Backtest a Hyperliquid Strategy by Describing It

A strategy you haven't tested is an opinion. Until now, testing it meant writing a backtesting engine first.

Strategy backtest skips the engine. Describe the strategy in plain English, or paste a Pine Script, and see how it would have traded on Hyperliquid, trade by trade.

Go From an Idea to a Tested Strategy in 3 Steps

Build. You describe what you want in plain English and the model asks for the asset and timeframe if you left them out. It returns a strategy definition: a readable summary of the indicators, conditions, risk rules and tunable parameters, plus the executable code behind them. Pasting a Pine Script from TradingView gets you to the same place.

Run. The server pulls Hyperliquid candles, runs the strategy candle by candle in an isolated context, and streams the engine log back as it goes, so you watch the run rather than wait for it. Costs are modeled on standard Hyperliquid trading costs: a flat 0.085% per side, made up of a 0.035% taker fee plus 0.05% slippage. Entries and signal exits fill at the candle close.

Study. The result arrives as a composite grade and a one-line verdict, a metrics strip, price, equity and drawdown charts, the full trade log, and a comparison against your previous run. The grade is there so a first read takes a second, and the metrics are there so the second read can disagree with it.

Skip the 3 Things That Stall Every Backtest

Testing an idea yourself runs into the same 3 problems every time.

Writing the engine. The machinery around a strategy has to be correct before the strategy itself can be judged.

Reading someone else's Pine. TradingView scripts only run on TradingView, so porting one by hand means reimplementing the indicator maths and hoping you matched it.

Tuning. Moving a parameter only tells you something if you can still see the run before it.

Why Testing a Strategy Used to Take Weeks

Write it yourself. Pull candles, build the indicator, write the loop, model fills and fees, produce the metrics, and then debug whether a bad result is a bad strategy or a bad engine. Most of that work is identical from one strategy to the next.

Use a hosted backtester. These work, and the strategy stays inside them. What they do with fees and fills is usually not visible.

Paper trade and wait. Honest, but slow, and it answers the question weeks after you asked it.

Tune Parameters Without Losing the Last Run

The workbench keeps parameter sliders next to the result. Move one and rerun, and the new run is shown against the previous one rather than replacing it.

Presets save a set of parameters you want to come back to. Saved runs keep the history of what you have already tried, so a session is cumulative.

There is also an Ask AI panel over a finished result, for asking why a drawdown happened where it did, and a chat for refining the strategy itself rather than its parameters.

Trade on Evidence, Not Memory

• Test an idea the same day you have it.

• Reuse a strategy you already trust, without rewriting it to find out.

• Settle on a parameter set because the numbers moved, not because you remember liking it.

• Read a strategy's trade log directly, rather than inferring behaviour from a summary metric.

Run Your First Backtest

• Strategy backtest

• Hyperliquid Info API documentation

• Hyperliquid data with zero rate limits

Part of our Hyperliquid series. Start with Lessons From Building Infrastructure for Hyperliquid.