Analytics won’t get you home
During a postgame press conference in the Knicks' Game 2 win against the Cavs, Josh Hart once described analytics with a line that has stuck with me:
"I'm never a huge analytics guy. At a certain point, they're a lamppost to a drunk person. You can lean on them, but it won't get you home."
Hart attributes the quote to Jay Wright, who likely found his rendition off of Andrew Lang’s:
“Most people use statistics like a drunk man uses a lamppost; more for support than illumination".
Hart’s interview got a chuckle from me, but it also captures one of the biggest challenges with data in sport.
We have more information available than ever. Athlete monitoring systems, force plates, tracking technologies, wearables, testing batteries, dashboards, and increasingly sophisticated models can quantify almost everything we do. The temptation is to believe that more data inevitably leads to better decisions.
It doesn’t.
Analytics can provide evidence, identify patterns, challenge assumptions, and help us see things we otherwise might miss. But the numbers do not understand the athlete, the environment, or the decision we are trying to make.
That part is still our job.
Context is key
A number rarely means much on its own.
In responding to Hart’s comment, Colin Cowherd made a related point:
Analytics can be incredibly useful across the long sample of a regular season, but the playoffs become much more situational. Matchups change. Rotations shorten. Roles shift. Individual possessions matter more. The context surrounding the numbers changes, and their interpretation has to change with it.
The same applies to performance data. Statistics are representations of reality, not reality itself.
When we summarize something that happened, we often strip away some of the context that made it happen.
Take something as simple as a team average for training load. It might accurately describe the average exposure of the group, while representing the actual exposure of almost no one. One athlete may have completed the entire session, another may have been limited, another may have accumulated substantially more high-intensity work because of their position or role, and another may have spent part of practice in rehabilitation. The average is mathematically correct. It just may not describe any individual athlete particularly well.
The same problem exists when we interpret changes in testing. A drop in jump height could represent fatigue. It could also represent a change in strategy, testing conditions, intent, pain, or simply normal day-to-day variation. A spike in external load could represent excessive demand—or exactly the exposure an athlete needed.
The statistic is not necessarily wrong. It is incomplete.
This is why collecting data and interpreting data are fundamentally different skills. The measurement gives us information. Context gives that information meaning.
Without context, analytics can create the illusion of precision without actually improving our understanding.
Assumptions need to be understood
Every metric comes with assumptions.
Before interpreting a number, we should understand what was actually measured, how it was measured, and what we are inferring from it.
A wearable does not measure “readiness.” A force plate does not measure “performance.” A GPS or LPS system does not measure “fatigue.” These systems measure specific signals from which we derive metrics and, eventually, make inferences about broader constructs.
Each step away from the original measurement introduces assumptions.
That does not make the metric useless. It makes understanding the measurement essential.
If we do not understand where a number comes from, its reliability, its limitations, and the assumptions required to interpret it, the sophistication of the dashboard displaying it does not matter.
Decision frameworks must come first
One of the easiest mistakes in sport science is to start with the data.
We collect everything we can, build a dashboard, establish thresholds, and then ask: What should we do with this?
The order should be reversed.
What decision are we trying to make? What information would meaningfully influence that decision? What would cause us to change the plan? And if something changes, what mechanism do we actually have available to respond?
The framework should determine what we measure, not the other way around.
That means defining the question, understanding the possible decisions, identifying the information required to distinguish between them, and establishing how that information connects to action.
Otherwise, we risk building increasingly sophisticated systems for answering questions nobody was asking and with no mechanism for response.
Analytics should support not replace decision making
The goal of analytics is not to remove humans from the decision.
It is to make human decisions better informed.
Coaches see things that sensors cannot. Athletes provide information that cannot always be reduced to a metric. Clinicians understand injury history, symptoms, and biological constraints. Strength and conditioning coaches understand the intent behind training. Sport scientists can help connect those observations with objective information and identify patterns across time.
None of those perspectives should exist in isolation.
Good decision-making happens when objective information is integrated with experience, observation, communication, and an understanding of the environment.
Analytics should challenge our intuition when necessary. They should help us recognize patterns our eyes may miss. They should create a shared language across disciplines and reduce uncertainty where possible.
But they should never become a substitute for thinking.
Analytics still have tremendous value…
None of this is an argument against analytics. It is an argument for using them better.
Human observation has limitations too. We are influenced by recency, expectations, previous experiences, and what happens to capture our attention. We cannot reliably remember every exposure, recognize every trend, or simultaneously track every variable that may matter across an athlete’s season.
Analytics can help fill those gaps.
They allow us to quantify what we otherwise might only describe. They can establish baselines, track changes over time, identify patterns that are difficult to see in individual observations, and challenge the stories we tell ourselves about what is happening.
Most importantly, they can reduce uncertainty.
They rarely give us the answer. But when the right information is collected, interpreted within the right context, and connected to a clear decision framework, analytics can help narrow the range of plausible explanations and give us greater confidence in the decisions we make.
That is their value, but we can choose to use them for either…
Support: Use numbers or data to prop up an argument
or
Illumination: Shed light on a problem or teach us something new
Josh Hart was right.
Analytics are like lampposts. They may not get you home…
But they can light the way.
TOMMY OTLEYPhysical therapist, sport scientist, educator, and researcher working at the intersection of rehabilitation and performance.
READ MORE
Returning to competition is an important milestone, but it does not mean performance has been restored. This edition of The Film Room explores the gap between return and performance, why performance requires an outcome, and how role, opportunity, and context complicate what “back” really means. If we say an athlete has returned to performance, we have to be able to show it.