Readiness and strain scores feel like readings but are estimates built on assumptions about heart rate, duration, and recovery. Knowing the model behind the number is the only way to judge what it is worth.

A number that feels measured

Training load, readiness, and strain scores arrive looking like the heart rate or distance beside them, as if a sensor had read them off the body. They were not. They are the output of a model that takes measured inputs and combines them according to assumptions about how effort and recovery work. The distinction is easy to miss and important to keep in mind, because measured and modeled numbers deserve different kinds of trust.

What goes into the model

Most load models start from inputs that genuinely are measured, heart rate over time, duration, sometimes power or pace, and sometimes sleep or heart rate variability. They then apply a formula that weights recent effort against accumulated fatigue to produce a single figure. The inputs are real; the way they are combined is a designed choice that reflects a theory of training, not a law of nature.

Why the assumptions matter

Two devices can read identical heart rate and duration yet report different load scores, because they weight the inputs differently. Neither is wrong in an absolute sense; they encode different assumptions. This is why load numbers are most useful as a consistent signal within one system over time rather than as an absolute value to compare across brands. The trend a model produces is usually more trustworthy than its specific magnitude.

Using modeled metrics well

A modeled score is a useful summary when its limits are understood. Watching how a readiness figure moves week to week within one system can surface patterns a raw number would obscure. Treating that same figure as a precise, transferable measurement invites overconfidence. The practical skill is reading modeled metrics as informed estimates that compress many inputs, not as direct readings of the body's state.

Transparency is the differentiator

The most useful systems are open about what their scores represent and roughly how they are computed, so a user can judge whether the model's assumptions fit their situation. Opaque scores that cannot be interrogated ask for trust they have not earned. This site describes how such metrics are constructed in neutral terms, to help readers calibrate their confidence rather than to recommend any particular product.