Performance Data Primer: Back to the Future

We’ve been here before

Every generation seems to believe it is standing at the edge of a revolution.

A new technology arrives. A new analytical method becomes accessible. A new way of measuring the athlete promises to change how we train, rehabilitate, predict, and perform.

And sometimes it does.

The tools available to sport science today would have been difficult to imagine not that long ago. We can track athletes continuously through practices and games. Cameras can estimate human movement without markers. Force plates can automatically process thousands of tests across a season. Information that once required hours of manual work can move from collection to dashboard almost instantly. Artificial intelligence creates another layer of possibility, allowing us to process, organize, and interrogate amounts of information that would be impossible for any individual practitioner to manage.

There is plenty to be excited about.

But look closely enough at many of our “new” problems and they start to look strangely familiar.

The technology changes. The terminology changes. The processing becomes faster. The visualization becomes prettier.

Yet underneath it all, we continue asking many of the same questions.

What are we actually measuring? What does it represent? What assumptions did we make along the way? Who should have access to the information? Does the information matter? How should we communicate it? And ultimately, what should we do?

Maybe moving into the future requires occasionally looking backward.

Not because the past was better.

Because we have already learned some of these lessons.

We keep reinventing the dashboard

The modern performance dashboard can feel like an entirely new way of understanding athletes.

In some ways, it is.

We can automatically integrate information from multiple technologies, calculate hundreds of variables, establish individual baselines, identify changes, and deliver those results to practitioners before an athlete has finished breakfast.

But strip away the interface and the process starts to look familiar.

We collect several imperfect representations of the athlete. We compare them against some reference. We decide which changes matter. We combine that information with other things we know. Then someone tries to determine what it means.

Twenty years ago, that might have happened in an Excel spreadsheet.

Today, it might happen in an athlete management system.

Tomorrow, an AI model may summarize the information for us.

The capability has changed enormously. The underlying problem has not.

That doesn't make the innovation meaningless. Faster processing, automated pipelines, improved visualization, and greater accessibility can fundamentally improve how information moves through an organization.

But improved processing does not automatically mean improved understanding.

If we don't understand the construct, automating it doesn't help.

If the measurement is poor, putting it into a better dashboard doesn't fix it.

If the decision isn't clear, adding another metric may only give us another metric.

Sometimes we improve the tool without improving the problem the tool was supposed to solve.

Prediction has been the future for a long time

Prediction might be the clearest example of our tendency to believe the newest method has finally solved an old problem.

Sport has spent decades searching for ways to identify what will happen next.

We identified individual risk factors. Then we built screening batteries. We combined variables into scores. We examined training-load relationships. We built regression models. Machine learning allowed us to model increasingly complex relationships. Artificial intelligence now promises another enormous expansion in what is possible.

These methods are not equivalent. Our computational capabilities have genuinely changed, and new approaches may allow us to identify patterns that previous methods could not.

But many of the questions surrounding prediction remain remarkably familiar.

What exactly are we trying to predict? How frequently does it occur? How representative is the population used to build the model? Are the relationships stable across environments? Does the model perform when we move it outside the population in which it was developed? What happens when our behavior changes because the model told us someone was at risk?

And perhaps most importantly:

What are we actually going to do with the prediction?

A model can become dramatically more sophisticated without making the decision on the other side any clearer.

We should be careful about looking back at yesterday's prediction models as simplistic while assuming today's have escaped the same fundamental problems simply because the mathematics have changed.

The future may give us better answers.

It does not free us from asking good questions.

Prediction has been the future for a long time…and probably will continue to be.

New technology…old assumptions

Markerless motion capture feels futuristic.

A camera records an athlete performing a task. Computer vision identifies the athlete. A pose-estimation model identifies key points. From those points, we begin estimating joint positions, angles, velocities, and other biomechanical variables.

No markers. No laboratory. Potentially very little disruption to the athlete.

That represents real progress.

But we didn't eliminate assumptions.

We changed them.

We still have to ask how joint centers are being defined. We still need a coordinate system. We still need to understand what happens when a body segment becomes occluded. We need to know how the model was trained, what population it was trained on, and whether the task we are using resembles the tasks on which it performs well. We still need to determine whether the variable produced by the system represents the construct we actually care about.

We replaced markers with pixels.

We didn't replace measurement theory.

The same will be true of many technologies that come next.

New tools may remove old limitations while introducing entirely different ones. Our responsibility is not to resist those tools because they are imperfect. Every measurement is imperfect.

It is to avoid assuming that novelty somehow frees us from the fundamentals.

If anything, the faster our technology advances, the more important those fundamentals become.

We keep trying to remove the human

There has always been something appealing about the idea that enough information might eventually make the decision obvious.

Collect the right variables.

Build the right model.

Set the right thresholds.

Green.

Yellow.

Red.

Train.

Modify.

Rest.

Return.

Artificial intelligence makes that possibility feel closer than ever. If a system can simultaneously process training load, force-plate data, strength testing, wellness, sleep, medical information, video, and competition schedules, perhaps it can finally tell us what to do.

But that assumes the hardest part of the problem is processing the information.

Often, it isn't.

The difficult part is deciding how much each piece of information should matter in the situation sitting in front of us.

A player can be fatigued and still need to train.

