Tutorial

This is the whole product in one page. If you read nothing else, read this — it is written to be sufficient on its own.

It takes about fifteen minutes end to end. Every step states what you do, what you should see, and what it means. Where a step hides a mechanism, a collapsible panel underneath carries the detail.

What you are actually doing

The product does one thing repeatedly, at three levels of ambition:

Everything below is those three verbs, in that order.

Step 1 — Start the workspace

npm run dev

Open http://localhost:8080 .

You should see: a dark trading workspace — a market chart on the left, a Pine editor on the right, a slim control bar across the top.

Step 2 — Choose a market

Use the controls in the top bar. The symbol picker’s search button queries Binance and ranks pairs by 24-hour volume, so liquid markets appear first. Timeframes are grouped, with a custom field for anything unusual.

You should see: the chart redraw with candles.

Step 3 — Load a strategy

Pick one from the strategy selector. Each file in app/strategies/ appears there.

For this tutorial use ema_cross.pine.

You should see: Pine source in the editor, and the parameters it declares listed in the parameter area.

Step 4 — Read the parameters before you run anything

The editor shows a parameter for every input.* call in the script. ema_cross.pine has four:

These four are the search space. Nothing else about the script changes between trials. This is the single most important idea in the product: you are not searching over strategies, you are searching over this list of numbers.

// 1. Positional default — the value comes first, and that is what marks it
//    as tunable. The optimiser reads the first argument.
fast_len = input.int(9, "Fast EMA Length")
atr_len  = input.int(14, "ATR Length")   // two spaces — still fine here

// 2. Emit the signal as data, not only as a picture.
plotshape(long_signal, "Buy", shape.triangleup, location.belowbar, color.green)
plot(long_signal ? 1 : 0, "Long Signal", display = display.none)   // ← this line

Step 5 — Run one backtest

Press Ctrl + Enter, or use the Run button.

You should see: a lower panel with three tabs — Performance, Trades Analysis, Trades Log — plus a metric card row and an equity curve.

Step 6 — Read the result

Look at three numbers:

1. Net profit — did it make anything at all, over the period?
2. Max drawdown — how much of the peak was given back at worst? A strategy that is up 20% and gave back 60% is a different animal from one that is up 5% and gave back 1%.
3. Trade count — how many times the logic actually fired. A great-looking result on four trades is noise.

If trade count is very low, the result is not informative yet. That is normal, and it is exactly what the optimizer exists to fix.

Step 7 — Open the optimizer

Switch to the Optimizer workspace. The chart and editor stay mounted; the lower panel is replaced.

The configuration column shows the four dimensions from Step 4, already discovered from your source. You choose:

You should see: a grid size — how many distinct parameter sets the space contains — and an estimate of how many trials remain.

Step 8 — Launch the run

Click ▶ Run again. Configuration locks, and the run captures the source, the bars and the search space it was given — so the result is attributable to exactly this configuration.

You should see: a results table filling in progressively. Trials complete in parallel across your CPU cores.

Step 9 — Read the results table

Sort by profit factor or your chosen metric. The table lists every completed trial with its parameters and metrics.

Three questions, in order:

1. Where is the best trial? Click it. Note its parameters.
2. Is it an outlier or a trend? A single spike in an otherwise flat distribution is usually luck. A broad cluster of good results is usually real.
3. Do the parameters make sense? If your “best” result requires fast_len: 3, slow_len: 58, the optimiser has probably found noise, not an edge. You are looking for regions that are stable, not a single maximum.

Step 10 — Open the analysis

Click Open analysis. This replaces the popup with a full Analysis workspace for the current run.

Three things worth doing straight away:

Then Open in Snipe to inspect one specific trial in detail — this opens the exact trial you had selected, not just the run it came from.

Step 11 — Export

The results panel offers three actions, all inside a Result actions menu (⋮):

CSV respects the filters and sorting you have applied — use it for “the top ten under these criteria”. JSON always contains the complete run regardless of what is on screen — use it when you want to re-analyse offline. The share card is an image rather than data; it never includes your Pine source or the winning parameter values.

Full detail in Sharing & export.

Where to go from here