How Training and Food Move the Curve

How does MaxDex model training load?

With a Banister two-component model. Each workout injects a training impulse that raises both acute fatigue, which decays fast, and chronic fitness, which decays slowly over a 288-hour constant. Net effect = fitness lift − fatigue drag.

The Banister systems model dates to 1975-76 and remains the standard impulse-response framework in training-load research: performance at any moment is the difference between a slow-decaying positive adaptation and a fast-decaying negative one. That single structure explains why you feel worse the day after a hard block and better a week later.

In our implementation the fitness lift is capped at 1.1 and weighted at 0.36, while fatigue drag is weighted at 0.45. Fatigue therefore outweighs fitness in the short term, which is the behaviour you want from a planner: it will not tell you a hard session made you sharper this afternoon.

Why does recovery time depend on the type of workout?

Because the physiology differs. Fatigue decays on a 4-hour constant for low-intensity aerobic work, 14 hours for moderate aerobic, and 30 hours for anaerobic work — with a separate 20-hour constant for central nervous system fatigue.

The split follows the two lactate turn points. Easy Zone-2-style work sits below the first turn point, produces minimal metabolite accumulation and minimal central fatigue, and clears within a few hours. Hard interval and resistance work pushes past the second turn point and drives sustained sympathetic activation and vagal withdrawal, which is why full autonomic restoration takes closer to a day or two.

Sleep debt stretches these constants by up to 50 per cent. Under-slept recovery is genuinely slower, and a model that ignored that would keep recommending hard sessions into a hole.

Why am I tired after eating?

MaxDex models a meal as a fuelling lift followed by a drag whose size depends on composition. Fast carbohydrate with little fibre or protein produces the largest dip; the effect is amplified by sleep debt and damped by movement.

The appearance curve uses a log-normal-style kernel whose time constant grows with the meal’s macronutrients — 0.6 hours as a base, plus 0.02 per gram of fat, 0.015 per gram of protein and 0.03 per gram of fibre. Fat, protein and fibre all slow gastric emptying, so a mixed meal releases more gradually and dips less sharply than the same calories as fast carbohydrate.

Protein also blunts the drag through a second mechanism: competition at the blood-brain barrier. Fernstrom and Wurtman established that brain serotonin synthesis is regulated by the ratio of tryptophan to other large neutral amino acids in plasma, and a protein-containing meal shifts that ratio away from the sleepiness-promoting direction that a pure carbohydrate load produces.

Why does sleep debt make food hit harder?

The meal drag is amplified by up to 50 per cent under accumulated sleep debt, because short sleep measurably reduces insulin sensitivity and impairs glucose tolerance.

Spiegel, Leproult and Van Cauter demonstrated this directly: restricting healthy young men to four hours in bed for six nights produced glucose tolerance in the pre-diabetic range, which recovered after sleep extension. The same lunch genuinely does hit harder on a sleep-deprived day, and the model reflects that rather than treating meals as context-free.

Movement damps the drag in the other direction, by up to 30 per cent — which is the modelled basis for the app suggesting a short walk after a heavy meal rather than as generic wellness advice.

Can it predict a crash before it happens?

Yes. When you log a meal, the model returns a crash risk of low, moderate or high, plus the number of hours until peak drag — for example, a sharp dip likely around 1.4 hours after eating.

This is the most directly useful output of the metabolic driver, because it is actionable in the window where action still helps: it lets the planner avoid putting your hardest work immediately downstream of a meal that is going to cost you, and it is why a walk suggestion arrives before the dip rather than during it.

How does daily movement factor in?

Movement contributes in two directions: an afternoon lift centred around 2pm, and a sedentary drag that ramps from 11:00 to 21:00 when step count falls well short of your target.

The drag is eased quadratically rather than linearly, so a modest shortfall costs very little and a severely sedentary day costs meaningfully more. The intent is to avoid nagging someone who walked slightly less than usual while still registering a day spent entirely at a desk.

Frequently asked questions