Women's Soccer Injury Risk: How Fatigue and Multi-Week Load Predict Harm

Women's Soccer Injury Risk: How Fatigue and Multi-Week Load Predict Harm

Three recent studies on women's soccer injury monitoring converge on the same finding: accumulated training load over two to four weeks, combined with fatigue and recovery state, tracks more consistently with injury outcomes than the acute-to-chronic workload ratio that many programs currently emphasize. None of the studies show that changing monitoring practices prevents injuries. They identify which variables produced the stronger associations and, in one case, demonstrated the feasibility of injury forecasting using athlete monitoring data.

A January 2026 preprint applied a DeepHit machine-learning survival model to elite female football monitoring data and found that elevated fatigue ranked as a primary contributor on days flagged as highest injury risk, alongside elevated stress, low mood, and high-intensity running volume. Greater readiness and longer sleep showed modest protective effects, according to the preprint. The paper has not yet been peer reviewed.

A separate four-season observational study published in December 2025 found that injuries during menstruation were associated with more than three times the days-lost burden compared to injuries in non-bleeding phases, even though incidence rates between phases were similar, per Frontiers in Sports and Active Living. An NCAA Division I study found that injured players carried significantly higher accumulated player load and total distance over two-, three-, and four-week windows than uninjured teammates, while ACWR showed no association with injury in either rolling-average or exponentially weighted moving-average form, per research indexed on PMC.

How cumulative fatigue in women's soccer shows up over multiple weeks

The NCAA study compared injured and uninjured players on the same team across the same time windows using wearable GPS data. Injured athletes had significantly higher two-week (7,242 vs. 6,613 m/s²), three-week (10,533 vs. 9,718 m/s²), and four-week (13,819 vs. 12,892 m/s²) accumulated player loads than uninjured teammates, the research found. Total distance followed the same pattern: injured players covered more ground over two weeks (62.40 vs. 57.25 km), three weeks (90.97 vs. 84.10 km), and four weeks (119.31 vs. 111.38 km), with statistical significance across all three windows.

No significant difference in ACWR emerged between injured and uninjured groups in either calculation method. Sustained accumulation over multiple weeks was the variable that tracked with injury status, not the ratio between recent and chronic loads, the study reported.

A Division I severity study covering 1,560 athlete observations adds a finding worth separating out. Higher daily mileage was associated with 67% higher odds of progressing to a more severe injury or illness outcome (OR = 1.67, 95% CI: 1.19-2.34), according to the data. Structured training load pointed the other direction: each unit increase was associated with 42% lower odds of more severe injury or illness (OR = 0.58, 95% CI: 0.41-0.83). Session-RPE showed no significant association (OR = 0.96, 95% CI: 0.65-1.42). Raw mileage accumulation and structured training load produced opposite associations with severity in the same dataset.

That divergence matters for how programs interpret load data. A player accumulating high weekly mileage through unstructured running is not in the same situation as a player completing a similar volume through organized, progressive training. The research does not explain the mechanism behind that difference, but the association was present across 1,560 observations.

Fatigue and the broader risk picture in women's soccer injury forecasting

The January 2026 preprint used a 21-day input window and a 7-day prediction horizon. On a sample high-risk prediction day examined through SHAP explainability analysis, fatigue ranked as a front-line contributor alongside stress, mood, and high-intensity running volume, the authors found. The combined model architecture produced better forecasting performance than simpler configurations, and the authors describe the work as demonstrating feasibility of injury forecasting using athlete monitoring data rather than a deployable prevention tool.

The preprint also notes that a coach-facing interface would be needed before the model could move from research prototype to practical decision support. That interface does not currently exist.

A classification study in soccer found that pairing deviation-from-baseline metrics with ACWR features produced accuracy of 0.78, sensitivity of 0.73, and specificity of 0.85 for classifying injury-risk periods. ACWR features alone yielded accuracy of 0.76, functional but lower, per that research. The feature set that achieved the higher accuracy included deviation metrics for sprints, training load score, and time in high heart-rate zones, alongside ACWR features for total distance, high-speed running, sprint distance, and training load score.

That same study found surges in high-speed running and sprint volume, combined with elevated heart rate-based internal load metrics, may contribute to muscle injuries, the research noted. The type of work accumulated appears relevant alongside total volume.

Taken together, fatigue in this body of research is not functioning as a standalone input. It appears as a summary output of several interacting variables: training volume, psychological stress, sleep quality, and recovery state. When those variables deteriorate simultaneously, the predictive signal in the machine-learning model strengthened. That pattern has a practical implication for what programs choose to collect. External GPS load metrics capture volume and intensity. They do not capture fatigue, mood, stress, or readiness directly. Subjective wellness surveys do, and the DeepHit model treated those inputs as meaningful alongside the GPS data.

Menstrual cycle phase and injury severity: a separate question

The December 2025 Frontiers study addresses injury severity during menstruation, not whether players are more likely to be injured during that phase. Keeping those two questions distinct matters for how the finding should be read.

Across four seasons of data in elite female football, injury incidence was not significantly different between bleeding and non-bleeding phases. Injury burden during the bleeding phase was 684 vs. 206 days lost per 1,000 hours of exposure (p = 0.0027), the study found. Injuries sustained during menstruation were associated with substantially more severe consequences. That is a severity gap, not an incidence gap, and standard load monitoring does not capture it.

The four-season dataset makes this one of the larger longitudinal records available on this specific question in elite female football. The study does not prescribe an operational protocol, and the authors do not recommend removing players from activity based on cycle phase. What the finding does suggest, as an interpretation rather than a study conclusion, is that injury-management and return-to-play decisions made without cycle phase information may be working with an incomplete picture, particularly when estimating recovery timelines.

Programs that do not currently track menstrual cycle data have no mechanism for incorporating this dimension into injury management. Adding that information to athlete health documentation would not require new technology, though applying it usefully in return-to-play conversations would require medical staff familiar with the research.

What the studies found and where the evidence stops

Across these four studies, accumulated load over two to four weeks and daily recovery state showed more consistent associations with injury outcomes than week-to-week workload ratios. The NCAA work and the Division I severity study are observational, working with association data from single institutional datasets. The DeepHit preprint has not yet been peer reviewed and describes its contribution as demonstrating feasibility rather than confirming a prevention effect. The menstrual cycle study measured severity rather than incidence, across four seasons of elite-level data.

None of these studies show that tracking these variables changes injury rates. They identify which variables tracked most closely with injury outcomes in their respective datasets, which is a different and more limited claim.

For programs using GPS monitoring and wellness platforms, the association evidence points toward multi-week load accumulation and daily fatigue, stress, mood, and readiness as variables that appeared together in the stronger models, treated as an interpretation of the research rather than a study finding. For programs without GPS hardware, weekly running totals and brief pre-session self-reports approximate the same inputs. Whether either monitoring approach changes outcomes remains a question the current evidence has not answered.

The most useful near-term step for any program, based on what this research found rather than what it proves, is reviewing load accumulation over rolling two- to four-week windows alongside whatever wellness data is already being collected, rather than relying on the previous week's ratio alone. That does not require new software. It requires looking at a longer window and treating converging signals as a flag worth acting on before a player misses time.