Seven things to look for in an elite athlete’s performance |

Seven things to look for in an elite athlete’s performance - and the hype to ignore

Prefer listening? Play the audio version:

Walk any sports-tech trade floor or scroll any industry newsletter and you’ll find the same story on repeat: bigger market, more sensors, AI everywhere, the future is now. Most of it reads like it was assembled from vendor press releases. We wanted to write the opposite of that: a list put together the way a performance department actually evaluates kit, which is to say sceptically, with one eye on the budget and the other on whether the thing survives contact with a real training week.

So, this is less a map of what’s new and more a working view of what’s worth your attention in 2026 — and, just as important, what isn’t. It’s the introduction to a seven-part series; each piece that follows goes deep on one area, names the teams using it, reads the actual evidence rather than the marketing, and tries to answer the only question that matters to a head of performance: would this change a decision I make on a Monday morning?

A word on the framing before we start. The honest through-line across all seven areas is not “AI is transforming sport.” It’s quieter and more useful than that.

The bottleneck in elite performance stopped being collecting data a long time ago; most departments are drowning in it (Weaving et al., 2019; Wilson & Kiely, 2023). Marco Cardinale, now Executive Director of Research at Aspetar and the man who ran the science for Team GB across three Olympic cycles, has described the problem as a “data tsunami”, one that forces sports scientists to find new ways “to analyse and visualize information to be able to improve decision-making” (Cardinale, 2021). That is the real frontier: not gathering more, but turning what we already have into a decision someone trusts enough to act on. On that measure, most of the technology below is still some distance from earning its place. Keep the tsunami in mind for everything that follows.

1. AI for injury and load — useful, oversold, occasionally dangerous

Start with the area carrying the most hype, because it’s the one most likely to separate good departments from credulous ones.

The pitch is seductive: feed a model your GPS, your wellness questionnaires and your medical history, and it tells you who’s about to get hurt. The reality is more sobering. One of the most visible industry case studies comes from Zone7, which retrospectively analysed workload and injury data across 11 professional football teams. The system flagged increased risk one to seven days before 306 of 423 recorded injuries, a detection rate of 72.4% (Zone7, 2022). That figure is interesting, but it should not be read as proof that the model can “predict injuries” in the way vendors often imply. The study was retrospective, vendor-led and not designed as peer-reviewed scientific research. More importantly, sensitivity alone does not tell a performance department how useful the tool will be day to day. To judge that, you still need specificity, alert frequency, positive predictive value, prediction window and evidence that acting on the alerts actually reduces injury burden without disrupting training.

The deeper problem is what’s under the bonnet. A lot of these models still lean on workload features descended from the acute:chronic workload ratio, a metric that drove a decade of practice before Franco Impellizzeri and colleagues showed it was mathematically leaky and unstable when applied forward rather than backward. Dropping a gradient-boosted model on top of a flawed input doesn’t launder the flaw. And the academic community has been blunter than the vendors: one widely cited paper argued that black-box injury-prediction methods deserve “a red card for reckless practice.” The interesting work in 2026 isn’t a cleverer algorithm — it’s the unglamorous plumbing of getting medical, GPS and tactical data into one place so a human can reason across it. Watch cycling here. The WorldTour is among the most sensor-saturated environments in sport — every rider produces continuous power, heart-rate and positional data — and in 2026 it turned into a live experiment in what to actually do with all of it. INEOS rebranded as the Netcompany INEOS Cycling Team around a five-year partnership built on PULSE, a real-time AI platform the Danish firm had previously deployed to run things like Heathrow and Munich airports. Within the same window, Visma–Lease a Bike announced its own tie-up with the AI lab Mistral. Two of the sport’s strongest teams are now openly betting that the edge lies in integrating data faster than the rival — turning power, weather, position and logistics into one decision in real time — rather than in collecting more of it. It’s a genuine test of the integration thesis, and the result will play out on the road over the next two seasons. Worth watching precisely because, for once, we’ll get to see whether it works.

The take: treat an AI risk flag as a reason to go and look at someone, never as a verdict. And if a vendor can’t tell you the positive predictive value of their model at your squad’s base rate — not the accuracy, not the sensitivity, the PPV — you’re not being sold a tool. You’re being sold a story.

2. Wearables — the easy problem is solved, the hard one isn't

“The sensor war is over” is too neat — anyone who’s tried to get clean GPS data indoors, or trusted a non-invasive lactate reading, will tell you the hardware still has plenty to prove. But for the metrics most departments actually live on — heart rate, sleep, gross external load — the devices are good enough, and have been for a while. The validation literature has matured to the point where we can be specific about what they do and don’t do well, and “well” depends entirely on which metric you ask about. Take sleep. An Oura-funded study against polysomnography found the ring nailed sleep-versus-wake detection, with sensitivity above 95%, but slipped to the high 70s when sorting actual sleep stages; rival devices in the same study fared worse, with one badly underestimating deep sleep(Robbins et al., 2024). That’s the pattern: good enough to track trends over time, not good enough to treat as a diagnostic(Cosoli et al., 2021; Herberger et al., 2025).

