LONA :: terminal

● Connect, read-only

Paste your public Hyperliquid address below. It is a public identifier — the same thing a block explorer shows — and it lets this page read your balances, positions and funding directly from Hyperliquid. Never paste a private key or seed phrase into any web page, including this one. Nothing here can trade or withdraw; order placement lives in the LONA execution service, which runs on your own machine and holds its key in an environment variable.

Hyperliquid address

The address is kept in this page’s URL, not in browser storage, so bookmarking the page remembers it and closing the tab forgets it.

▸hyperliquid perpetuals

market	mark	24h	24h volume	open interest	funding /8h	annualised	max lev	LONA

Funding is quoted per hour on Hyperliquid, not per eight hours like most venues — the 8h column here is the hourly rate multiplied by eight so it can be compared with everywhere else. Annualised is the current rate extrapolated forward, which is why a market in a short squeeze can read −3,000%: it is a snapshot of one hour, not a forecast of a year. The carry book on the signals page ranks on the trailing 7-day mean for exactly that reason.

⚠ Computing from live candles

Both books below are recomputed in this page from Hyperliquid’s own daily candles every time you load it. Nothing is cached and nothing is pre-baked, so what you see is what the rules say right now.

▸finding 008 · breakout momentum long only

market	weight	close	50d high	days held	30d volume	carry cost

Rank the perps by trailing 30-day dollar volume, take the top 20. Hold one for five days after it closes at a new 50-day high, at 1/20 of capital each. Dividing by the size of the universe and not by the number of signals is deliberate: it means a quiet market leaves you in cash instead of putting the whole account into one coin.

▸finding 009 · hedged funding carry market neutral

market	weight	7d mean	annualised	mark	spot leg

Short the ten richest funding payers on the perp and hold an equal size on spot, so the price cancels and only the funding is kept. Unhedged this loses 62% a year — the coins paying the most funding are the ones going up hardest, so shorting them without the spot leg is shorting momentum.

▸near misseshow close is the rest of the universe

market	close	50-day high	gap to trigger	progress

A market at 100% is at a new high right now. This table is here so a quiet signals page reads as “nothing qualified” rather than “something is broken” — the rule is in cash about a third of all days and that is the design, not a fault.

▸the two candidates2023-08-11 → 2026-08-10, after fees and funding

strategy	CAGR	Sharpe	worst DD	Calmar	t
breakout + hedged carry	+49.5%	+2.03	−11%	4.52	—
breakout (008)	+22.9%	+1.10	−15%	1.51	+1.5
hedged carry (009)	+21.7%	+11.24	−1%	30.5	+2.4
buy and hold BTC	+30.3%	+0.80	−53%	0.57	+1.2
buy and hold the top-20 book	−15.9%	+0.16	−83%	−0.19	+0.2

The t column is the Sharpe divided by its own standard error, and that error depends on the number of years, not the number of rows. Three years gives roughly ±0.7. The breakout’s 1.5 is a hint, not a proof — which is the entire reason the forward record exists.

▸year by yearwhere the honesty lives

year	breakout	carry	both	BTC
2023 (part)	+38.6%	+22.1%	+171.7%	+52%
2024	+43.4%	+27.9%	+61.4%	+121%
2025	+13.3%	+24.8%	+5.3%	−6%
2026 (to Aug)	+0.4%	+11.6%	−13.7%	−26%

The breakout is a bull-market engine with a cash brake: it earns close to nothing in a bear and loses very little. Nov 2025 → Aug 2026 it returned −3% at a −8% drawdown while Bitcoin fell 48% with a −43% one. The carry is the flat income stream that does not care. Neither alone is a business.

▸what each rule survived

attack	result	verdict
parameter surface	31 of 32 cells positive — a region, not a spike	survived
fees to 30 bps a side	positive; venue charges 4.5	survived
actual hourly funding	costs 8.4 points of CAGR	survived
adverse selection, measured	held coins pay +25.7% carry vs +16.9% for the average	survived
delete the best 3 coins	Sharpe still +0.87	survived
remove delisted names	they contribute −0.3%	survived
fill one day late	Sharpe +1.13 · two days, +0.42	survived
broken-shift control	+427% a year, Sharpe 9.2 — the look-ahead guard is load-bearing	survived
2020–2023, never fitted	Sharpe +1.55 to +1.73	survived
carry: 7-day-stale data	+17.1% — funding is persistent	survived
carry: drop the best month	+15.1% from what is left	survived
carry: hedge leaks 25%	+16.9%, Sharpe +1.57	survived
short side: new 50-day LOW	−9.5% a year	failed
funding filter on entries	non-monotonic — fitted noise	failed
carry at 50 bps a leg	−17.5% a year	failed

The failures are listed on purpose. Long-only is correct here, and unlike finding 006 we know that because the short side was measured, not because it was never looked at.

