We Watched Trader Mane’s Entire ICT Course. Then We Backtested It.

We pinned the discretion before seeing the result

The original audit stopped at an honest boundary: a test could not choose “the” higher-timeframe zone without adding a rule the course does not contain. We have now added those rules openly. These are research translations, not a claim that Mane would take every signal they produce.

• Structure: strict three-candle swings, visible only after the third candle closes; trend changes only on a close through the latest unbroken swing.
• Location: the newest completed dealing range, its 61.8–78.6% OTE, and the newest fresh same-direction FVG that overlaps it. No later zone can replace the one selected at the decision time.
• Displacement: the middle candle body must be at least 1.5 times the prior 20-bar median, at least 60% of its range, with an FVG at least 0.10 ATR wide.
• Execution: one-minute limit fills after the signal candle closes, stop-first whenever one bar spans both levels, one position at a time, and two points of round-turn NQ friction. GBPUSD cost was normalized to H1 ATR.
• Evidence: development, validation, and forward-style splits; 1,000 daily-block bootstrap samples; 200 matched-date placebo paths; and unchanged GBPUSD replication rules.

The strict Nasdaq variants barely traded

The primary dataset contains 2,656,313 one-minute bars from January 2018 through August 2026. Yet the full top-down rules reduced more than eight years to five total trades across four variants. The positive V2 and V4 averages look attractive only until you read the sample column.

V1’s only trade lost. V2’s only trade was a 2020 time-exit winner. V4 had two development trades, no validation trades, and one forward-style winner. Its average was +0.950R, but the bootstrap interval ran from -1.041R to +1.975R and it reached only the 91.5th placebo percentile, below the registered 95th.

The controls had more observations—and no edge

C2 kept the weekly/daily/H4 direction and the 07:00–12:00 M5 trigger, but removed H4 OTE and H1-zone selection. It averaged -0.198R over 109 trades, with profit factor 0.74 and a bootstrap interval of -0.441R to +0.075R. C5 kept the AM sweep and M5 logic but removed D1/H1 direction and H1 premium/discount. It averaged -0.042R over 37 trades, profit factor 0.94, at the 53rd placebo percentile. Neither control was statistically distinguishable from a matched date shuffle.

GBPUSD did not replicate a usable edge

We reran the rules unchanged on 918,555 GBPUSD minute bars, scaling only friction and minimum risk to H1 ATR. The strict continuation rules produced no trades. V3 produced one winner while producing no Nasdaq trade. Strict V4 produced three trades and all three stopped.

The one V3 winner cannot validate either reversal stop. The larger replication controls pointed the other way: C2 averaged -0.081R over 35 trades and C5 averaged -0.293R over 14.

Why this series looked more testable than most

After the Silver Bullet lecture and the TradingView ICT guide, we have a working definition of “mechanical.” A strategy is tradeable, in the sense this blog uses the word, if a second person given only the stated rules produces the same trades. Confluence you cannot write down is not a filter. It is an unfalsifiable escape hatch: every loss can be explained as missing context.

Mane is unusually explicit about wanting the opposite. Episode 15: early on, “as robotic and mechanical as possible.” Episode 13: five steps, in order, or it is not a setup. Episode 1: a three-candle swing, not a five-candle one. That is more specification than most ICT derivative courses ever write down. It is also why the remaining holes are so easy to see. He drew the lines around them himself.

Twenty videos, one payload

Episodes 1–11 are a glossary with worked charts: structure, liquidity, dealing ranges, fair value gaps, order blocks, Fibonacci, Power of Three. Useful, and not a strategy. The framework he promised in episode 1 arrives in 12–14. Everything after that is risk, filters, and timing.

The live Bitcoin trade he shows in episode 14 — posted in Telegram as it happened — is the nested top-down loop, not a session-timed Silver Bullet. That is how he actually trades.

The 5-minute model, step by step

Episode 13 is the closest thing in the series to a checklist. The 15-minute and hourly versions in episode 14 are the same five steps on a slower chart. He is clear that none of it works in a vacuum: you do not open a 5-minute chart until weekly, daily, H4 and H1 work is done.

1. 1

   Wait for price to trade into the higher-timeframe zone

   Not close to it. If the zone is never tagged, there is no trade. The zone is an order block, fair value gap, breaker, swing-failure pattern, equal highs or lows, or the 0.62–0.79 optimal trade entry band — whichever you marked on the H4 after reading the weekly and daily.

2. 2

   Require a 5-minute market-structure shift in the higher-timeframe direction

   If you are looking for longs, the 5-minute is usually in a downtrend as it pulls into the zone. The trigger is a break of the last 5-minute lower high. Inverse for shorts. If the 5-minute keeps printing lower highs, the zone is not being respected.

3. 3

   That break must come with displacement

   A large middle candle that leaves a three-candle fair value gap. Displacement is how he tells a real shift from a stop hunt. He never gives a size: not an ATR multiple, not a body-to-range ratio. Larger is “more probable.”

4. 4

   Enter on the pullback to the fair value gap

   Limit or market at the gap, not on the displacement candle. If price never returns, you do not chase. An order block stacked on the same gap is confluence, not a requirement. “The fair value gap alone is enough.”

5. 5

   Stop at invalidation; target the higher timeframe; skip under 2:1

   Continuation: stop beyond the 5-minute swing that made the shift. Reversal: stop beyond the higher-timeframe zone. Targets are equal highs and lows and dealing-range extremes from the H4 and daily, with internal liquidity used for partials. If the math is under 2:1, pass.

