiple languages for live match broadcasts almost
instantly. This is huge for global
audiences and particularly for inclusivity like helping people
with hearing impairments. Follow the commentary.
(32:29):
That's a fantastic use case. What about other content?
Automatic sports content translations.
Think about the sheer volume of articles, social media posts,
video descriptions Allajet produces.
Using Gen. AI to translate this content
quickly and accurately into manydifferent languages massively
increases their global reach andfan engagement worldwide.
(32:49):
And making content more relevantto individual fans.
Yes, that's the third pilot area.
Match briefings and personalizedcontent instead of generic match
reports. Imagine getting a summary
automatically generated by AI that focuses specifically on
your favorite team, maybe highlighting stats for players
you follow. Compare them to rivals, all
tailored to your known interests.
This deep personalization aims to build stronger fan loyalty
(33:12):
and reduce churn. Very cool.
Beyond Gen. AI, what other technologies are
on the horizon for a training and simulation?
The big ones mentioned are augmented reality AR and virtual
reality VR. The potential here for training
is immense. Imagine players using VR
headsets to run through specifictactical scenarios, defending a
(33:32):
particular set piece, practicingpenalty shootouts under
simulated crowd noise and pressure without the physical
fatigue of doing it repeatedly on the training pitch.
Arkansas could overlay tactical instructions or ghost players
onto the real training ground. It promises much more immersive
and efficient tactical learning.Wow.
From tracking dots 25 times a second to VR training
(33:54):
simulations, it's an incredible journey.
So wrapping this up, the move from media coach to sports and
performance, this deep commitment from Eliga, it seems
aimed at securing the entire league's future competitiveness.
That's the core strategy. By providing this advanced,
standardized platform and consultancy to all 42 clubs,
Eliga acts as a crucial democratizing factor.
(34:16):
It helps level the playing fieldsomewhat, ensure that
competitive advantages aren't soley based on which club has
the biggest budget for Tech. It lives the analytical
capability of the entire league.Ricardo Resta really emphasized
this. R&D, the university
partnerships, it all flows back to making the teams better on
the pitch. Exactly.
It translates directly, they believe, into increased
competitiveness, more sophisticated tactics league
(34:38):
wide and ultimately improve teamperformance across the board.
It's an investment in the quality and appeal of La Liga
football itself. A profound transformation using
data not to replace humans, but to empower them to refine
decisions and to push the boundaries of the beautiful
game. Beautifully put.
It's about making the game smarter without losing its soul.
(35:00):
OK, so you've heard how data is dissecting tactics, optimizing
players, even predicting microbehaviors.
But if this powerful sportion performance platform becomes the
standard, the baseline for everyone, it leaves us with a
fascinating question for you to ponder.
In this new era of data-driven strategy, what single,
unquantifiable human element, something the algorithms can't
(35:21):
truly capture, will ultimately become the most valuable
competitive edge in the next decade of football.
Think about that.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
The Breakfast Club
The World's Most Dangerous Morning Show, The Breakfast Club, With DJ Envy, Jess Hilarious, And Charlamagne Tha God!