Reproducibility

Every backtest in Quanthop is fully reproducible.

The Guarantee

Given:

Same strategy code
Same parameters
Same data range
Same configuration

The result will be identical, including:

Every trade entry/exit
All metrics (PnL, Sharpe, drawdown, etc.)
The exact sequence of fills

How It Works

No Hidden State

The engine maintains no state between runs. Each backtest starts fresh.

Deterministic Execution

Order evaluation follows fixed rules
Fill prices are calculated, not simulated
No random number generation

Pre-computed Indicators

Indicators declared in init() are computed once before onBar() runs:

function init(ctx) {
  ctx.indicator('ema', 'EMA', { period: 20 });
}

// By the time onBar runs, ctx.ind.ema contains all values
// for all bars - computed deterministically

What Breaks Reproducibility

These patterns will cause different results between runs:

Random numbers

// FORBIDDEN
function onBar(ctx, i) {
  if (Math.random() > 0.5) {
    ctx.order.market('ASSET', 1);
  }
}

External data

// FORBIDDEN
function onBar(ctx, i) {
  const signal = await fetchExternalAPI();
}

Wall clock time

// FORBIDDEN - uses current time
function onBar(ctx, i) {
  if (new Date().getHours() > 12) {
    ctx.order.market('ASSET', 1);
  }
}

Verification

You can verify reproducibility:

Run a backtest
Note the exact metrics
Run again with identical settings
Results must match exactly

If they don't, there's a bug in your strategy (or the platform — please report it).

Why This Matters

Reproducibility enables:

Valid comparisons — Different parameter sets compared fairly
Debugging — Isolate issues by reproducing them
Auditing — Prove results to yourself and others
Trust — Know that "good" results aren't luck

Overview — Platform philosophy