That is the same structure we found in ICT’s Silver Bullet lecture: a precise trigger hanging off a preselected draw. The windows are mechanical. The draw is not. Mane replaced the draw with a higher-timeframe point of interest and told beginners the rest is robotic. The robotics start after the human has already picked the level.

What is specified tightly enough to code

Building blocks and filters first. A strategy is the combination. Several of these are cleaner than ICT’s own wording.

Three places the course disagrees with itself

Stop placement

Episode 12: the stop goes below the H4 point of interest, not the 5-minute low, because the 5-minute shift can fail and reprint inside a still-valid higher-timeframe zone. Episode 13: the stop goes below the 5-minute swing, with the higher-timeframe stop as an optional wider choice. Episode 14: use the wide stop for reversals and the tight stop for continuation. That split is a judgment about whether the daily higher-low is already in. A backtest has to pick one and hold it. He does not.

Timeframes

Weekly / daily / H4 / H1 / M5 is an example, not a requirement. “I don’t care what timeframes you use” as long as they nest. His public charts often collapse this to H12 and H1 because he does not want to stare at a 5-minute chart. A test that uses M5 on Nasdaq and a test that uses H1 on Bitcoin are both “the model.” They are not the same strategy.

Trend versus the live example

Beginners are told: weekly and daily bullish, only longs; bearish, only shorts. In episode 14 he takes a Bitcoin long while both are bearish, justified by a weekly swing-failure at range lows and a draw into a fair value gap overhead. He immediately says beginners should not take that trade. The series therefore contains both a rigid rule and a recorded exception. The exception is how he actually traded.

The seven disqualifiers

Episode 15 is the closest thing to a mechanical filter list. Any one of them, he says, means no trade. Four of the seven survive contact with a compiler. The other three are still a person.

The Silver Bullet he names is ICT’s

Episode 14 introduces it as a time-based version of the same 5-minute model, inside three one-hour windows, all New York time: 03:00–04:00, 10:00–11:00, 14:00–15:00. Episode 20 widens those to kill zones (London 02:00–05:00, New York 07:00–10:00) and says a setup at 10:00 on a Tuesday is not the same caliber as the same shape at 03:00 on a Saturday. He is explicit that the clock does not create the entry. It grades it.

He is also explicit that he does not know whether ICT’s algorithm story is true, that crypto follows the windows less than FX and futures, and that you should test it yourself. We first did that on the lecture ICT actually gave. We have now also tested a frozen version of Mane’s nested-zone model.

In that study, direction comes from a preselected draw on liquidity, not from the sweep. We mechanized all seven transcript-listed draw categories over 2.6 million Nasdaq one-minute bars. At two points of round-turn cost, no literal draw produced a statistically defensible edge. The least-bad version still lost 0.021R per trade. A non-causal draft that could skip an early gap and pick a later one looked slightly profitable until first-qualified-order sequencing erased it.

Mane’s contribution is not a new window. It is the claim that if your 5-minute model happens to print inside one, the trade is better. That is a filter on a model whose zone is still unchosen in the lectures. Once we pinned it, the confluenced AM variant produced only three Nasdaq trades. It beat the raw AM control numerically, but failed the sample, bootstrap, placebo, period-stability, and GBPUSD gates.

What the earlier tests measured on the overlapping pieces

The course is ICT’s PD arrays, nested, with a 5-minute FVG retest as the trigger. We have already run the arrays without Mane’s nesting. Dropping the discretionary layer biases those tests against a skilled reader of his charts. It also tests the part a rules trader could automate — which is the part he told beginners to treat as robotic.

None of that proves a discretionary reader of Mane’s nested charts would match those numbers. It does bound the claim that the 5-minute pattern, by itself, is the edge. Every time we have isolated that pattern, the edge has not been there.

Every registered decision gate rejected the models

We wrote the pass/fail criteria before reading the outputs. A plausible-looking average was never enough. A serious forward-test candidate needed sample size, stability across time, a profit factor above costs, a positive bootstrap floor, a placebo win, a directionally positive second market, and evidence that its confluence beat the negative control.

No model qualifies for a serious forward-test allocation under that rule. If the strict framework is pursued, the clean next step is prospective signal collection—not tuning these five historical trades. Any relaxed definition should be registered as a new experiment and judged on new data.

What this study is not

• It is not a backtest of Trader Mane’s live results, his Telegram calls, or the Bitcoin trade in episode 14. One narrated winner is a demonstration.
• It is not a claim that nested higher-timeframe context has no value. We tested explicit causal translations of that nesting; their samples were too small to estimate its effect. That is not the same as proving the discretionary idea has zero value.
• It is not a literal transcription of every visual judgment in the course. The point-of-interest tie-breaker and displacement thresholds are ours, disclosed in advance because the videos do not provide unique values.
• It is not a personality piece. He is unusually willing to show a counter-trend exception and to tell beginners not to copy it. That honesty does not close the specification gap.
• Auto-captions are imperfect. Where a sentence was garbled we checked it against the surrounding argument; we did not invent a rule the audio does not support.
• The primary report is price-only. No immutable economic-calendar file was available across the full span, so we did not fabricate a historical news filter from memory.

The finding we would defend is narrower than “ICT does not work.” Mane teaches an executable discretionary framework, but not one unique strategy. When we froze the ambiguities into four complete causal strategies, none produced enough stable, replicated evidence to call tradeable. Positive averages over one or three trades do not change that conclusion.