How wallet-copy works

The problem with every leaderboard

If you rank 30,000 traders by return, the trader at the top is usually the luckiest, not the most skilled — this is the multiple-testing problem, and almost no copy-trading product corrects for it. Scienter’s core trick is scoring every candidate with the Deflated Sharpe Ratio, using n_trials = the entire scanned population, not just the shortlist that made a “top 100” cut somewhere else. The first full live run of this pipeline (23,000+ candidates scanned, 150 deep-judged) found zero candidates that proved skill beyond population luck. The resulting watchlist is explicitly the WATCH tier: the cleanest available candidates, labeled best-available-not-proven, with the real deflated score shown on every card.

The pipeline

SCAN (every 6h)   40,000+ Hyperliquid traders + HL vaults
                  + OKX lead traders + Polymarket
   -> hard gates: copyable size, green both windows, real volume,
      no one-hit-wonders, no wash-churn
JUDGE             Deflated Sharpe vs. the whole population -> GOLD/SILVER/BRONZE/WATCH
   ->
WATCH (every 5min) the watchlist traders' live positions
   -> NEW / SIZED-UP / FLIPPED
SUGGEST           a card: confidence 0-100 (deflated skill, consistency,
                  crowd agreement, leverage sanity, freshness, copy-lag,
                  funding regime)
   ->
YOU DECIDE        dashboard buttons or Telegram: Took it / Pass / Mute
   ->
RECEIPTS          every suggestion tracked, even ones you passed on;
                  the bot grades itself publicly, fees included

The late-entry guardrail

A suggestion card is only as good as how fresh it is when you act on it. Every tap re-validates at tap time, refetching the trader’s live position:

Check	Result
They already closed the position	Hard block — nothing left to copy
They flipped side	Hard block — the card is inverted
They cut more than 50% of the position	Hard block — de-risking, not entering
Re-validation itself fails	Hard block — fail-closed: unverifiable means refused
Price already ran >0.75% their way	Soft warning — you’d be buying their profit, not their idea
Card older than 90 minutes	Soft warning — this may be a different trade than the one flagged

Hard blocks can’t be overridden. Soft blocks require a second, deliberate tap.

What it does not do by default

Scienter never trades, never touches your keys, and needs no wallet, by default. Buttons only log your decision — every trade is yours to place or not. If you choose to enable approve-gated execution, a tap still triggers every single trade; there is no full-auto mode in this codebase.

Honest confidence scores

A signal component with no data is marked N/A and dropped from the score entirely — it does not get averaged in as a neutral or perfect value. This was a real, fixed bug: Polymarket cards used to score a perfect “entry timing” component simply because the lookup found no price data at all. The score now renormalizes over only what’s actually measurable, and the card says which parts it couldn’t measure.

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.

Get Started

Concepts

Guides

How wallet-copy works

The problem with every leaderboard

The pipeline

The late-entry guardrail

What it does not do by default

Honest confidence scores

​The problem with every leaderboard

​The pipeline

​The late-entry guardrail

​What it does not do by default

​Honest confidence scores

The problem with every leaderboard

The pipeline

The late-entry guardrail

What it does not do by default

Honest confidence scores