Hyperliquid needs no API key, so this works on a fresh checkout with nothing configured.

​
1. Pick an address

Any Hyperliquid trader address works. The leaderboard gives you candidates:
The top name on a leaderboard is the worst starting point for judging skill — it is the most extreme outcome of a large multiple-comparisons experiment. That is the entire reason the next step exists. Pick a few names from different ranks.

​
2. Get the scorecard

Read these three fields together and never separately:
If every DSR on the board is exactly 0, that is almost certainly a var_sr misconfiguration silently zeroing the statistic, not a population without skill. See Wallet scorecards.

​
3. Ask where the alpha lives

A good DSR says the record is unlikely to be luck. It does not say what the trader is good at.
Compare the two modes:
  • Sharp realization concentration, flat exposure — credited for exits, not for holding.
  • Both concentrated in the same window — a genuine session-specific edge.
  • Neither concentrated — a generalist, or too few trades to tell.
Check the p-value before believing any label. A 40-trade account concentrates by accident, and the module attaches a permutation p-value precisely so you do not have to take the magnitude at face value.

​
4. Check identity

Two things to expect:
  1. Zero clusters is the normal result. It means no evidence of a link, not proof of independence.
  2. On Hyperliquid, deployer and gas_fingerprint skip every pair — the venue exposes no contract-creation or transaction-metadata class. Six heuristics ran, not eight.
On a master checkout this returns module_absent. Treat that as “not checked”, not “no links found.”

​
5. Write the judgment

A defensible summary states the correction, the concentration with its p-value, and the gaps:
Address 0x… has a raw Sharpe of X and a deflated Sharpe of Y over N trades, so the record survives correction for a scanned population of M. Realized P&L concentrates in the Asia session (permutation p = 0.02); exposure attribution does not concentrate, so the edge appears to sit in exit timing rather than in holding. Entity clustering was not run — the module is absent on this deployment.
The last clause is what makes it defensible. A judgment that omits what it did not check cannot be audited.

​
What this does not tell you

  • Nothing about the future. Every figure describes a past record.
  • Nothing about capacity. A strategy that worked at 100kmaynotat100k may not at 10m.
  • Nothing about survivorship beyond the scan. Traders who blew up and left are not in the population.
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.