ICT Silver Bullet Backtester

The Silver Bullet runs inside three one-hour windows, all in New York time: 3:00–4:00 AM (the London open), 10:00–11:00 AM (the New York morning) and 2:00–3:00 PM (the New York afternoon). Most people who say "Silver Bullet" mean the 10 o'clock hour on NQ or ES, and that is the one worth testing first.

Inside the window you wait for price to take liquidity — run a nearby high or low that the market has been resting against — and then leave a fair value gap as it displaces back the other way. The entry is the retrace into that gap. The target is the next obvious pool of liquidity in the direction of the displacement. No gap inside the hour, no trade.

That last sentence is the whole reason this model is testable. It has a no-trade condition, and a strategy with one can be counted day by day instead of cherry-picked from the days it worked.

The three Silver Bullet hours. Times are New York local — they move with US daylight saving, not with yours.

Every Silver Bullet video shows a gap that filled and ran. What none of them show is the denominator: how many 10 o'clock hours produced no gap at all, how many gaps formed on the wrong side of the sweep, and how many retraced straight through the gap and kept going. A win rate without those days is not a win rate.

The second problem is timing drift. Scroll back through a chart you have already seen and it is very easy to decide that the 10:52 gap "counts" and the 9:58 one does not. A replay that shows you nothing past the current candle takes that decision away from you: at 10:00 you do not know whether the hour will give you anything.

The third is tape. The Silver Bullet window sits right after the 9:30 cash open and, often, a 10:00 economic release. A test that ignores which days had a release is mixing two different markets into one number — see backtesting with news on the chart.

The button on this page opens NAS100 on a 1-minute chart with two of the model's three ingredients already marked: fair value gaps and liquidity sweeps, both detected candle by candle and never from a bar past your replay cursor. Switch on Sessions on the right rail and the killzone bands show you where each window starts and stops, in New York time.

Then the routine is simple and boring, which is the point. Start the replay before 10:00. When the hour opens, step one candle at a time and write down, before the next candle prints: did price take liquidity, did a gap form after it, did price come back into the gap. Mark the entry, stop and target the way your rules say. At 11:00 the window closes whether you traded or not — log the day either way.

Twenty trading days is a month of 10 o'clock hours and one evening of replay. The honest number usually shows up somewhere around the hundredth day: the sample size calculator tells you how many you need for the edge you think you have.

Five columns are enough: date, did a sweep happen, did a gap form after it, result in R, was there a 10:00 release. The empty days are not wasted entries — they are how you find out whether the model gives you a trade three days a week or three days a month, which changes everything about how you size it.

Once you have fifty logged days, split them. Release days against quiet days. Gaps that formed in the first twenty minutes against the last twenty. Long against short. The Silver Bullet that survives those splits is a strategy; the one that only works in the aggregate is usually one lucky week.

The model was taught on index futures, and NAS100 is where it is tested most: the 10 o'clock hour on the Nasdaq is fast enough to leave clean gaps and deep enough to fill them. SP500 is the calmer sibling and a good second sample. For the 3 AM window, gold and EURUSD are the natural candidates — that is when their day actually starts.

Run the same written rules on two markets before you trust either. If the edge only exists on one instrument in one month, you have found a description of that month.