Part 1: Understanding the fair value gap · Chapter 5 of 14 · 9 min
Which FVGs hold: the criteria, one at a time
The criteria most often used to tell a strong FVG from a weak one (middle candle, range half, D1 trend, sweep, break of structure, order block), each measured on the 161 real gold FVGs. Only the strength of the middle candle holds up.
The criteria
For each criterion we compare the visited FVGs that meet it with those that do not. The outcome is the chapter 1 one: the zone holds if price reacts by 1.5 ATR before closing beyond the far side. Overall: 133 visits, 46% hold, margin of error 8 points.
| Criterion | With the criterion | Without |
|---|---|---|
| Middle candle with a body of 2 ATR or more | 21 of 33 (64%) | 40 of 100 (40%) |
| Right half of the H4 range | 18 of 43 (42%) | 43 of 90 (48%) |
| Aligned with the D1 trend (against D1) | 26 of 53 (49%) | against D1: 14 of 29 (48%) |
| Sweep before the gap | 23 of 45 (51%) | 38 of 88 (43%) |
| Break of structure | 20 of 40 (50%) | 41 of 93 (44%) |
| Nearby order block | 12 of 26 (46%) | 49 of 107 (46%) |
The margins of error are 10-19 points. With six criteria tried it is easy to find one that seems to work by chance, and it must be read with that caution.
The middle candle
It is the only criterion with a clear difference: 64% against 40%, which is 24 points. It agrees with chapter 2 (a body of at least 2 ATR: 64%, a body under 1 ATR: 37%) and with book 2 (a strong displacement goes with zones that hold more). A violent move leaves a gap the market has more trouble closing right away.
The criteria that showed nothing
The range half goes the opposite way to the common idea (42% with the right half, 48% with the wrong one). The D1 trend changes nothing (49% against 48%). The sweep and the break of structure add 8 and 6 points: they are signs in the expected direction, with margins of error of 15 points. The nearby order block changes nothing (46% and 46%), on only 26 cases.
An FVG with many criteria that holds
Figure 1 shows a supply FVG on H4, from 4 to 7 July 2026. It runs from $4,168.00 to 4,175.50 (7.50 dollars, 0.82 ATR; ATR 9.2 dollars). The middle candle (6 July at 00:00) has a body of 2.16 ATR. On 4 July at 16:00 there was a high at 4,179.85; on 5 July at 08:00 a candle climbs to 4,181.15 and goes past it: a sweep. The gap sits in the upper half of the range and D1 is falling, so it has three criteria. On 6 July at 16:00 price climbs to 4,170.00, 27% of the gap, and closes at 4,165.90. In the next ten candles the low is 4,093.34. Until 5 August no close goes above 4,175.50.
Read the chart in words
H4 gold chart from 4 to 7 July 2026. A dashed pink line at 4,179.85 dollars marks the high taken by a candle with a wick. A yellow band between 4,168.00 and 4,175.50 is a supply FVG left by a long bearish candle. Price climbs barely into the band, marked as the first visit, and then falls.
| When | What | Price |
|---|---|---|
| 6 Jul 2026 00:00 UTC | Supply FVG | 4,168.00 – 4,175.50 |
| 6 Jul 2026 16:00 UTC | first visit | 4,170.00 |
| 4 Jul 2026 16:00 UTC | Swing taken | 4,179.85 |
Times in UTC.
An FVG with many criteria that gives way
Figure 2 shows the opposite case, on H4, from 10 to 13 May 2026. The demand FVG runs from $4,681.45 to 4,711.78 (30.33 dollars, 1.44 ATR; ATR 21.1 dollars). The middle candle (11 May at 12:00) has a body of 3.04 ATR. A candle on 11 May at 00:00 drops to 4,672.59, below the 4,680.80 low of 7 May: a sweep. A close on 11 May breaks the structure: a CHoCH. On 12 May at 04:00 price comes back with a low of 4,689.28 and closes at 4,703.12. At 12:00 a candle closes at 4,668.47, below the far edge: the FVG fails within eight hours.
Read the chart in words
H4 gold chart from 10 to 13 May 2026. A dashed pink line at 4,680.80 dollars marks the low taken by a candle with a long lower wick. A yellow band between 4,681.45 and 4,711.78 is a demand FVG. Price comes back into the band, marked as the first visit, and a candle closes below the low edge.
| When | What | Price |
|---|---|---|
| 11 May 2026 12:00 UTC | Demand FVG | 4,681.45 – 4,711.78 |
| 12 May 2026 04:00 UTC | first visit | 4,689.28 |
| 10 May 2026 04:00 UTC | Swing taken | 4,680.80 |
Times in UTC.
Why a score does not help
Summing the five criteria that can be computed without D1 (2 ATR body, right half, sweep, break, order block) over the 133 visits:
| Criteria present | Visits | Hold |
|---|---|---|
| 0 | 29 | 11 (38%) |
| 1 | 43 | 21 (49%) |
| 2 | 42 | 17 (40%) |
| 3 | 16 | 9 (56%) |
| 4 | 3 | 3 |
The ladder does not climb: after one criterion the hold rate wanders between 40 and 56%, with small groups. A long score mixes the only criterion that shows something (the strong candle) with criteria that add nothing, and gives an impression of precision the data does not have.
A four-step routine
- Measure the middle candle's body in ATR: 2 or more is the only signal in the data.
- Note the sweep, break of structure, range half and order block, but do not use them as filters until you have checked them on your own data.
- Do not sum the criteria into a score.
- After a hundred FVGs check which of your criteria changes the hold rate, and drop the others.
Exercises
Chapter exercises
1. On the chart the supply gap has a sweep criterion. Which letter is the candle that takes the liquidity, meaning it goes past the swing high marked by the dashed line with its wick?
Read the chart in words
| When | What | Price |
|---|---|---|
| 4 Jul 2026 16:00 UTC | Point A | 4,179.85 |
| 5 Jul 2026 08:00 UTC | Point B | 4,181.15 |
| 6 Jul 2026 00:00 UTC | Point C | 4,204.41 |
| 6 Jul 2026 16:00 UTC | Point D | 4,170.00 |
Times in UTC.
2. FVGs with a 2 ATR or more middle candle held 64%, the others 40%. How do you read it?
3. With 0, 1, 2, 3 and 4 criteria the hold rate was 38%, 49%, 40%, 56% and 3 out of 3. How do you use it?
4. The right range half gave 42% and the wrong half 48%. What do you conclude?
Have a question or want to share your exercise?
Post in the community, or join the free signals room on Telegram.
Educational content, not financial advice. Trading involves risk.