Menu

Original research

What 375,000 days of training data actually show

We tested the claims training watches make against what athletes actually did. These are the results, including the weak signals and inconvenient null findings.
Aggregated before the personal data was deleted. Read the methodology and limitations before citing a result.
athlete-days375,000+
activities81,000
published studies5

Five questions. Five measured answers.

Each study includes its cohort, method and limitations.

  1. S1

    daily predictions

    Garmin race predictor: what 22,422 daily predictions show

    All our athletes got a steeper slowdown curve than the textbook formula predicts.

    our athletes · 22,422 prediction-days

  2. S2

    median correlation

    Does Training Readiness predict how a run goes?

    Median within-athlete correlation with running efficiency: 0.056. Effectively nothing.

    our athletes · 5,136 runs

  3. S3

    points per change

    How much does Garmin move your VO2 max?

    It changes about every sixth day — by 0.2 points. A year of training moves it 2.0.

    our athletes · 180+ days each

  4. S4

    of athletes agree

    Body Battery, HRV and sleep score: three metrics or one?

    They agree for 98% of athletes. Largely one signal wearing three interfaces.

    our athletes · 30+ paired days each

  5. S5

    median correlation

    Does last night’s sleep score predict today’s run?

    Median correlation 0.04 across 7,492 runs. The control worked; the signal was not there.

    our athletes · 7,492 runs

Run the numbers on yourself

Using these numbers

Cite them freely with a link back. Read the caveats first; every result has real limits.

Questions, corrections, or a dataset worth testing: hello@gneta.app