Why Your ICT Trades Keep Failing — and How to Fix It

If you know what a fair value gap is, can name every kill zone, understand liquidity sweeps and market structure shifts — and you are still losing — the problem is almost certainly not that you need to learn another concept.

I spent about two years in exactly that position. I had the whole vocabulary by then. I could label a chart beautifully after the fact. What I could not do was string a month together, and every time I lost I went looking for the concept I must have been missing.

There wasn't one. What I was missing was a model, and the difference between having a vocabulary and having a model is the entire subject of this page.

Mistake 1: You took the first fair value gap you saw

This is the most common single error and it is worth being precise about why it fails.

On any given morning a 1-minute chart will show you a dozen fair value gaps. They are not equivalent. Most of them are noise — three candles that happened to leave a small imbalance while price was doing nothing in particular. A gap only carries meaning when it was created by displacement that means something, and "means something" has criteria.

A gap worth trading generally has all of these behind it:

• It formed after a liquidity sweep. Something got taken — a session high, relative equal lows, the previous day's extreme. Without that, price had no reason to reverse and the gap is just a gap.
• It formed on displacement. A decisive move, not a drift. If the candles that created it look like every other candle on the chart, no institutional order flow left that mark.
• It sits on the correct side of the range. Buying a gap in premium means paying above the midpoint. That is the wrong half.
• It agrees with your higher-timeframe bias. A bullish gap in a bearish day is a place price passes through on its way down.
• It formed inside a kill zone. A gap at 1:15 PM on a quiet Tuesday has thin participation behind it.

Count those. Five conditions. The first gap you see on a chart satisfies, on average, one or two of them. That is why taking it feels like following the method and produces results that look random — because you have selected a gap on the single criterion of being visible.

Mistake 2: You traded against your bias, or never formed one

Ask yourself honestly: before this morning's session, did you write down which direction you expected price to be delivered, and why?

If the answer is no, every setup you saw looked equally valid, because with no directional filter both sides always look reasonable. There is always a bullish gap and a bearish gap on the same chart. Without a bias you are picking between them on feel, and feel reliably picks the one that just moved.

The second version is worse and more common: you formed a bias, then took a setup against it because the setup looked good. That is not flexibility. It is the bias being abandoned at the exact moment it was supposed to do its job, which is to stop you taking the pretty countertrend setup.

Daily bias comes from the higher timeframe — where price sits in the range, which pool of liquidity is unfilled, what the draw on liquidity is. It does not come from the 1-minute chart, which during the pre-session is engineered to look convincing in whichever direction is about to reverse.

Mistake 3: You entered before the confirmation closed

The market structure shift requires a body close through the opposing swing. Not a wick through it. Not price trading there and looking like it will close there. A closed candle body.

Entering on the wick feels like getting a better price. What it actually does is remove the only mechanism that distinguishes a genuine shift from the algorithm testing a level and rejecting it. Wicks through levels are constant. Body closes through them are not, and that difference is the entire filter.

The same applies further up the sequence. Price approaching a level is not price reacting to it. Price tagging an order block is not the order block holding. The 2024 material is explicit about waiting for price to give a clue rather than trading the level on arrival, and the clue is always a close, never a touch.

If you find yourself thinking "it's basically going to close there" — that thought is the mistake. Wait eleven seconds and know.

Mistake 4: You chased the one you missed

This one is emotional rather than technical, and it does the most damage.

The sequence is always the same. You watch a setup form perfectly. You hesitate, or you were making coffee, or you wanted one more candle of confirmation. Price leaves without you. And then, forty points later, you take an entry that has none of the criteria, because being out while price runs feels worse than being in a bad trade.

It does not feel like FOMO in the moment. It feels like recognising that you were right about direction. But the trade you have just taken is a different trade from the one you analysed: the level is gone, the risk-to-reward has collapsed, and your stop is now wherever it happens to fit rather than where the structure says it belongs.

The honest reframe is that a missed setup costs you nothing. A chased one costs you real money, and it usually costs you the next hour of judgement as well.

The real problem: you have a vocabulary, not a model

All four of the above are symptoms. This is the cause.

A vocabulary is a set of concepts you can recognise. A model is a written sequence with entry criteria, invalidation, and a stop and target rule, that you have tested on enough historical sessions to know how it behaves — including how often it produces nothing.

The last row is the one that separates people. Without a tested model you have no idea how often your setup is supposed to appear, so a day with nothing feels like failure and you go looking. With a tested model you know roughly what proportion of sessions qualify, and an empty morning is simply a data point that matches expectation.

Why backtesting is the actual fix

I know backtesting sounds like homework you can skip. It is not, and it is not primarily about proving the model works.

The real value is that it tells you what your setup looks like when it does not qualify. That is the knowledge that stops you taking the first fair value gap you see, because you have already seen two hundred gaps that failed the criteria and you recognise them on sight.

Work through sessions in order, one model, one instrument, and log every session — including the ones with no trade, which is where most of the learning is. What you are building is not a statistic. It is a reference library of what qualification actually looks like.

A useful discipline: mark your levels before you scroll forward. A chart annotated after the move always agrees with you, and it teaches you nothing at all.

Why you need far less than you think

A lot of the FOMO and the over-trading traces back to a belief that you need big wins. Look at what the framework actually asks for.

ICT has talked for years about modest daily objectives — a handful of points on ES rather than a home run. ES moves $50 per point per contract, so the arithmetic is simple:

I want to be careful with that table, because tables like it are how people get sold things. It is arithmetic, not a projection. It assumes twenty winning sessions in a month, which nobody has. Most days the model declines and you take nothing. Losing days subtract. Nobody, including me, is telling you those numbers are achievable.

The reason it is worth showing is the opposite of the usual reason. It demonstrates how little you need from any single session. Five points is a fraction of a normal ES day's range. It is well inside what one clean setup in a kill zone can produce. You do not need to catch the whole move, you do not need three trades a day, and you certainly do not need the one you missed at 09:47.

Almost every mistake on this page comes from believing the opposite — that you need to be in, that you need more, that a small target is not worth taking. Take the five points and go and do something else. The model does not require heroics and neither does the maths.

The routine that removes most of this

Nothing here is clever. It is the same short sequence every morning, and it makes the four mistakes above structurally harder to commit.

1. Before the session — write the bias sentence. Direction and reason, or you watch today.
2. Mark the levels — liquidity above and below, unfilled gaps, the draw. Before the open, not during.
3. Write the criteria you need — the actual checklist for your one model, visible on screen.
4. Wait for the window — if the kill zone is not open, there is no trade to consider.
5. Place a limit, not a market order — at the price your model says. Fills or does not.
6. Log the session either way — including "no setup", which is the most useful entry you will write.

If that sounds boring, that is the point. Every one of the mistakes above happens in the gap between seeing something and acting on it, and a written routine is what fills that gap with a decision you already made when you were calm.

Frequently Asked Questions