The unsolved problem is everything that happens after the data lands. A number only means something once you know its noise and its smallest worthwhile change — a discipline sports scientists have understood for years(Schneider et al., 2018) and the consumer-wearable market routinely ignores. A morning readiness score that’s “down” might be signal, or it might be the normal day-to-day wobble of the measure, or it might be that the athlete slept badly because of a newborn. Strip away that context and the score is theatre — and athletes learn fast to game a metric they don’t trust.

So, the real divide in 2026 isn’t between brands. It’s between departments where someone’s actual job is to turn wearable output into a specific coaching or medical decision, and departments where the kit was bought on impulse, lived on a dashboard nobody opened, and was quietly forgotten within a season. Most people reading this have seen the second version up close. Sam Robertson — who has worked across the AFL, FIFA and the San Antonio Spurs — recently helped set up an institute aimed at validation standards for sports technology, on the view that the industry should be held to the same account we hold our athletes to(Robertson et al., 2023). He’s right that the kit has outrun the scrutiny. Before you buy another sensor, the only question that matters is who, by name, will own the interpretation. If the answer is “the platform,” you haven’t bought a solution. You’ve bought another dashboard.

3. Recovery and sleep — the one area where the science got ahead of the marketing

Of everything on this list, recovery is the area where we’d argue the evidence is strengthening on its own merits rather than being retrofitted to justify a product. The most useful recent finding is also the least commercial: a 2024 study using hamstring muscle biopsies on professional footballers found that key markers of performance and injury risk had not recovered three days after a match(Carmona et al., 2024). The textbook 48-hour recovery window, for several of the things that actually get players hurt, looks closer to wishful thinking than physiology.

Sit with the implication, because it’s bigger than any gadget. If hamstrings aren’t right at 72 hours, then the congested fixture calendar — midweek cups, international windows, Thursday-Sunday turnarounds — is built on a physiological assumption the data no longer supports. No temperature-controlled mattress or red-light panel fixes that. The sleep tech and the photobiomodulation kit have their place at the margins, and some of it has reasonable evidence behind it, but the headline isn’t the hardware.

Recovery has become a scheduling and load-planning discipline. The single highest-value recovery decision you’ll make this season is probably a fixture you choose not to play someone in — and no device on the market will make that call for you.

4. Computer vision — the most underrated shift on this list

While everyone argued about AI chatbots, optical tracking moved from telling you where a player was to telling you what every joint was doing, in three dimensions, from ordinary camera feeds. That’s a genuine step change, and the same fundamental capability that powers elite broadcast tracking is starting to reach clubs that could never afford a marker-based lab — though how far down the pyramid it has actually travelled is something we examine properly in a later piece.

The honesty required here is about the gap between the lab and the pitch. In controlled conditions, markerless motion capture gets close to lab-grade for joint angles. Drop it into a live competition — fast, ballistic, awkward lighting — and the agreement with proper measurement can collapse. One validation at a real athletics meet found pose-estimation tracking so unreliable for take-off mechanics that the authors concluded it did not allow for valid quantification(Cronin et al., 2024). The technology is real and improving; the “screen every athlete’s biomechanics from match footage” promise is not there yet.

The take: vision is the area most likely to change your practice in the next two years. Pilot it on something specific and measurable, and don’t believe the in-competition accuracy claims until you’ve checked them on your own footage.

5. The mind — two different things wearing one label

Lumping “brain and mind performance” into a single category is how budgets get wasted, because it’s really two things, and they’re not close.

One half — athlete mental health — has crossed a real threshold. After a decade of athletes speaking openly and leagues mandating clinical staffing, support that was once an afterthought is becoming infrastructure. Fund it without hesitation; the limiting factor now is whether athletes use the services, not whether they exist.

The other half — the EEG headbands and brain-stimulation devices sold for “focus” — is where scepticism earns its keep. The most careful recent meta-analysis of transcranial direct-current stimulation on athletic performance found a real but small effect, swamped by enormous variation between studies and devices (Winker et al., 2024). A broader umbrella review was blunter still, concluding that once you account for publication bias and analytical choices, the evidence does not reliably support tDCS improving performance at all (Holgado et al., 2024). That’s not nothing — but it’s a long way from a research-grade lab finding to a consumer headband on a wet training pitch, and the marketing collapses that distance for you. Treat the whole neuro-gadget category as an experiment with a clear exit criterion — not as performance kit, and certainly not as a budget line you defend two years running because you’ve already sunk money into it.

