Every training platform will give you a load number. It compresses duration and intensity into a single figure, stacks those figures into a trend, and offers conclusions about whether you are building fitness or digging a hole.

The models are genuinely useful. They are also narrower than the confident dashboards suggest, and one of the most popular has been seriously challenged since it became popular.

TRIMP: load from heart rate

TRIMP — training impulse — goes back to work by Eric Banister in the 1970s. The core idea is simple: multiply how long you trained by how hard, where "how hard" comes from heart rate.

The naive version is duration × average heart rate, which badly undervalues hard sessions. Refined versions weight the heart rate term exponentially, so time near threshold counts for much more per minute than easy running. Most implementations use heart rate reserve rather than raw heart rate, so the number is individualised to your resting and maximum rates.

TRIMP's strength is that it works for any activity where you wear a heart rate strap. Its weaknesses are heart rate's own weaknesses, which are covered in the heart rate zones article: it lags behind effort, so short intervals are systematically undervalued, and it drifts upward in heat, so a hot easy run scores as a harder session than it was.

TSS: load from power

Training Stress Score came out of cycling, where power meters made intensity directly measurable rather than inferred. It is built on normalised power relative to your functional threshold power, scaled so that one hour at threshold equals 100 TSS.

That anchor is what makes TSS interpretable. A 200 TSS week and a 600 TSS week mean something specific, and the scale is consistent between athletes because each is measured against their own threshold.

Because it uses power rather than heart rate, TSS has no lag and no drift. It reflects the work done rather than the body's delayed response to it, which makes it markedly better for interval sessions.

Running has equivalents — rTSS derived from pace against threshold pace, and hrTSS derived from heart rate where no pace or power data exists. These are approximations of an approximation, and the further you get from measured power against a known threshold, the more the number is a model rather than an observation.

Running power deserves a specific warning. Unlike cycling power, which is measured directly at the crank or pedal, running power is estimated from motion and terrain by a proprietary model. Different vendors produce different numbers for the same run, and there is no shared standard. Running power is useful as a consistent relative signal within one system; it is not comparable across systems the way cycling watts are.

CTL, ATL and form

Single-session load becomes a training picture through three derived series:

  • CTL (chronic training load) — an exponentially weighted average over roughly six weeks. A proxy for accumulated fitness.
  • ATL (acute training load) — the same over roughly one week. A proxy for current fatigue.
  • TSB (training stress balance, or "form") — CTL minus ATL. Positive means fresher, negative means carrying fatigue.

The underlying logic is sound and old: fitness accumulates slowly and decays slowly, fatigue accumulates quickly and clears quickly, and performance reflects the gap. This is why a taper works — cutting volume drops ATL much faster than CTL, so form rises while fitness barely falls.

Used as a planning tool, this is genuinely valuable. Used as a scoreboard, it invites the mistake of chasing CTL as a goal in itself, which mainly produces a lot of moderate-intensity volume that raises the number without making anyone faster.

The acute:chronic workload ratio, and why to be careful

ACWR divides acute load by chronic load. The widely repeated version says that a ratio between roughly 0.8 and 1.3 is a "sweet spot" and that going above about 1.5 sharply raises injury risk. It spread quickly after being popularised in sports science literature in the 2010s, and it now appears in consumer apps as settled fact.

It is not settled. The concept has been substantially criticised in the years since, on grounds that are methodological rather than ideological: the ratio shares a term between numerator and denominator, which can generate apparent relationships as a statistical artefact; several influential analyses have been questioned over their handling of that and other issues; and attempts to reproduce the injury-risk findings have been inconsistent.

The sensible position is neither "ACWR predicts injury" nor "ACWR is worthless". It is that sharp increases in training load are worth noticing, which was sound coaching advice long before anyone computed a ratio, and that the specific thresholds carry far less authority than the confident colour-coding in apps implies.

If your app turns a number red, treat it as a prompt to think about the last fortnight, not a diagnosis.

What no load model sees

Every model on this page is built from duration and cardiovascular intensity. The list of things that stress an athlete and are absent from that is long:

Mechanical load. Thirty minutes of downhill running and thirty minutes on a flat road can score identically while doing very different things to your legs. Eccentric loading is where much of the muscle damage in running comes from, and no heart-rate or power model registers it.

Strength training. A heavy session may barely raise heart rate and still be one of the most demanding things you do that week. Most load models score it near zero.

Heat, altitude and illness. These inflate heart-rate-derived load for the same work, so a hot week looks harder than it was in training terms while genuinely being harder in other ways.

Sleep and life stress. Recovery is not a function of training alone. A week with three bad nights and a work deadline is a different week at identical load.

Cumulative history. CTL looks back about six weeks. A runner with ten years of consistent base and one returning after a year off can show identical CTL and are not in the same place.

Using load numbers well

  • Watch the trend, not the value. Whether your CTL is 60 or 90 matters less than whether it has been climbing steadily or spiking.
  • Compare yourself to yourself. Load numbers are not comparable between athletes, and with running power they are not comparable between devices either.
  • Log how you felt. A session rated hard that scored low, or the reverse, is information the model does not have. Over a season that subjective record catches things the number misses.
  • Let the model lose. If load says you are fresh and you feel wrecked, you are wrecked. The model is a summary of two variables; you have access to all of them.

The value of these tools is that they make a training block legible at a glance and stop you fooling yourself about a week that felt hard but was not. That is worth having. It is just not the same thing as knowing how your body is doing.