Weighted Power

Weighted Power gives variable power a fourth-power-weighted summary over complete 30-second windows.

Calculation

Build the one-second timer-running power sequence. For every complete trailing 30-second window, calculate its arithmetic mean M. Exclude windows containing any missing power sample. Then:

Weighted Power = (mean(M⁴))^(1/4)

The result is watts. The first 29 seconds cannot form a complete window. Paused time is removed before the trailing windows are formed, so a window can span the join around a recorded pause in compressed timer time. This is an implementation choice, not a promise that every platform treats pauses identically.

Coverage and sampling

Power is held forward for at most ten seconds during projection; longer gaps remain missing. Recorded zero watts are valid coasting and remain in means. At least 80% power coverage is required by activity analysis. A recording with no complete window is unavailable. Less than ten minutes of timer duration is marked low confidence. The positive recorded timer/moving duration is the coverage denominator, or the projected count if that summary is missing.

Verified examples

For 3,600 seconds constantly at 200 W, all 3,571 complete windows average 200 W: Weighted Power is 200 W. For 30 seconds at 200 W followed by 30 seconds at 0 W, the 31 complete rolling means produce approximately 135.3952 W; the full-recording average is 100 W. Replacing the zeros with missing values would exclude windows instead of treating the second half as coasting.

Weighted Power does not need FTP by itself. Power Intensity divides it by effective FTP, PSS adds duration, and Power Variability compares it with average power. A weighted score is a modeled summary, not measured metabolic power.

Sources and implementation scope

Coggan, power-based training metrics provides background, not a specification of every EffortSignal rule. The formulas, windows, quality gates and examples above describe EffortSignal 0.1.13, checked September 10, 2026. Product-specific thresholds and rounding should not be attributed to the background source.

All explanations · All features