ICT and SMC concepts, systematically tested: every result in one table
Nine mechanical tests of ICT and Smart Money concepts on MNQ, NAS, MGC and XAU with realistic costs, one row each. The short version: most of them carry a small, real edge after costs (profit factors 1.2–1.5), none carry the edge the content around them implies, and every number below is attached to the trade count that produced it.
Every concept, one row
Each row links to the full test with the definition used, the instrument table and the drawdown profile. Win rate and profit factor are after commission and slippage on the sample stated; where a page is a method guide without a published result, the row says so.
| Concept | Sample | Win rate | Profit factor | Verdict |
|---|---|---|---|---|
| Fair Value Gap entry | MNQ 12 mo, 218 trades (4 instruments total) | 58% | 1.45 (1.18–1.45 across instruments) | Small, real edge after costs; strongest on MNQ and NAS, weakest on gold |
| Inverted FVG (iFVG) entry | MNQ 12 mo, 67 trades | ≈ +6 pp vs FVG | +0.15–0.17 vs FVG | Better per trade, too few trades on one instrument to stand alone |
| Liquidity sweep reversal | 2,000+ trades, multi-instrument | 53% | 1.32 | The dynamic is real; the 70–80% win rates in marketing are not |
| CRT (Candle Range Theory) | MNQ 5 min, 142 trades (4 instruments) | 54% | 1.31 (1.31–1.42 equity-style, 1.08–1.18 commodities) | Works on equity-style instruments, marginal on gold |
| Order block long entry | MNQ 12 mo, 5 min | 52% | 1.21 | Marginal; ICT mechanical rules land at PF 1.2–1.5, high variance |
| Combined “perfect” ICT setup | 38 trades | 64% | 1.78 | Sample too small to mean anything yet |
| Order blocks: how to test | method page | — | — | Parameter list and the random-zone control most tests skip |
| ICT Silver Bullet | method page | — | — | Fully codeable rule set and the control it has to beat; no headline number published |
| VWAP mean reversion, anchored vs static | strategy comparison | 65–75% (standard) | — | Fits the prop profile; trend days produce the few large losses |
Exit rules tested
Four common discretionary exits, each checked against the question that decides them: would acting on it have improved the system’s result, or only its feel?
| Exit trigger | Finding |
|---|---|
| Opposite FVG fills against you | May raise the odds of hitting the stop; still not an exit unless the system defines it |
| Order block forms against you | Visible only after the move; not actionable in time |
| Liquidity sweep against you | Compelling story, usually the wrong action |
| VWAP cross against you | Not a signal unless VWAP is in the strategy logic |
How to read this table
Three things decide whether a row means anything. The trade count: under roughly 100 trades the profit factor is noise, and two rows above are flagged for exactly that. The control: an entry with a tight stop and a 2R target has a characteristic win rate from geometry alone, so a result only counts against a random-placement control, which the order block guide explains. The instrument: every edge above is strongest on MNQ and NAS and weakest on gold, so a result quoted without the instrument is a result quoted without its sign.
For the comparison between a discretionary SMC process and a systematic one under prop firm rules, see SMC vs systematic prop trading; for how the variance in these numbers plays out across thousands of trade sequences, how to read Monte Carlo results.
The method, applied to the whole catalogue.
Every strategy and portfolio with modeled and live shown separately and the forecast behind each one. One PDF, then our letters with what the systems did.
By email, from Puerto Viejo, Costa Rica. Unsubscribe anytime.
Related
FAQ
Does ICT work when backtested?
Mechanically defined ICT entries show small positive expectancy after costs, profit factors in the 1.2–1.5 range on liquid futures, with high variance and strong instrument dependence. That is a real edge, not the one described in the content that popularises it.
Which ICT concept tested best?
Per trade, the inverted FVG on MNQ (higher win rate and profit factor than the plain FVG); across the whole sample, the plain FVG and liquidity sweep tests have the most trades behind their numbers.
Why are the samples so small?
One instrument on one timeframe produces 50–220 trades a year for most of these setups. Below roughly 100 trades the profit factor is noise, which is why every page states the trade count next to the result.
Are these results live or modeled?
Modeled: mechanical backtests on MNQ, NAS, MGC and XAU with commission and slippage, as stated on each page. The production strategies on this site are a different, separately reported set.
All figures are hypothetical, modeled outputs from mechanical backtests with commission and slippage, as published on the linked pages; none are live results. Past performance does not guarantee future results.