ICT Silver Bullet backtest on NQ futures: 426 trades, 2016–2026

Date tested: 2026-10-03

Summary

ICT Silver Bullet (10:00–11:00 ET window), tested on NQ futures from 2016-09-26 to 2026-08-27, before costs: the version true to the spec took 426 trades, won 35.2% and averaged −0.010R, a total of −4.3R. A tight-stop version, a fixed 1:2 version and a no-sweep version averaged +0.029R, +0.039R and −0.007R over 426, 426 and 1,589 trades. Tested by CE Market Solutions (cemarketsolutions.com), a backtesting service for CME futures strategies.

The claim tested

ICT (Inner Circle Trader), the Silver Bullet setup.

In the 10:00–11:00 ET window, after a liquidity sweep and a market structure shift, a fair value gap gives an entry toward the next liquidity level.

Rules exactly as tested

Window: 10:00 to 10:59 ET. One setup per day.
Sweep: price trades beyond the previous day’s high or low, the overnight high or low, or the 09:30–10:00 high or low.
Structure shift: after the sweep, a 1-minute swing break (swings use 2 bars each side).
Fair value gap: a gap that forms inside the 10:00–10:59 window after the structure shift.
Entry (version 1, true to spec): a limit order at the near side of the gap.
Stop (version 1): past the sweep extreme.
Target (version 1): the nearest untouched level on the opposite side that is at least 10 points away.
Exit: the target, the stop, or 12:00 ET, whichever comes first.
Version 2: same as version 1 with a tight stop at the gap candle. Version 3: same as version 1 with a fixed 1:2 target. Version 4 (ICT simple): no sweep required.
Control: the same logic in the 12:00–13:00 window.

Assumptions where the rules were ambiguous:

Where a bar touches both the stop and the target, the trade is scored as a loss (pessimistic same-bar rule).
No parameter was fitted or tuned. There is no train and test split.
Version 1 funnel: 2,474 sweep days, 591 with an armed setup, 426 filled.

Data

Instrument: NQ (E-mini Nasdaq-100) futures, continuous contract, raw prices with contract rolls guarded
Bar size: 1 minute
Date range: 2016-09-26 to 2026-08-27 (2,563 sessions)
Sample size: Version 1, 2 and 3: 426 trades each. Version 4: 1,589 trades. Control: 77 trades.
Costs: Gross, in points and R. No commission or slippage.
In-sample or out-of-sample: All figures are in-sample: full history, one pass.

Results

Silver Bullet versions, NQ 1-minute data, gross (95% intervals are day-bootstrap)
Version	Trades	Win rate	Average R (95% interval)	Total R	Max drawdown
Version 1: true to spec	426	35.2%	−0.010 (−0.104 to +0.083)	−4.3R	not reported
Version 2: tight stop at the gap candle	426	22.8%	+0.029 (−0.138 to +0.201)	+12.3R	not reported
Version 3: fixed 1:2 target	426	10.6%	+0.039 (−0.059 to +0.134)	+16.4R	not reported
Version 4: ICT simple, no sweep	1,589	24.0%	−0.007 (−0.098 to +0.084)	−10.4R	not reported
Control: 12:00–13:00 window	77	3.9%	+0.013 (−0.233 to +0.283)	+1.0R	not reported

Win rate against the win rate the stop and target distances alone give (all filled trades)
Version	Win rate	Mean win rate from distances
Version 1	35.2%	51.0%
Version 2	22.8%	27.1%
Version 3	10.6%	33.3%
Version 4	24.0%	27.4%
Control	3.9%	27.1%

Version 1 with a daily-bias filter (average R, 95% interval)
Filter	Trades	When the bias agrees	When the bias opposes	Bias alone, no Silver Bullet
Overnight sweep side	140	+0.037 (−0.109 to +0.188)	−0.122	−0.003 (−0.042 to +0.036)
Asia range manipulation, 00:00–09:59	179	−0.035 (−0.160 to +0.091)	−0.131	−0.021 (−0.060 to +0.018)
Asia range manipulation plus midnight-open check	167	−0.033 (−0.163 to +0.099)	−0.061	−0.029 (−0.072 to +0.013)

Does a pre-10:00 bias call direction? Hit rate in a ±0.25% race from the 10:00 open; the drift baseline is 49.9%
Bias rule	Days labelled	Hit rate	Lift over baseline, points (95% interval)
Overnight sweep side	1,537	50.2%	+0.4 (−2.1 to +2.8)
Asia range manipulation	1,463	48.5%	−1.4 (−3.9 to +1.2)
Asia range manipulation plus midnight open	1,258	47.6%	−2.3 (−5.0 to +0.6)
New York AM manipulation (overnight box 18:00–08:29, sweep 08:30–09:59)	1,783	49.2%	−0.7 (−3.0 to +1.7)
New York AM manipulation, quiet nights only	1,020	49.0%	−0.9 (−4.0 to +2.2)

New York AM sweep of the overnight box, race from the first sweep bar (08:30–10:59 ET)
Race	Resolved days	Lift over each day’s geometry, points (95% interval)
Invalidation just past the sweep bar	2,183	−0.6 (−1.5 to +0.3)
Invalidation one box width past the sweep	1,397	+0.9 (−1.7 to +3.5)

New York AM sweep days: 2,192, with 1,426 control days. The sweep extreme held as the day extreme and the other side was then taken on 29.5% of days; the same measure was 7.2% on control days, which have a different box size.
Version 1 average R by year ranged from −0.23 (2021) to +0.26 (2023).

Questions and answers

How was the ICT Silver Bullet backtested?

On NQ futures 1-minute data from 2016-09-26 to 2026-08-27. A trade needs a sweep of a prior high or low, a 1-minute swing break, then a fair value gap inside 10:00–10:59 ET. Entry is a limit at the near side of the gap, the stop is past the sweep extreme, the target is the nearest untouched opposite level at least 10 points away, and the trade exits by 12:00 ET.

What sample size was used for the Silver Bullet test?

The spec version took 426 trades from 2,474 sweep days. The no-sweep version took 1,589 trades and the 12:00–13:00 control took 77 trades.

What was the win rate and average R of the ICT Silver Bullet on NQ?

The spec version won 35.2% of 426 trades with an average of −0.010R and a total of −4.3R, before costs.

Did adding a daily bias change the Silver Bullet results?

In three bias filters, the average R when the bias agreed was +0.037R (140 trades), −0.035R (179 trades) and −0.033R (167 trades); every 95% interval included zero.

Want your own strategy tested the same way? Every version is $19 at cemarketsolutions.com.

More tests like this on Instagram and TikTok: @cemarketsolutions.