⚠ What these numbers do not contain

Three years is one and a half cycles. No amount of re-reading the same history adds evidence.

A carry Sharpe of 11 is not a real Sharpe. Carry income arrives in a trickle and the loss arrives all at once, from something not in the sample: venue failure (FTX paid excellent funding right up until it did not exist), a delisting while short, a liquidation on a wick before the hedge can be topped up, a depeg of the stablecoin that is 100% of the collateral. Volatility cannot see that shape.

The carry needs a second venue for the spot leg — two sets of fees, two sets of collateral, and transfer risk between them. Until that is measured, the +21.7% is a ceiling.

This is research, not advice. Every number here is a backtest. Losses are possible and leverage makes them faster.

⚠ Read this before arming anything

You have chosen to let LONA place orders through an agent wallet. My honest read, once, and then I will help you do it properly: the breakout’s t-statistic is 1.5, which is below the bar this project set for itself, and the forward record is eight days old. The carry is stronger statistically but its real risks — venue, hedge, collateral — are exactly the ones three years of data cannot show.

If you go ahead anyway, the architecture below is what makes it survivable: an agent wallet that can trade but cannot withdraw and can be revoked in one click, a key that lives in an environment variable and never in the repo or this page, dry-run by default, and hard limits that are checked in code before every single order rather than remembered by a human.

▸position sizingwhat the rules would actually send

Account equity (USDC)

Notional multiplier — 1.0×

Breakout book, deployed today—

Per breakout position—

Carry book, perp leg—

Carry book, spot leg required—

Total notional—

Backtested worst drawdown at this size —

Estimated round-trip cost per rebalance —

The multiplier scales notional, not risk appetite. Sharpe does not improve with leverage — only the drawdown grows. At 3× the backtested worst loss goes from −11% to roughly −36%, on three years that contained no venue failure.

▸hard limitschecked in code, before every order

Max single position, % of equity

Max gross notional, % of equity

Daily loss kill switch, % of equity

Total drawdown kill switch, %

Arm for

There is no permanent arm option and that is on purpose. Every arming expires. A system that is on until someone remembers to turn it off is a system that is on.

▸execution servicenot connected

This page never holds a key and never signs anything. Orders are signed by Lona.Execution, a small service that runs on your own machine, reads the agent key from the LONA_HL_AGENT_KEY environment variable, enforces every limit above in code, and exposes a local endpoint this page can talk to on 127.0.0.1:7717. It is not built yet — it is the next thing I write, and it will ship with its own test suite proving each limit blocks what it claims to block.

Endpointhttp://127.0.0.1:7717

Key sourceenv: LONA_HL_AGENT_KEY

Wallet typeagent — trade yes, withdraw no

Default modedry run

Statusoffline

To create the agent wallet: Hyperliquid → More → API, generate one, and revoke it from the same screen the moment anything looks wrong. An agent wallet cannot move funds off the exchange even if the key leaks — that is the entire reason to use one.

Looking for a published briefing

…

▸the briefing

▸every measurement behind it

measurement	value	source

● Why this is a file and not a live connection

The analyst runs on the PC at home. This page does not ask it anything — the PC writes a small JSON file and pushes it here, and never accepts a connection. That machine will one day hold a signing key, and a listener on it would be reachable from the whole internet so that a phone could see a number. The blast radius of a bug on this page is a wrong number. The blast radius of a bug in a listener is the wallet.

It also means everything here is a snapshot, so its age is the first thing shown rather than a footnote. Only allowlisted market measurements are ever written to the public file; anything about the account or a planned order stays on the PC, in a second file this site never sees.

Looking for a published leaderboard

…

▸funding leaderboard

market	now, annualised	30-day mean	hours positive	24h volume	open interest	spot hedge

Checking…

Looking for a published plan

…

▸forward record vs the expected band

peak

▸open paper positions

market	left	entry	now	move	P&L

▸orders the rule wants

market	size	why

▸reconciliation

This compares the book the rule wanted against what the wallet actually holds, read from its public address. Without it the paper log measures one strategy while the account runs another, and nothing would say so.

● Nothing here has been traded

Every figure on this page is the book finding 008 wants, at an assumed notional, marked at live prices. No order has been placed and no number here is an account balance. The execution service has no command that can send an order, and a unit test fails the build if anyone adds one.

Positions are held for five daily bars from entry, so left is how many bars remain before the rule closes the trade.

▸open positions

▸margin

▸recent fillslast 50

▸funding paid and received