Equity (net, USD)
Is it realistic? Live vs research expectation
Coins
Recent fills
Events
Settings green = as researched · amber = scales the result · red = a different strategy · blue = risk / reporting only
Token scope which coins trade, and the lead filter per coin
Coins outside the tested list are marked; the research covers few of them.
1. Fund the account (in the HyperLiquid app)
- Deposit USDC to HyperLiquid (Arbitrum bridge, in the HL app). This page never moves funds.
- Main-dex coins (BTC, ETH, …) use the main perp balance. xyz coins use the xyz dex balance: transfer collateral to it in the HL app, or the xyz coins will fail the margin check below.
- Create an API wallet: HL app → More → API → generate and authorize. An API wallet can trade but cannot withdraw. Copy its private key (shown once by HL). Never use your main wallet's key here.
2. Credentials
Keys never cross the web. On the VPS, off-stream:
python -m trader.wallets set <name> (the key goes into a hidden
prompt). It writes this strategy's own wallet file (gitignored, 0600); another strategy uses another file. Only
an API wallet's key is accepted: the main wallet's key is refused. Risky actions here need the authenticator
code (python -m trader.auth setup once): 3. Sizes and limits for live edit in Settings; the first-live preset uses small sizes
4. Pre-flight checks
| not run yet |
5. Switch
Only a live service (its own config, live_allowed) can switch; the paper twin never sends
an order. Switching cancels every quote first and starts a new run. Positions stay on both exchanges when you
go back to standby (each still hedged by the other leg): flatten first if you want to be flat. A restart of the
service always comes back in standby.
6. Live vs paper twin
| no comparison yet |
Diagnosis: where the PnL came from
Full report with hours and risk:
python -m trader.report. Raw market data
for replaying these hours: Paper audit
Every simulated order, fill and cancel checked against the live market it saw: price on
the right side of the book, the real trade (by HL trade id) at or through our price, after the order went live
and before its cancel landed, size no larger than what the print had left after the queue ahead. Deeper check
against HL's own tape:
python -m trader.audit --clickhouse.