Applying ANF to SoccerMon Data: A First Look at Load, Wellness, and Readiness

 As a first independent proof of concept, I applied the Area of Normal Functioning — ANF — idea to open SoccerMon athlete-monitoring data.

This work continues the broader ANF direction I described in my earlier post:
https://www.trub.in/2026/06/expanding-area-of-normal-functioning.html

The data source for this experiment is the open SoccerMon dataset:
https://zenodo.org/records/10033832

The goal was not yet to build a final predictive model. The goal was simpler: to see whether ANF can be defined from real sport-science data and whether it can help us interpret training load, wellness, readiness, and possible overload patterns.

For this first experiment, I selected one player with relatively complete longitudinal data. I built two weekly IT-control-style charts:

1.      Weekly training load

2.      Weekly wellness/readiness risk

Both charts use the same ANF logic. For each week, the athlete’s current value is compared with the previous 12 weeks of history.

The control limits are percentile-based:

·         CL = rolling median, or P50

·         UCL = rolling P95

·         LCL = rolling P05

The area between LCL and UCL represents the athlete’s recent Area of Normal Functioning.

Weekly load as external demand

The first chart looks at weekly training load. I calculated weekly load as the sum of daily load values within each week.

In ANF language, this chart shows the athlete’s normal range of external training demand. When weekly load stays inside the ANF band, it means the athlete is training within the range that has recently been normal for her.

When weekly load rises above the UCL, it means the athlete is experiencing a higher load than expected based on her recent history. This may indicate overload, an aggressive training ramp-up, or a temporary excursion beyond the normal functional range.

Several above-UCL weeks appeared in the data, especially during periods of rapid increase in training load. Some of these high-load excursions occurred near weeks with injury reports. This does not prove causality, but it suggests that the upper ANF boundary may be useful as an overload-warning signal.

The LCL is also shown for consistency. However, low-load excursions are harder to interpret. A week below the LCL may simply represent rest, tapering, recovery, disrupted training, or off-season behavior. In this first example, below-LCL weeks did not appear to align with injury reports.

So for training load, the upper boundary appears more informative than the lower boundary.

Wellness/readiness risk as internal response

The second chart looks at a composite wellness/readiness risk index.

This index combines normalized values of:

·         fatigue

·         soreness

·         stress

·         inverse readiness

·         inverse sleep quality

·         inverse mood

Higher values mean worse subjective condition or lower readiness.

This chart represents the athlete’s ANF for internal response rather than external demand. It asks a different question: not “how much work did the athlete do?” but “how did the athlete appear to be responding?”

Several weeks were above the wellness/readiness UCL. In ANF terms, these weeks suggest that the athlete’s subjective state moved outside her recent normal range. This may indicate reduced recovery, higher fatigue burden, elevated stress, or a temporary maladaptive state.

Interestingly, in this first sample, elevated wellness/readiness risk did not directly coincide with injury-report weeks. That is still useful. It suggests that wellness ANF may capture short-term internal state changes, while load ANF may better capture external stress. They are related, but not identical.

Why both charts matter

The most important lesson from this first experiment is that athlete ANF is probably multidimensional.

A player can have normal training load but poor wellness.
A player can have high training load but still report good readiness.
A player can show subjective stress without immediate injury.
A player can experience injury after load patterns that look unusual.

That means ANF should not be defined by one metric alone.

In sport, ANF may need at least two dimensions:

1.      External load — what the athlete is asked to do.

2.      Internal response — how the athlete appears to absorb and adapt to that load.

This is very close to the broader ANF idea: systems do not only have inputs; they also have internal state, adaptation capacity, boundaries, and delayed responses.

First conclusion

This SoccerMon experiment is still preliminary, but it supports the basic idea that ANF can be operationalized on real human-performance data.

The ANF framework allows us to ask:

When is training load normal for this athlete?

When is subjective wellness outside the usual range?

Is the athlete adapting, recovering, or moving toward overload?

Are we seeing noise, anomaly, change, trend, or possible breakdown?

This is exactly why sport science may be a promising first application area for ANF beyond technical systems. Athletes are living dynamic systems. They have individual boundaries, adaptation capacity, stress responses, recovery patterns, and changing normal ranges.

The next step should be to repeat this analysis across multiple players and test whether sustained or repeated departures from ANF boundaries are associated with injury, illness, poor recovery, or performance decline.