Backtesting

Test trading strategies against historical data — tune thresholds, hold times and execution, or post-process trades with your own JavaScript.

Parameters

Lookback

Pair

Strategy

Timeframe

Signal threshold ×

Max hold (hours)

Exposure (0=auto)

Fee Rate

Slippage (bps)

Initial Equity ($)

Long only

Extra setting: post-process trades with custom JavaScript (runs in your browser)

// transformTrades(trades, params) is called with the simulated trades
// before metrics are computed. Return the trades you want to keep;
// you can also adjust them. Runs locally in your browser only.
//
// Each trade: { pair, side, entry_time, exit_time, entry_price,
//               exit_price, pnl_pct }  (pnl_pct is a fraction: 0.01 = 1%)
function transformTrades(trades, params) {
    return trades
        .filter(function (t) {
            // Example: skip trades held longer than 48 hours
            var heldHours = (new Date(t.exit_time) - new Date(t.entry_time)) / 36e5;
            if (heldHours > 48) return false;

// Example: only trade when the entry isn't a Monday (UTC)
            if (new Date(t.entry_time).getUTCDay() === 1) return false;

return true;
        })
        .map(function (t) {
            // Example: assume an extra 2bps of slippage per side (4bps round trip)
            return Object.assign({}, t, { pnl_pct: t.pnl_pct - 0.0004 });
        });
}

Define transformTrades(trades, params). params carries your form settings (days, pair, thresholdScale, maxHoldHours…). Summary, win rate and equity curve are recomputed from the returned trades.

Running backtest...

Example result — this is a real 30-day backtest over the top pairs with default settings, run automatically so you can see what to expect. Tweak the parameters above and hit Run Backtest.

-- Total Return

-- Win Rate

-- Max Drawdown

-- Trade Count

Replay

Prices rebased to 100 at the window start. Solid = in the walk-forward basket, dotted = tried but not selected. ▲ entry, ✕ exit. Space plays, ←/→ steps (shift = 24 bars).

-- show tried pairs

Equity Curve

Trade History

Pair	Side	Entry Time	Exit Time	Entry Price	Exit Price	PnL %

Research agent

An autoresearch loop: the model proposes one change at a time, runs the backtest tool, keeps a ledger and picks a winner behind a 25% drawdown gate. Usage is billed per token at OpenPaths price +20%; the default model is free.

Ask for an improvement and watch the runs land in the replay above. Each run is one backtest; the ledger on the right keeps every result.

Disclaimer: Backtesting results are hypothetical and do not guarantee future performance. Past results are not indicative of future returns. Trading cryptocurrency involves substantial risk of loss. Always do your own research before making investment decisions.