Fair Value Gap (imbalance): how it forms, how to measure it, how to handle it

The Fair Value Gap (FVG), also called imbalance or inefficiency, is probably the easiest concept to define in a completely objective way: it's based on three consecutive candles and an inequality between prices. That makes it ideal to measure, test and program (we'll do it in Python). Let's see how it forms, how it's used and what the data really say.

The three-candle definition

Take three consecutive candles C1, C2, C3. The second, C2, is the "strong" one (large body). An FVG forms when between the high of C1 and the low of C3 (or vice versa) there remains a non-overlapping space:

• Bullish FVG: the high of C1 < low of C3. The zone between the two is the gap.
• Bearish FVG: the low of C1 > high of C3.

Bullish FVG:  C1.high < C3.low     zone = [C1.high , C3.low]
Bearish FVG:  C1.low  > C3.high    zone = [C3.high , C1.low]

The gap isn't a "hole" between two closes (like Monday's opening gap): it's a space in which, in candle C2, price moved so fast that no "overlapping" trading between C1 and C3 ranges was left.

Why it matters (the hypothesis)

The hypothesis is that such a fast move leaves unexecuted orders ("inefficiency"), and that price tends to return to fill at least part of that area. In practice a return is often observed, but:

• Not always: many FVGs are never reached.
• Not all the way: price often fills only part.
• Not at the same speed: sometimes it happens after a few minutes, sometimes after weeks.

These very percentages (how many fill, how much, how quickly) make the FVG an excellent experimentation ground.

Filtering FVGs by size

On a chart there are hundreds of micro-gaps, and most are noise. A widely used filter: consider only FVGs whose size is at least a fraction of the ATR, for example 0.3-0.5 × ATR(14). This way:

• you avoid meaningless 1-pip gaps;
• the threshold adapts to volatility.

Partial fill and "consequent encroachment"

Price doesn't always fill the whole gap. Useful conventions (to put in writing):

• Partial fill: price enters the gap without crossing all of it.
• *50% fill (consequent encroachment, CE):* price reaches the gap's midpoint. It's a much-used level as an entry point or a "reaction" threshold.
• Full fill: price closes beyond the opposite side: the gap is invalidated (in ICT logic it can become an inverse FVG, see below).

You can measure the statistic for your pair: "in how many cases does price reach at least 50% of the FVG within 20 candles?". It's a number that varies enormously across pairs, timeframes and periods: that's why you shouldn't trust the one you read elsewhere.

Inverse FVG (IFVG)

When an FVG is crossed on a close from the opposite side, some traders consider it "flipped": a bullish FVG broken downwards becomes an area that can act as resistance on the return (and vice versa). It's this language's equivalent of the role reversal principle of support and resistance. As always, it's a hypothesis to test.

How it's used: three applications

1. As an entry area: after an impulse that breaks structure and leaves an FVG, wait for the return into the gap in the direction of the move and enter (with confirmation or a limit at the CE).
2. As confirmation of strength: an FVG created by the move that breaks structure is a clue to displacement. A BOS without an FVG is, in this logic, weaker.
3. As a target: an unfilled opposite FVG, in the direction of your trade, is a point where price might be attracted.

What the data say (and what they don't)

In the analyses you can run yourself on historical data, these aspects typically emerge:

• A consistent but not total share of FVGs is filled at least partly, often within a few dozen candles.
• That share depends on the size filter, the timeframe and the regime (in strong trends, fewer counter-trend FVGs fill).
• An FVG that fills doesn't imply that price then resumes the impulse's direction: they are two distinct events.
• Many FVGs fill simply because price oscillates and passes through there, as it would for any random level.

That's why it's always wise to compare with a control: how often does price reach a randomly chosen level (at the same distance) in the same interval? If the FVG fill rate is similar, the gap has nothing special.

Common mistakes

• Treating all gaps as FVGs without a size filter.
• Expecting a guaranteed bounce at the first touch.
• Ignoring context: an FVG against the high-timeframe trend carries less weight.
• Confusing the FVG with the opening gap.
• Not recording the cases where the gap is never reached (price runs away and doesn't return).

In short

• Bullish FVG: high of C1 < low of C3; bearish: low of C1 > high of C3.
• Filter by size (for example ≥ 0.3-0.5 × ATR) to remove noise.
• The CE (50% of the gap) is a useful level; a full fill invalidates the gap.
• Use it as an entry area, confirmation of displacement or target.
• Measure how many FVGs fill and compare with a control: without a test there is no evidence.

Practical exercise

1. Mark all FVGs of the last week on M15 with a 0.4 × ATR threshold.
2. Calculate in how many cases price reaches at least 50% within 20 candles, and in what median time.
3. Repeat on three pairs and compare with a control level at the same distance.