What a fair value gap is
A fair value gap (FVG) is a three-candle pattern: the middle candle moves so far that the candles on either side of it don’t overlap. The price range between them was traded through quickly and only one way. A common view is that price tends to come back to “fill” that range, and that a gap left by a strong move can act as support or resistance when it does.
Every FVG indicator draws those gaps. Almost none of them tell you what gaps like the one on your screen actually did afterwards. So we counted.
How we counted
For every gap we asked two questions:
- Did it fill? Did price come back and cover the gap within a set number of bars?
- Did it hold? Would a simple trade from it have worked: a limit order halfway into the gap, a stop just past the move that created it, and a target at one, two or three times the risk (1R, 2R, 3R)?
Then we split the gaps two ways: by size compared with recent volatility, and by how they were made. An opening gap is one that appeared between two candles, overnight or over a weekend, without any candle actually making the move. A candle gap is one made by a single candle’s move, with the whole gap inside it.
We ran it on gold, GBPNZD, AUDUSD and NAS100. On intraday charts the same picture came back on all four.
Intraday: every gap is the same gap
On gold’s 15-minute chart, across about 3,200 gaps:
- 85% filled
- 48% held at 1R
And no kind of gap did better than any other. Large or small, with the trend or against it, the hold rate barely moved. When we scored each gap from the ones before it and checked those scores against what happened later, out of sample, the skill came out at zero.
That is the honest answer that most FVG content won’t give you: on intraday charts there is no such thing as a “high-quality” gap. Roughly half hold and half don’t, whatever they look like. If you trade intraday gaps, the gap is a coin flip and your edge has to come from somewhere other than the gap.
Against a plain level
A gap could hold half the time simply because price comes back to any level that close. So we compared every gold 15-minute gap with two plain levels of the same size, distance and stop: one on the other side of price at the same moment, one on the same side weeks away. On three data feeds:
- Fill: 85-87% for the gaps, 85-87% for the plain levels. No gap effect.
- Hold at 1R: 47-48% for the gaps, 45-48% for the plain levels. A small, consistent edge for the gap, 1 to 3 points, and not enough to trade on its own.
Daily: a correction
An earlier version of this page said that on daily gold 73% of gaps are opening gaps that hold only 22%, against 48% for gaps a candle made. That was wrong. Scrolled all the way back, the daily gold history on some feeds starts decades ago with single-price bars, and between two dots every move counts as an opening gap. On daily bars with a real range, opening gaps barely exist, and gaps hold about as often as a plain level at the same distance.
What to do with it
- On the 15-minute chart: stop waiting for the “perfect” gap. Treat FVGs as places price is likely to revisit, and look elsewhere for the reason to trade.
- On any timeframe: before you trust a count over long history, look at the oldest bars. A bar with no range makes fake gaps.
- Everywhere: the gap alone is not a reason to trade. A gap inside a higher-timeframe order block tested no better after costs either.
A few caveats. These figures come from the loaded chart history and ignore spread and commission. Gaps that form close together aren’t fully separate events. And a hold rate is a description of the past, not a promise about the next gap.
GAP DIGGA does this on your own chart
GAP DIGGA is free on TradingView. It runs the count above on whatever chart you open, any symbol and any timeframe, labels every gap with its fill and hold rates, splits gaps by size and by how they were made, and shows a panel that says plainly whether its own scores have been right on that chart: yes, a little, no, or overconfident. On intraday gold it says no, and that is the correct answer.