Presets
Playback
Calendar: one cell per day, bar = load (blue build, orange taper, gold race day), underline green when fitness exceeds fatigue. Phase plane: the (fatigue, fitness) path; diagonals are lines of equal performance. Teaching model, not a coaching prescription.
Plan shape
Plan and model parameters
Layers
Model check
The science
Training-theory model, not a coaching prescription. Parameters are typical orders of magnitude from the sports-science literature, and the load unit (AU) and performance index are arbitrary. Individuals differ widely; this page is for understanding why tapering works, not for planning your own training.
The Banister fitness-fatigue model
Each training session of load w_s adds a slow, positive "fitness" effect and a fast, larger "fatigue" effect. In daily steps: p_n = p0 + k1 ∑ w_s e^(-(n-s)/τ1) − k2 ∑ w_s e^(-(n-s)/τ2) with τ1 (fitness, typically 30-60 days) much longer than τ2 (fatigue, typically 5-15 days) and k2 > k1. This is a linear, time-invariant impulse-response system. Early in a block fatigue dominates, so performance dips; with rest the fast fatigue clears first and performance rises above baseline. The page computes it with the recurrence F_(n+1) = e^(-1/τ)(F_n + w_n).
Analytic checks
- Single session: net response
k1 e^(-t/τ1) - k2 e^(-t/τ2)peaks att* = τ1τ2/(τ1-τ2) ln(k2τ1/(k1τ2)). - Optimum taper (constant load for B days, then complete rest x days): race-day performance is
p0 + w (a1 e^(-x/τ1) - a2 e^(-x/τ2))witha_i = k_i (1 - e^(-B/τ_i)) / (e^(1/τ_i) - 1), maximised atx* = τ1τ2/(τ1-τ2) ln(a2τ1/(a1τ2)). The page scans taper lengths by simulation and compares the best integer with x*. - Steady state: constant load w for ever gives
p0 + w (k1/(e^(1/τ1)-1) - k2/(e^(1/τ2)-1)). - The recurrence is also compared with a direct double-sum convolution of the load history.
Simplifications
The model is linear: no diminishing returns, injury, illness, sleep, nutrition, session type or psychological state; real athletes have changing parameters that must be fitted from data and fits are often ill-conditioned. The phase plane plots the two model states against each other; iso-performance lines are the diagonals fitness - fatigue = constant, and the ring marks the steady state a constant load would reach. The maths runs in JavaScript.