6. FemTech — the biggest gap between current and best practice in all of sport

For most of the modern era, “the athlete” in sports science was implicitly male — the protocols, the load models, the injury research, the device validation cohorts. Closing that gap is, in our view, the single largest unrealised performance opportunity in the field right now, and it’s moving fastest where women’s teams are being built new rather than retrofitted onto male templates.

A note of discipline, though, because this area attracts as much hype as it deserves attention. Female footballers tear their ACLs at something like 1.5 to 1.7 times the male rate — and depending on the sport, the cohort and how you count, other studies put the disparity higher still. The exact multiplier is genuinely contested; that the gap is real and substantial is not(Montalvo et al., 2019). But the popular claim that performance swings dramatically across the menstrual cycle is not well supported: the strongest meta-analysis found only trivial differences in performance between cycle phases(McNulty et al., 2020). The honest case for cycle-aware practice is about managing load, recovery and symptoms intelligently, and closing a genuine data gap — not about predicting a 30% performance dip on a given Tuesday. Overclaim here and you hand sceptics a reason to dismiss the whole field, which would be the worst outcome.

The take: if your female athletes are still training on protocols built from male physiology, that’s a bigger, cheaper performance win than anything else on this list. Just make the case on the real evidence, not the inflated version.

7. GenAI and immersive training — automating the analyst, not the coach

The last area splits cleanly. Generative AI is starting to absorb the lower-value end of analysis — tagging clips, cutting footage, drafting opposition summaries — and where it works, that’s a real efficiency, freeing analysts for the judgement work software can’t do. VR for decision-training and rehabilitation has earned a place in some elite environments(Richlan et al., 2023; Yun-chao et al., 2023). While some evidence suggests transfer of physical skill from VR to real-world, its generalizability and robustness are still under investigation(Juliano & Liew, 2020; Markwell et al., 2023).

The bottleneck across all of it is the same one from area one. The compute works. The models work. What doesn’t yet work is a coaching staff trusting a recommendation they can’t interrogate. That’s a human problem, and no amount of model improvement solves it on its own.

The take: let AI take the grunt work off your analysts. Be far more careful before letting it near a decision your coaches are accountable for.

Where this leaves us

If there’s a single thread, it’s that the winners over the next two years won’t be the clubs with the most kit. They’ll be the ones who got disciplined about the question every vendor would rather you didn’t ask: does this change a decision, and can we trust it enough to act? Most of the technology above is genuinely useful in the hands of a department that asks that question relentlessly, and genuinely wasteful in the hands of one that doesn’t.

The deep-dives that follow take each area in turn, with the named clubs, the actual numbers, and an honest read of where the evidence runs out. We’d rather tell you what we don’t know than sell you certainty we can’t back — and if you’re working at the sharp end of this and disagree with any of it, we genuinely want to hear from you. That conversation is the point.

References

Bullock, G. S., Mylott, J., Hughes, T., Nicholson, K. F., Riley, R. D., & Collins, G. S. (2022). Black box prediction methods in sports medicine deserve a red card for reckless practice: A change of strategy is needed. Sports Medicine, 52(8), 1729–1735. https://doi.org/10.1007/s40279-022-01655-6

Cardinale, M. (2021). The sport scientist perspective on the use of wearable sensors in sport and exercise [Conference presentation]. IEEE. Retrieved from https://www.marcocardinale.com

Carmona, G., Moreno-Simonet, L., Cosio, P. L., Astrella, A., Fernández, D., Cadefau, J. A., Rodas, G., Jou, C., Milisenda, J. C., Cano, M. D., Arànega, R., Marotta, M., Grau, J. M., Padullés, J. M., & Mendiguchia, J. (2024). Hamstrings on focus: Are 72 hours sufficient for recovery after a football (soccer) match? A multidisciplinary approach based on hamstring injury risk factors and histology. Journal of Sports Sciences, 42(12), 1130–1146. https://doi.org/10.1080/02640414.2024.2386209

Cosoli, G., Scalise, L., Poli, A., & Spinsante, S. (2021). Wearable devices as a valid support for diagnostic excellence: Lessons from a pandemic going forward. Health and Technology, 11(3), 673–675. https://doi.org/10.1007/s12553-021-00540-y

Cronin, N. J., Walker, J., Tucker, C. B., Nicholson, G., Cooke, M., Merlino, S., & Bissas, A. (2024). Feasibility of OpenPose markerless motion analysis in a real athletics competition. Frontiers in Sports and Active Living, 5, 1298003. https://doi.org/10.3389/fspor.2023.1298003

Herberger, S., Aurnhammer, C., Bauerfeind, S., Bothe, T., Penzel, T., & Fietze, I. (2025). Performance of wearable finger ring trackers for diagnostic sleep measurement in the clinical context. Scientific Reports, 15, 93774. https://doi.org/10.1038/s41598-025-93774-z

