Backtestify
BACKTEST EXISTING2024-12-14 → 2026-08-31 · 1.7 yearTested windowBTCUSDT (asked for "BTC") · 5m · history 2024-12-14 → 2026-08-31 (625 days), identical for all timeframes · $10,000 account

Backtest the ICT Silver Bullet strategy on BTCUSDT

Rules: ICT Silver Bullet (New York windows 03–04 · 10–11 · 14–15) · as published luxalgo.com·ictkillzone.com

Win rate

NOT RECOMMENDED

27.8%

126 trades · 5M · 625 days · Binance

Data through 2026-08-31 00:00 UTC

NOT RECOMMENDEDA fixed set of measured gates decides this. It is not an opinion and not advice.

Not recommended on this evidence

Tested an approximation of your rules — see 'What we actually tested'Part of your description could not be executed exactly, so the verdict is capped.Win rate 27.8% is below 51%Share of closed trades that ended in profit, against the minimum this app requires.Lost money after costs: profit factor 0.32Checks that gross profit covers gross loss by the required margin.

EvidenceWeak evidence

✓ Sample ≥ 100 trades126 trades in 625 days✓ Unseen sample ≥ 30 trades43 trades✗ Profit factor ≥ 1.30.32✓ Drawdown ≤ 25%18.5%✗ Edge on unseen data > 0-0.755R✗ Edge with doubled costs > 0-1.369R✗ Rules tested exactlyapproximated

ReturnTotal simulated return

-18.5%

Account change over the tested window

Profit factorProfit per loss

0.32

Gross profit ÷ gross loss (above 1 = profitable)

Max drawdownDeepest closed-trade fall

-18.5%

Worst fall from a previous account peak

ReliabilityConfidence in test

64/100

How much this test can be trusted

Unseen-data edgeUnseen-data edge

n/a

Average result per trade on data never optimised on

Asked for "BTC" · tested BTCUSDT

What we actually tested

  • · 5M candles inside a New York Silver Bullet window (03:00–04:00, 10:00–11:00 primary, 14:00–15:00)
  • · Step 1: the 20-candle low is swept and the candle closes back above it (sweep armed)
  • · Step 2: a later candle in the same window completes a bullish three-candle fair value gap
  • · Limit entry at the 50% level of that gap, valid until the window closes, one trade per window
  • · Stop at the low of the sweep candle; target 2R

Buys whenThe exact entry conditions that were executed for long trades.

inside 03:00-04:00 / 10:00-11:00 / 14:00-15:00 America/New_York time, price first sweeps the 20-candle low and closes back above it, then a LATER candle in the same window completes a bullish fair value gap

StopPosition sizing and stop rules exactly as executed. Nothing was added.

the low of the sweep candle that armed the sequence

Target / exitTarget, trailing and time exits exactly as executed.

2R (R = entry-to-stop distance)

Risk per tradePosition sizing and stop rules exactly as executed. Nothing was added.

1% of equity risked to the stop

One real trade, most recent

One real trade from the tested ledger, drawn on the same candles the engine used.
FROM TESTED HISTORY

BTCUSDT · 5m · actual exchange candles executed by the engine

LONGLOSS · -1.19RStop hit (SL)
Bought at 79,130.20 on 2026-08-28 14:45, stop at 78,622.20, target at 80,146.21. Closed at 78,606.48 on 2026-08-28 15:50 — Stop hit (SL). Result: LOSS (-1.19R)
  • Up candle
  • Down candle
  • Entry price
  • Stop
  • Target
  • Exit price

Entry

79,130.20 · 2026-08-28 14:45

Initial stop

78,622.20

Target

80,146.21

Exit

78,606.48 · 2026-08-28 15:50

Each candle is one 5M bar: the thin line is the high-low range, the thick body is open-to-close. Vertical marks show when the trade opened and closed. One real trade, picked by recency — an illustration of the executed rules, not proof they will work again.

How $10,000 would have changed

The account balance after each closed trade, starting from the stated budget.

Account value after every trade closed, in dollars

  • Account value

Left axis is dollars, bottom axis is time. The line only moves when a trade closes, so flat stretches are periods the rules stayed out of the market.

Simulation on historical candles, not live trading. Full methodology in Pro depth.

This is the free view of a shared report. Run your own strategy on the same engine.

Backtest a strategy