This walks through reading one real signal end-to-end — not just fetching JSON, but understanding what the number in front of you actually means before you act on it.

​
1. Pull the live snapshot

Look at signals. If it’s empty, that’s not broken — the ensemble only publishes above its conviction bar (min_conviction in the same payload), and below_bar tells you how many candidates existed but didn’t clear it. A quiet market produces an empty list on purpose.

​
2. Read one signal properly

Before you treat 71 as “71% confident,” know what it’s actually built from:
  • contributing_signals tells you which independent groups agree. funding and liq both read Hyperliquid’s own positioning data — so this is corroboration within one group, weighted lower (0.35) than a second, genuinely independent group would be. If only one group is measurable, conviction is capped at 78 regardless of how strong that one read is — see Signal tiers.
  • evidence_ts vs ts — the age of the underlying evidence, not just when it was scored. Funding evidence expires within an hour; liquidation maps within 90 minutes. A signal scored five minutes ago on evidence that’s 55 minutes old is close to expiring, not fresh.
  • coverage_capped — for liquidation-cluster data specifically, this tells you whether the sample of accounts scanned was large enough to trust the number, or whether it’s a floor on a partial view.

​
3. Check whether it’s actually worth following

Pull the backtest for this exact signal kind before treating conviction alone as a track record:
Read biased in the response — if you set min_conviction below the bar this ensemble actually published at, the result understates real selectivity (see Backtest & Tier Routes).

​
4. If it’s a wallet-copy suggestion instead

Signals from the edge engine and copy-trading suggestions are two different things sharing one dashboard. A copy suggestion looks like a card with a wallet address, not a pair/direction signal — see How wallet-copy works for the confidence formula and the late-entry guardrail that re-validates a card the instant you tap it.

​
5. Decide, and let it grade itself

Every decision is logged, including passes — the point is a track record you can check later, not just the wins.
action is one of took, passed, mute. This only logs — nothing executes unless you’ve separately armed approve-gated execution, and even then a fresh tap is required per trade.
Scienter publishes impersonal market signals of general and regular circulation. Nothing here is investment, legal, or tax advice, or a recommendation to buy, sell, or hold any asset. Scienter is not a registered investment adviser or broker-dealer. Trading digital assets can result in the total loss of your funds — see Disclaimers for the full text.