Holgado, D., Sanabria, D., Vadillo, M. A., & Román-Caballero, R. (2024). Zapping the brain to enhance sport performance? An umbrella review of the effect of transcranial direct current stimulation on physical performance. Neuroscience & Biobehavioral Reviews, 164, 105821. https://doi.org/10.1016/j.neubiorev.2024.105821

Impellizzeri, F. M., Tenan, M. S., Kempton, T., Novak, A., & Coutts, A. J. (2020). Acute:chronic workload ratio: Conceptual issues and fundamental pitfalls. International Journal of Sports Physiology and Performance, 15(6), 907–913. https://doi.org/10.1123/ijspp.2019-0864

Juliano, J. M., & Liew, S.-L. (2020). Transfer of motor skill between virtual reality viewed using a head-mounted display and conventional screen environments. Journal of NeuroEngineering and Rehabilitation, 17(1), 48. https://doi.org/10.1186/s12984-020-00678-2

Markwell, L. T., Cochran, K., & Porter, J. M. (2023). Off the shelf: Investigating transfer of learning using commercially available virtual reality equipment. PLOS ONE, 18(10), e0279856. https://doi.org/10.1371/journal.pone.0279856

McNulty, K. L., Elliott-Sale, K. J., Dolan, E., Swinton, P. A., Ansdell, P., Goodall, S., Thomas, K., & Hicks, K. M. (2020). The effects of menstrual cycle phase on exercise performance in eumenorrheic women: A systematic review and meta-analysis. Sports Medicine, 50(10), 1813–1827. https://doi.org/10.1007/s40279-020-01319-3

Montalvo, A. M., Schneider, D. K., Yut, L., Webster, K. E., Beynnon, B., Kocher, M. S., & Myer, G. D. (2019). “What’s my risk of sustaining an ACL injury while playing football (soccer)?” A systematic review with meta-analysis. British Journal of Sports Medicine, 53(21), 1333–1340. https://doi.org/10.1136/bjsports-2016-097261

Richlan, F., Weiß, M., Kastner, P., & Braid, J. (2023). Virtual training, real effects: A narrative review on sports performance enhancement through interventions in virtual reality. Frontiers in Psychology, 14, 1240790. https://doi.org/10.3389/fpsyg.2023.1240790

Robbins, R., Weaver, M. D., Sullivan, J. P., Quan, S. F., Gilmore, K., Shaw, S., Benz, A., Qadri, S., Barger, L. K., Czeisler, C. A., & Duffy, J. F. (2024). Accuracy of three commercial wearable devices for sleep tracking in healthy adults. Sensors, 24(20), 6532. https://doi.org/10.3390/s24206532

Robertson, S., Zendler, J., De Mey, K., Haycraft, J., Ash, G. I., Brockett, C., Seshadri, D., Woods, C., Kober, L., Aughey, R., & Rogowski, J. (2023). Development of a sports technology quality framework. Journal of Sports Sciences, 41(22), 1983–1993. https://doi.org/10.1080/02640414.2024.2308435

Schneider, C., Hanakam, F., Wiewelhove, T., Döweling, A., Kellmann, M., Meyer, T., Pfeiffer, M., & Ferrauti, A. (2018). Heart rate monitoring in team sports—A conceptual framework for contextualizing heart rate measures for training and recovery prescription. Frontiers in Physiology, 9, 639. https://doi.org/10.3389/fphys.2018.00639

Weaving, D., Beggs, C., Dalton-Barron, N., Jones, B., & Abt, G. (2019). Visualizing the complexity of the athlete-monitoring cycle through principal-component analysis. International Journal of Sports Physiology and Performance, 14(9), 1304–1310. https://doi.org/10.1123/ijspp.2019-0045

Wilson, M. R., & Kiely, J. (2023). Developing decision-making expertise in professional sports staff: What we can learn from the Good Judgement Project. Sports Medicine – Open, 9(1), 69. https://doi.org/10.1186/s40798-023-00629-w

Winker, M., Hoffmann, S., Laborde, S., & Javelle, F. (2024). The acute effects of motor cortex transcranial direct current stimulation on athletic performance in healthy adults: A systematic review and meta-analysis. European Journal of Neuroscience, 60(5), 5086–5110. https://doi.org/10.1111/ejn.16488

Yunchao, M., Mengyao, R., & Xingman, L. (2023). Application of virtual simulation technology in sports decision training: A systematic review. Frontiers in Psychology, 14, 1164117. https://doi.org/10.3389/fpsyg.2023.1164117

Zone7. (2022). Injury risk forecasting: Retrospective validation across professional football teams [Industry report]. Zone7. Retrieved from https://www.zone7.ai