24 Jun Return to performance: machine learning insights into how absence time following muscle injuries affects match running performance in LaLiga soccer players
Returning to play is not the same as returning to performance.
This study moves beyond the classic medical clearance question and addresses something much more relevant for elite environments: how does absence time after a muscle injury affect match running performance in LaLiga players?
A total of 110 lower-limb muscle injuries from 90 professional players competing in LaLiga during the 2022–2023 season were analysed. For each injury, four pre-injury and four post-injury matches were compared. Then, machine learning models were applied to determine which performance metrics were most strongly associated with absence time.
The results are highly practical.
Not all performance variables behave the same after injury. And not all of them are equally influenced by how long the player was absent.
Maximal speed is the variable most affected by absence time.
Both regression models used in the study identified maximal speed as the strongest predictor associated with longer recovery periods. The multiple linear regression model showed a coefficient of -7.94 for maximal speed differences. The random forest model confirmed this finding through SHAP values, where maximal speed had the highest contribution to predicting absence duration.
In simple terms: the longer the absence, the greater the drop in maximal speed during matches.
This is not a trivial detail.
Maximal acceleration and maximal deceleration also decline as absence time increases. The correlations showed moderate negative relationships between absence time and maximal acceleration (r = -0.303) and maximal speed (r = -0.355). Deceleration capacity also demonstrated meaningful reductions.
However, one variable behaved differently.
Sprint count did not show a significant relationship with absence time.
Players performed fewer sprints after injury compared to pre-injury matches. But this reduction was not dependent on how long they were absent. A short absence and a long absence produced similar effects in sprint frequency.
This is a critical nuance for practitioners.
It suggests that maximal speed and acceleration/deceleration capacity deteriorate progressively with longer recovery periods, while sprint frequency may decrease regardless of absence duration.
The composite performance index, which summarizes acceleration-related actions, high-intensity running and medium-intensity actions, also declined after injury. Moreover, longer absence time was associated with greater reductions in this overall performance index.
In other words, prolonged recovery affects not only isolated speed metrics but global match output.
From a methodological perspective, the random forest regression model outperformed multiple linear regression, explaining 44.2% of the variance in absence time (R² = 0.442 in the test set). This reinforces the value of machine learning approaches in capturing non-linear relationships between recovery duration and performance loss.
But the key value of this paper is not methodological.
It is applied.
Two players with the same hamstring injury but different absence durations should not follow identical return-to-play pathways.
The player with a longer recovery timeline is more likely to show substantial reductions in maximal speed and acceleration/deceleration capacity during matches.
Therefore, return-to-play criteria should not be uniform.
For longer absences, practitioners should prioritize:
– Restoration of pre-injury maximal speed
– Restoration of maximal acceleration capacity
– Restoration of maximal deceleration ability
These capacities should be objectively tested before competitive reintegration.
Interestingly, the study also suggests that maximal acceleration deficits may partly explain why maximal speed is reduced during matches. In soccer, most sprints are short (10–30 m). If acceleration capacity is compromised, players may simply not have enough distance or time to reach their previous peak speeds.
This highlights the importance of assessing mechanical sprint properties, not just peak velocity.
Another relevant finding relates to deceleration capacity.
The machine learning model identified maximal deceleration as one of the most modifiable variables depending on absence time. This may be linked to tendon involvement in longer injuries. High-intensity braking actions rely heavily on eccentric and tendon capacity.
If rehabilitation does not adequately restore eccentric braking ability, players may return with compromised deceleration output. This has implications not only for performance but potentially for secondary injury risk.
The study reinforces the concept of return-to-performance.
Medical clearance does not guarantee performance restoration.
Players may return available but still exhibit meaningful deficits in maximal speed and acceleration capacity, especially after prolonged absence.
The practical message is clear.
Absence time is not just a medical variable. It is a performance variable.
The longer the recovery period after muscle injury, the greater the expected decline in maximal speed and acceleration/deceleration capacity during matches.
Rehabilitation must adapt accordingly.
Availability is the first step.
Performance restoration is the real objective.