Skip to main content
Back to Glossary

Statistical Terms

Statistical terms in training are the measures used to smooth, compare and classify data series: rolling mean, standard deviation, z-score and baseline.

Definition

“Statistical terms” here means the set of measures obseed uses to turn individual readings into reliable statements. A single heart rate or HRV value says little; it only becomes information when compared with the usual range.

The six most important terms at a glance:

TermMeaningWhere it appears in obseed
Rolling meanAverage of the last n values, all weighted equally30-second smoothing for Normalized Power, 7-day HRV
Exponentially weighted moving average (EWMA)Average in which recent values count moreChronic Training Load (42 days), Acute Training Load (7 days)
Standard deviation (SD)Measure of spread around the meanNormal range of resting heart rate and HRV
z-scoreDistance of a value from the mean in standard deviationsClassifying a daily value as normal or unusual
Coefficient of variation (CV)Standard deviation divided by mean, as a percentageComparing spread across people and metrics
BaselineIndividual reference range, usually mean ± 1 SD over 28–60 daysBasis of every deviation alert

Why it matters

Physiological values fluctuate from day to day, even without a cause. Interpreting every swing means reacting to noise. Never reacting means missing the cases where the body is actually sending a signal.

Statistics draws the line between the two. It answers whether an HRV of 48 ms in someone with a mean of 62 ms and a standard deviation of 8 ms is still normal, or far enough outside to justify changing the training plan.

How obseed measures it

obseed computes the baseline for resting heart rate and HRV as a rolling mean and standard deviation over recent weeks and compares each new morning value against it:

z = (value − mean) / standard deviation
CV = standard deviation / mean × 100

For training load obseed uses the exponentially weighted moving average of daily TSS. The smoothing factor α = 1 − e^(−1/τ) with time constant τ in days sets how quickly old load loses weight: τ = 42 for CTL, τ = 7 for ATL.

z-scoreClassification in obseed
between −1 and +1Normal range, no action
between −1.5 and −1Watch, lean towards reducing intensity
below −1.5Unusual, adjust training
above +1.5 for resting heart rateUnusual, sign of infection or overreaching

Example

An athlete has the following HRV baseline over 28 days: mean 62 ms, standard deviation 8 ms. The coefficient of variation is 8 / 62 × 100 = 12.9 %.

On the morning after a hard block she measures 48 ms. The z-score is (48 − 62) / 8 = −1.75. The value lies below the −1.5 limit and is flagged as unusual. The planned intervals are swapped for an easy base run.

Two days later HRV is 58 ms, z = −0.5, back in the normal range. Her CTL that week: with 95 points the day before and 120 TSS today, α = 1 − e^(−1/42) ≈ 0.0235 gives a new CTL of 95 + 0.0235 × (120 − 95) ≈ 95.6. A single hard day barely moves chronic load; that is exactly the purpose of exponential smoothing.

Categories

  • Datenanalyse
  • Technische Begriffe