×

Strategy vs. The Random God

Write a JavaScript strategy. Backtest it on 800 candles of seeded history. Tune the parameters until the equity curve looks heavenly. Then click ✷ Reveal Future and watch random god rip your edge apart.

API

Define two functions:

function setup(api, memory) { /* called once before bar 0 */ }
function loop(api, memory)  { /* called at every candle close */ }

api

• api.i — current candle index
• api.bars[k] — {o,h,l,c,vol} for k ≤ i
• api.price — close of candle i
• api.position — current qty (+long, −short)
• api.indicators.macd — {macd[], signal[], hist[]}
• api.indicators.rsi[] — RSI(14)
• api.indicators.kdj — {K[], D[], J[]}
• api.indicators.ema(period) — EMA on closes
• api.openBuy(qty) · api.openSell(qty) · api.close()
• api.log(msg)

memory

An empty object you can attach state to. Persists across loop() calls within a single backtest/forward run.

Mechanics

• Trades fill at the candle's close. No slippage. No costs.
• One candle = one trading day (for AR / Sharpe annualization with 252).
• The historical 800 candles are the same every reload (seeded). The forward future is fresh randomness each time.
• Pan: drag chart horizontally. Zoom: scroll. Double-click: snap to latest.

The Lesson

In-sample tuning produces beautiful curves. Out-of-sample, the random god gives back exactly your costs. If the simulator had costs.