Methodology
A number you can
argue with.
The stress score is five measured quantities, each scaled against a stated threshold and then weighted. There is no fitted parameter, no training set, and no vendor input. Below is the whole thing.
Inputs
| Component | What it measures | Saturates at | Weight |
|---|---|---|---|
| Drawdown depth | Distance from the highest close in the window to the latest close. A token at its high scores 0; one 30% below scores 1. | −30% | — |
| Realised volatility | Standard deviation of daily log-free returns, annualised by √252.
Measures violence of movement, direction-blind. |
120% | — |
| Time underwater | Share of sessions closed below the running maximum. One high on day one and a grind down scores near 1; a steady climb scores near 0. | 100% | — |
| 30-day trend | Return over the last 30 sessions, mapped so that
+20% → 0 and −20% → 1. |
±20% | — |
| Relative to the pack | Window return minus the median window return across all 34 feeds. The market-wide part of a decline is not this token's stress. | −20pp | — |
Bands
| Band | Range | Reading |
|---|---|---|
| settled | 0.00 – 1.99 | Near its own high, moving quietly. |
| calm | 2.00 – 3.99 | Mild drift, within the range this feed normally travels. |
| watch | 4.00 – 5.99 | A real drawdown, or unusual movement, or both. Worth inspecting the composition. |
| stressed | 6.00 – 7.99 | Deep below the high, or getting there violently. |
| capitulation | 8.00 – 10.00 | Sustained drawdown, elevated volatility and persistent underperformance at once. |
Read this before quoting a number
What the score is not
- It is not a forecast. Every input is backward-looking.
- It is not comparable across assets as a quality judgement. A stablecoin-adjacent ETF and a quantum-computing name have different normal volatility; the thresholds are the same for both, which is a choice, not a discovery.
- It does not know about liquidity, float or holders. Nothing on this site measures who holds what.
- It does not know about news. A −30% move on an earnings miss and a −30% move on a delisting scare read identically here.
The window
Why every feed shares one window
Feeds do not print at the same cadence. Measured on this chain, SPY writes roughly 1.1 rounds a day and IONQ writes roughly 52 — a 47× spread. Cross-asset statistics are only meaningful on sessions that every feed in the comparison actually printed, so Meridian intersects rather than fills.
That means one late-starting feed shortens the window for everyone, and the honest response is to say so rather than to interpolate. Sessions where no feed moved — weekends and US market holidays — are dropped rather than carried forward.