How TrainingPlan Chooses Training Phases

TrainingPlan's phase schedule is driven by the available planning window and goal type. The implementation reviewed for this article does not wait for an aerobic-efficiency plateau or a recovery score before moving from base to build.

That distinction matters when you read a generated plan: a phase label describes the scheduled emphasis, not a measurement of your body's readiness. Here is a reproducible example of what the software actually calculates.

Phase Labels and Weekly Sessions

For the general idea of organizing training emphasis over time, see periodization for normal runners. This article focuses on TrainingPlan's mechanism.

The current code uses BASE, BUILD, PEAK, TAPER and MAINTENANCE labels. Their order and duration come from scheduling rules. Weekly session targets are then calculated using the phase alongside other inputs, such as running-day count and runner-level rules.

A label alone is not a complete training prescription. For example, the PEAK label in the example below denotes the final week after the taper. It is not a finding that fitness has reached a peak. Longer planning windows can also contain additional cycles or an intermediate taper/peak block; the example is not a universal pattern for every plan.

A Worked Planning-Window Example

Synthetic software example, checked September 7, 2026. We ran the current phase-scheduling function with a half-marathon goal and 12 total planning weeks, then changed only the available weeks to 10. No customer data, fitness measurements or account edits were used.

What changed: the base allocation fell from four to three weeks and the build allocation from five to four. The two taper weeks and final peak week remained.

Why: for these inputs, the scheduler reserves two taper weeks for the half-marathon goal and one final week. It divides the remaining window between base and build using its allocation rules. Twelve weeks leave nine for that split; ten leave seven. This calculation explains the changed phase boundaries without assuming that the runner adapted faster.

In the weekly-target builder, the number of weeks is counted from the planning window, through the goal-race week when a race date is available. A changed date can therefore change the input to phase scheduling when a plan is generated. The example directly exercises that scheduling input; it does not demonstrate that editing a date in an account instantly replaces all saved sessions.

What the example does not establish: that ten weeks are sufficient preparation, that compressing a plan is appropriate after illness, or that either output improves race results. Those questions require information beyond the phase allocator.

Automatic Calculations and User Inputs

“Automatic” below means calculated when the relevant generation code runs. It does not mean every saved plan is continuously rewritten in the background.

The plan-adjustment guide explains how to separate inputs, proposed changes and review. That is a different task from reading the phase calendar.

Missing or Noisy Data

An incorrect goal or date can produce an internally consistent schedule for the wrong planning problem. A correct total-week calculation cannot fix that input error.

This phase allocator does not use sleep or heart-rate variability, so missing those measurements is not evidence that it can infer recovery from something else. Likewise, a noisy activity record and a scheduled phase transition should not be interpreted together as proof that a phase's physiological purpose has been achieved.

Estimated session duration is another separate calculation: the duration helper uses an explicit duration in the planned workout details when available, otherwise distance and pace or fallback estimates. A duration estimate does not select the phase or assess readiness.

Review the Result Before Following It

1. Check the goal and the planning dates against what you intend to do.
2. Compare the phase calendar with the actual weekly sessions, including recovery and other commitments.
3. If the available window changes, review the resulting plan rather than assuming fewer weeks simply require faster progression.
4. Keep recovery advice and changed circumstances in the decision even when the calendar says BUILD or PEAK.

Method note: this example was reproduced against TrainingPlan's phase-scheduling implementation at code revision 3152c04. It is a deterministic function demonstration, not an authenticated account walkthrough, a customer outcome or a validation study of the training rules. The phase and duration code was inspected separately from the weekly adjustment logic.