There is no formula for a deload. The research gives ranges, and this tool applies those of the best-documented autoregulated models: Renaissance Periodization for volume reduction, Tuchscherer (RTS) for intensity reduction.
The starting point is not how many weeks have passed, it is how you actually are. The tool doesn't ask you to name your fatigue: that's the hard part, and picking the wrong axis gives you an uncomfortable week that recovers nothing. It infers it from your answers, separating two families of signals: what comes from accumulated sets (muscle recovery, soreness, joint pain) and what comes from nervous system load (sleep, energy, motivation, falling performance).
Joint fatigue comes from accumulated sets: cut volume in half and keep the loads. Nervous fatigue comes from heavy loads and sets close to failure: take 20% off the load and barely touch the volume. When both families are present, reducing only one axis leaves the other fatigue untouched.
The tool can say “no”. A single signal doesn't justify a deload, and deloading for no reason costs a week of progress, exactly like deloading too late. Pain that limits movement isn't a deload matter at all: it needs to be examined.
Reps never change. Changing three variables at once makes the return week unreadable: the following week, you no longer know what worked.
Duration depends on how long the last block lasted and on the type of fatigue. Central fatigue clears more slowly than local fatigue: beyond ten weeks without a deload, or when both signals are present, a single week is usually not enough.
Loads are rounded down to the nearest 2.5 kg. It is the only direction that guarantees the deload load stays below your working load.