Neurofeedback in Sport: Brain Training or Expectation Effect?

The key points in 30 seconds

• Neurofeedback gives an athlete feedback on a chosen brain signal and invites them to learn to change it. Studies often report improvements in attention, emotional regulation or motor performance.
• A new systematic review included 56 studies and 1,711 participants, but the median completeness of reporting by CRED-nf was only 37%.
• A positive result was often obtained relative to no intervention: only 18% of studies showed an advantage over an active control, and only 4% over other experimental conditions.
• This is not proof that the method is useless. It means it is still hard to separate the specific training of a brain signal from practice, coach attention, motivation, expectations and repeated performance of the sport test.

What exactly neurofeedback trains

During EEG neurofeedback, electrodes record brain activity. The software extracts a pre-selected measure — for example the power of the alpha rhythm or the sensorimotor rhythm — and almost immediately turns it into a sound, an image or a change in a game object. The participant's task is to learn to obtain the desired feedback and, supposedly, to regulate the corresponding state better.

In sport the goals differ. Some protocols aim at concentration before a precise action, others at controlling anxiety and arousal, others at motor learning. That is why the phrase "neurofeedback works" is too general: a protocol for a marksman's attention, for a swimmer's emotional regulation and for motor imagery are not the same intervention.

Why a positive result does not yet prove a specific effect

Imagine a group of athletes who go through ten unusual high-tech sessions, talk with a specialist, concentrate on the task and perform the same test several times. The control group simply carries on with ordinary training. If the first group improved more, the difference could be produced by a neurophysiological mechanism — but also by extra attention, motivation, the expectation of benefit, training in self-regulation or familiarity with the test.

A passive control answers the question: is the whole package of intervention better than nothing new? An active control receives comparable time, attention and procedure, but without the supposedly active component. Sham neurofeedback reproduces the interface, but the feedback does not match the participant's target signal. The latter variants test specificity better, although it is hard to create a plausible sham without side differences.

What the new review did

Izac and colleagues systematically grouped studies by the main sport goal: attention, emotional-affective abilities or motor abilities. The review included 56 studies and 1,711 participants, found in Scopus, Web of Science and PubMed. Reporting quality was assessed with CRED-nf — a consensus checklist of requirements for the design and description of neurofeedback studies.

Most of the studies reported positive neurophysiological, cognitive, emotional or motor results. However, the structure of the comparisons was less convincing:

• 34% of studies reported a positive effect relative to a passive control;
• 18% relative to an active control;
• 4% relative to other experimental groups.

These percentages cannot be added together and turned into a "probability of success". They show how rarely the literature answers the strict question of a specific mechanism. The median CRED-nf result was 37%, and the protocols differed greatly in brain markers, instructions and feedback format.

The authors' conclusion

The authors do not conclude that the whole effect is placebo. Their synthesis suggests a combination of specific components — learning to regulate the relevant activity — and non-specific ones, such as concentration and engagement. Some combinations occurred more often: for example, raising the sensorimotor rhythm in protocols aimed at attention. But a frequent association does not yet confirm optimality or a causal mechanism.

The main recommendation of the review is to design the intervention starting from the sport goal, to link it in advance to a measurable neurophysiological indicator, and to report fully according to CRED-nf.

Why earlier reviews looked more optimistic

The 2024 review by Tosti et al. included 24 studies and described improvements in athletic and cognitive performance, emotional regulation and anxiety. The meta-analysis by Yu et al., published later, pooled 21 studies of motor tasks and obtained Hedges' g = 0.78 (95% CI 0.49–1.07). At first glance this is a strong argument in favour of the method.

But that meta-analysis also shows the problem of control. Among the 25 well-described studies, 12 used a no-intervention condition, nine used sham feedback, five used ordinary training without neurofeedback, and three compared different neurofeedback protocols; the categories could overlap. None of the 22 studies that underwent risk-of-bias assessment had low risk in all domains, and double blinding was used in only one.

A positive mean effect and a weak ability to establish its specificity can coexist. The earlier review asked whether performance improves after a programme; the new one asks how far the design allows the improvement to be attributed to the targeted neurophysiological learning.

What CRED-nf is and why it matters

CRED-nf requires goals, protocol, signal processing, feedback, the control condition and the analysis to be described in advance. It is especially important to demonstrate two links:

1. whether the participant actually learned to change the target brain indicator;
2. whether that change is related to the improvement in behaviour or sport performance.

If the second outcome improved but the first was not checked, the study demonstrated an effect of the programme but did not confirm the claimed neurophysiological mechanism. Pre-registration and disclosure of all outcomes also reduce the risk of choosing only the successful test for publication.

How reliable are today's conclusions

The volume of literature can no longer be called negligible, but its heterogeneity is great. Sports, athlete level, number of sessions, electrode placement, frequency bands, instructions and success criteria all differ. Many studies have small samples and short interventions, and it is unclear whether the effect persists.

There is also a practical measurement problem: an improvement in the accuracy of a few shots or in reaction time does not always transfer to competition. The result of a novice in the laboratory cannot automatically be transferred to an elite athlete, who has less room for improvement and a more complex competitive environment.

Practical significance for athletes and coaches

Neurofeedback is reasonably regarded as an experimental addition rather than a replacement for technical, physical and psychological preparation. Before buying a course it is useful to ask:

• which specific sport goal has been chosen;
• why precisely this EEG indicator is being trained;
• whether the ability to actually change it is measured;
• whether there is a result in the sport task, and not only on the screen;
• what the programme was compared with — expectation, an active control or sham;
• how long the effect lasts after the course ends;
• whether the specialist's qualifications and any commercial interest are disclosed.

If a programme promises a guaranteed improvement in results but cannot explain the protocol and show controlled data, the technological look of the interface does not compensate for the weakness of the evidence.

SportinLux conclusion

Neurofeedback may help some athletes, but the current literature more confidently supports the effect of a complex procedure than the specificity of a particular brain mechanism. The new review does not refute the positive results; it shows why they cannot be unreservedly attributed to training the chosen EEG rhythm.

What has not yet been proven

• No universally best protocol has been defined for any specific sport.
• It is unclear what proportion of the effect persists after training ends.
• There are too few rigorous sham- and active-controlled studies in elite athletes.
• It has not been shown that an improvement in a laboratory measure reliably raises competitive performance.
• It has not been established which athletes are most likely to respond to the intervention.