A metric can be abnormal without requiring intervention.

A rehabilitation test can look excellent while the athlete remains unprepared for the demands waiting on the other side.

A model can be correct about risk without making the appropriate response to that risk obvious.

This is where the ideas from the previous installment of the Performance Data Primer become increasingly important.

More information can improve our epistēmē.

Better tools can expand our technē.

Neither eliminates the need for phronēsis.

Perhaps the future isn't removing the practitioner from the loop.

Perhaps it is giving the practitioner a much better loop.

Nothing is new. And everything is.

It would be easy for this argument to become cynical.

We have seen this before. The new thing isn't actually new. Today's breakthrough will eventually encounter yesterday's problems.

But that misses what makes this moment genuinely exciting.

Some things really are different.

The scale is different.

The speed is different.

The resolution is different.

Our ability to collect information without disrupting the environment is different. Our ability to connect previously isolated sources of information is different. Our ability to process enormous datasets is different. Our ability to recognize patterns that would be impossible for a human practitioner to see is different.

And access may become one of the biggest changes of all.

Things that once required a biomechanics laboratory, expensive equipment, specialized personnel, and hours of processing may increasingly become available through a camera, wearable, or automated system.

That matters.

The opportunity isn't to decide between old wisdom and new technology.

It is to figure out what becomes possible when we combine them.

The future is integration

When we talk about the future of performance data, integration is often interpreted as a technology problem.

Put the force-plate data here. Bring the tracking data into the same database. Connect the athlete management system. Add wellness. Add medical information. Add video. Then perhaps place an AI layer over everything so it can summarize the result.

That may be data integration.

It is not necessarily integration into practice.

The integration that interests me more happens farther downstream.

Can we bring different sources of information together around a practical problem? Can we understand what each source contributes, where each is limited, and how much weight it deserves? Can that information enter the training, rehabilitation, or performance process in a way that actually changes what happens next?

That doesn't require artificial intelligence.

Sometimes it requires a sophisticated data architecture.

Sometimes it requires a well-designed report.

Sometimes it requires several practitioners standing around a table discussing an athlete.

And sometimes it requires one person recognizing that the information they have collected should change today's plan.

The future of integration shouldn't simply be about getting everything into the same place.

It should be about getting the right information into the right decision.

That distinction takes us beyond communication.

Communication asks whether the information reached someone.

Translation asks whether they understood what it meant.

Integration into practice asks whether that information could be combined with everything else we know about the athlete and environment to inform an appropriate action.

And then we need to see what happened.

Did the intervention produce the response we expected? Did the athlete adapt? Did our interpretation hold up? Should the plan continue, or should we pivot?

In that sense, integration isn't an endpoint.

It is a loop.

Information enters practice. Practice produces an outcome. The outcome creates new information. That information changes what we know and shapes what we do next.

AI may eventually make that loop faster and more powerful.

But AI isn't the loop.

Practice is.

Bringing it all together

This is where I think the future of performance data connects back to everything in this series.

We started with ethics.

Just because we can collect something does not mean we should. As technology becomes less intrusive and information becomes easier to collect, that question becomes more important, not less.

Then we moved to assumptions.

Every measurement, model, and prediction contains assumptions. New technology does not make those assumptions disappear. Sometimes it simply hides them deeper inside systems we understand less.

Then came translation.

Information does not become useful simply because it exists. We need the scientific foundation to understand it, the craft to make it usable, and the practical wisdom to decide what should actually be done.

And now we arrive at the future.

New technology does not invalidate any of those lessons. It makes them more important.

Progress or novelty?

Maybe one of the most useful distinctions as we move forward is between novelty and progress.

We tend to treat them as though they are the same.

They aren't.

Something can be new without making anything better.

And an old idea can become extraordinarily powerful when new technology finally gives us the ability to apply it differently.

So perhaps the question we should ask about the next technology, model, platform, or metric isn't simply:

What's new?

Maybe it is:

What's actually better?

Does it measure the construct better?

Does it reduce the burden on the athlete?

Does it allow us to understand something we couldn't before?

Does it make important information more accessible?

Does it acknowledge uncertainty rather than hide it?

Does it help practitioners make better decisions?

Does it improve what happens for the athlete?

Innovation should move us forward.

Novelty simply makes something different.

Back to the future

The future of performance data will undoubtedly look different.

Our sensors will become smaller. Our cameras will become measurement systems. Our datasets will become larger. Our models will become more sophisticated. Artificial intelligence will help us recognize patterns, integrate information, and automate work that currently consumes enormous amounts of time.

We should be excited about that.

But we should also remember that a more sophisticated tool does not free us from the fundamentals.

We will still need to ask whether we should collect the information.

We will still need to understand what was actually measured.

We will still need to interrogate the assumptions between the signal and the conclusion.

We will still need to translate information into the messy environment of sport.

And someone will still have to decide what to do.

Maybe that is the real lesson from going back to the future.

The tools will change. The responsibility doesn't.

The future of sport science doesn't require us to choose between what we have learned and what is possible. Progress happens when we bring both with us.

TOMMY OTLEY

Physical therapist, sport scientist, educator, and researcher working at the intersection of rehabilitation and performance.

QUICK LINKS

Free Downloads

Services

Work With Me

Ask A Question

READ MORE