Stress score reads exactly 0 for people with high HRV #631

Which metric?

Stress

What it showed, and why you think it's wrong

Stress shows 0 on all 8 nights I have data for, and Stress Nerd Stats lists 0.00 for each. I'd expect a score that varies with the data. A few other users have reported the same on Reddit.

Cause. onehz_pipeline.dart maps the Baevsky Stress Index to 0-100 as 100*(ln SI - ln 20)/(ln 600 - ln 20) and clamps. Any SI <= 20 gives exactly 0. My nightly SI is about 10-14 on every night (level "low"), so every night clamps. These look like plausible values for a high-HRV sleeper, not noise. The scale seems to have been calibrated on SI ~47-52 (derivation_engine changelog v8). Reproduced on an imported backup, algo v113.

So 0 means "below the scale", not "no data". The index and level are right; the score carries no information once you are under the floor.

Proposal.

1. Show the index/level/score in Stress Nerd Stats (small PR, no schema change, no new series key). It reads the day's own stored stress block.
2. Make the 0-100 score relative to the person's own trailing nights (ln SI vs ~28 prior nights, tonight excluded, robust spread, abstain with "-" under ~7 nights; keep level as the absolute label). Seed the baseline from stored day_result stress.si so nobody sees a week of "-". Needs a kAlgoVersion bump.

Question for maintainers: is a baseline-relative score acceptable for "Stress", given that 70 would then mean "high for you" rather than high absolute tension?

Happy to implement the second part once you've said whether the direction is acceptable.

What kind of wrong?

Other

App version

0.11.3 (78)

Which band?

WHOOP 4.0

Screenshots

No response