Athlete Performance Profiles and the Problem of Interpreting Testing Data in Practice
A practitioner-focused look at athlete profiling and the question of what actually constitutes “good” performance
If you have worked with athletes for long enough, we can probably agree on one thing: athlete performance testing gives us a lot of useful data.
The challenge is not data availability. It is using that data to make applied decisions.
Force plate testing is more common than ever in gyms and clinics and is widely used to guide performance and rehabilitation decisions. Yet even a single countermovement jump produces dozens of outcome metrics describing how an athlete performs, brakes, and moves.
At the same time, we know that different neuromuscular capacities are trained differently. Maximal strength is trained differently than maximal power. Maximal power is trained differently than plyometric function. Eccentric braking capacities are trained differently than concentric propulsive capacities.
As a result, bodyweight countermovement jumping alone, which reflects strength under relatively light loads, is insufficient to characterize an athlete’s broader envelope of function. Practitioners therefore rely on multiple tests that probe performance under different constraints, including loaded jumps, maximal strength assessments, plyometric tasks, and single-leg testing.
This depth of testing reflects a reasonable goal: to better understand the capacities that contribute to performance.
But the downside is obvious.
If a single CMJ produces dozens of metrics, imagine the interpretive challenge once maximal strength, maximal power, and plyometric assessments are added to the story.
The Normal Distribution Trap
Faced with this problem, a common response is to interpret athlete performance using isolated metrics through a lens of normative data. Athletes are compared to group averages, percentiles are assigned, and “good” performance is defined relative to the distribution.
Relying on normative data to guide the interpretation of sport science data is, in many ways, a step in the right direction. However, its usefulness depends on the degree to which the reference population truly represents the athlete being evaluated. In practice, defining what constitutes a sex-, sport-, and position-specific representative population is not straightforward.
There are also inherent challenges in how normal distributions function when applied to athlete performance.
Even in a population of relatively weak athletes, half of the group will still be characterized as stronger than average. Some athletes, by definition, fall into the upper percentiles regardless of whether the group itself is prepared for the demands of their sport.
Consider the implications.
The evidence suggests that female athletes have less access to high-quality strength training resources compared to their male counterparts. Evaluating strength using normative distributions derived from a population with a low training age may label athletes as “strong” when, in an absolute or sport-specific sense, meaningful strength development is still required.
The opposite problem also exists.
In a population of very strong athletes, normative distributions still require that half of the athletes be classified as weaker than average. In this case, athletes with excellent physical qualities may appear deficient simply due to the interpretation method.
The result is an unresolved question that normative benchmarking struggles to answer:
How good is good enough for athlete performance?
Even in large, representative distributions, the performance benchmark remains ambiguous. If an athlete’s maximal strength falls at the 50th percentile, is this sufficient? Or should training be prescribed to aim for the 75th percentile? The 90th percentile?
This logic is inherently flawed. Only a fixed number of athletes can occupy a given percentile, and not every athlete can or should be the strongest on the team.
These limitations compound when multiple tests and numerous performance metrics are used to characterize an athlete’s envelope of function across maximal strength, maximal power, and plyometric capacities.
From Isolated Metrics to Performance Profiles
Athlete profiling offers an alternative way to frame this problem.
Rather than interpreting each outcome metric in isolation using normative data, profiling focuses on the pattern expressed across multiple capacities. Metrics obtained from maximal strength, maximal power, and plyometric testing are not treated as separate silos, but as interacting components in the broader picture of performance.
This idea is not new.
Coaches have long recognized that distinct athlete phenotypes exist within similarly trained populations. In the early 2000s, D. J. Smith described this concept using four athlete phenotypes:
- The Thoroughbred: A fit and fast athlete who, assuming sufficient technical and tactical proficiency, has the highest likelihood of success in their sport.
- The Workhorse: A slow athlete with high fitness who performs well in the gym but may struggle to translate this fitness to sport-specific performance.
- The Bolter: A fast athlete with limited fitness who may experience short-term success but struggles to sustain long-term performance.
- The Wooden Horse: A slow, low-fitness athlete who may have a low training age or be returning to sport following injury.
What has been missing is a consistent, data-informed way to formally describe different athlete performance profiles within the context of multidimensional testing batteries.
Existing Profiling Tools: Useful, but Incomplete
Several profiling-based approaches already exist in applied sport science.
Before going any further, it is worth acknowledging an important point. Anytime practitioners consider more than a single metric together, they are already engaging in a form of profiling.
A simple example is the reactive strength index. In simple terms, RSI relates jump height to how long it took to achieve that height. Two athletes may jump to the same height, but the athlete who reaches peak height faster will exhibit a higher RSI.
RSI is a simple profile, but one with clear applications for sport performance and injury rehabilitation.
Radar plots represent another increasingly popular profiling approach. By plotting multiple testing metrics on a single figure, practitioners can generate a visual snapshot of an athlete’s performance profile, making large testing batteries easier to summarize and communicate.
Tools like ratios and visual profiles are useful. They represent a shift away from interpreting isolated metrics and toward a pattern-based approach.
But these methods also carry important limitations.
First, they do not account for the relative importance of different metrics when characterizing an athlete profile. The importance of specific testing metrics can vary substantially based on sport, position, age, and sex. These high-priority metrics are often referred to as key performance indicators.
Second, simple ratios and visual summaries do not account for complex, non-linear interactions between metrics. The scientific literature consistently shows that simply adding more isolated metrics does not meaningfully improve the prediction of sport performance or injury risk.
As a result, these tools are best viewed as a starting point for practitioners transitioning toward a profile-based interpretation of multidimensional performance data. They move us in the right direction, but there is still room to improve how athlete profiles are characterized.
Where We Go From Here
Emerging methods to profile athletes extend the logic of traditional profiling approaches but integrate larger sets of performance testing metrics, often 30 or more, to characterize more complex, non-linear, and weighted patterns of performance.
One way this has been achieved is through unsupervised machine learning, which can be used to identify subpopulations of athletes who express similar patterns across multiple neuromuscular capacities. These subpopulations can be interpreted as distinct athlete profiles or phenotypes.
The technical details of how to apply these methods in sport science deserve their own discussion. For now, what matters is why this approach helps resolve the problems associated with traditional performance testing interpretation.
When applied appropriately, profile-based analyses can add value in three practical ways.
First, communication. Profiles provide a structured way to summarize multidimensional testing data. Rather than reducing performance to a small set of isolated metrics, practitioners can describe a global pattern across capacities.
Second, benchmarking. Profile-specific benchmarks allow athletes to be compared based on how performance is achieved, rather than forcing all athletes into a single normative framework.
Third, prioritization. Profile-specific key performance indicators can help guide training emphasis or inform strategies aimed at shifting athletes toward different performance profiles when appropriate.
Athlete profiling does not replace traditional data interpretation methods. Instead, it provides the context needed to understand the pattern of capacities that define an athlete’s envelope of function and to support more informed training decisions.
If you found this useful, I explore these ideas further in McClean Performance Insights, a newsletter focused on applied sport science and real-world decision-making.
If you’re working with complex athlete data and want support translating it into actionable decisions, you can learn more about my consulting work below.