At a glance
The short version
Everyone can point at a fair value gap after the fact. Almost nobody knows their own fill rate — how often the gaps they'd actually have traded got filled, and how far price ran when they didn't. That is a countable number, and counting it takes a replay that marks gaps as they form rather than after you already know the answer.
What a fair value gap actually is
Three candles. The first candle's high sits below the third candle's low (bullish), or the first candle's low sits above the third candle's high (bearish). The middle candle moved so hard that a band of price was never traded on both sides — an imbalance. The idea traders test is that price often returns to that band before continuing.
That is the whole definition, and it is why it can be detected mechanically. The disagreement is never about what a gap is — it's about which gaps are worth trading, and that is exactly the question a backtest settles. Read the long-form explanation in what is a fair value gap if you want the theory first.
How CRTLAB detects them (and why the details matter)
The detector is a streaming reducer: it consumes one candle at a time and is never handed a bar past your replay cursor. That single rule is the difference between a backtest and a highlight reel — a gap can't be marked out of price you haven't been shown yet.
Two inputs control what gets drawn. Min size filters gaps smaller than a set percentage of price, which is how you stop the chart filling with one-tick imbalances that nobody would trade. Fill rule decides what counts as mitigated: first touch, or a complete fill. Those two settings will change your measured fill rate more than anything else you do, which is precisely why they're settings and not a hardcoded opinion.
A tapped gap stays on the chart — it stops extending right at the candle that filled it, so a live level and a spent one look different, but it isn't deleted. Half the point of marking gaps is seeing where price came back to.
Running an FVG backtest that means something
Set your min size first and don't touch it again — a filter you tune mid-test is a filter fitted to the outcome. Then step forward candle by candle, and every time a gap forms that meets your rules, log it: did price return, how far did it run first, did the fill hold or blow straight through?
You need enough occurrences for the number to survive contact with variance. Fifty is a feel; a few hundred is an estimate. The sample size calculator tells you how many you need for the edge you think you have, and the risk of ruin calculator tells you what that win rate survives.
What a bare fill rate is worth, measured
We counted every fair value gap on 13 markets since mid-2022 — 273,191 of them on the 15-minute and 1-hour charts. On the 15-minute chart 87.3% were filled completely within 24 hours. Then we drew a fake gap for each one, the same distance from price on the other side, and it filled 87.2% of the time. A gap fills because it is close to price, not because it pulls price back.
So the fill rate you measure here only means something against that baseline, and so does the trade: the textbook retest (limit at the near edge, stop at the far edge, 2R) won 26.2% of 189,423 trades for −0.21R before costs. The full tables, by market, size and session, are in the fair value gap statistics. What your backtest has to show is that your gaps — after a sweep, in a session, with displacement — beat those numbers.
Which markets to test it on
Imbalance behaves differently by instrument, and that is testable rather than debatable. Index futures gap hard off the New York open and fill with a rhythm; FX majors leave smaller, more frequent gaps and drift back into them across sessions; gold produces some of the widest imbalances on the board and does not always come back for them.
CRTLAB carries the same detector across every market, so the comparison is apples to apples: NAS100, gold, EURUSD, DAX — same rules, same replay, different fill rates. Run the identical test on three of them and you will learn more in a weekend than a year of reading about gaps.
Frequently asked questions
Is the FVG backtester free?
CRTLAB is free to try. You get fair value gap detection on every market, with every tool the product has, for your first 10,000 candles of replay — roughly two hours of stepping through a chart, no card needed. After the sample, unlimited replay is $14.99 a month, cancel any time. The min-size filter and the fill rule are editable from the start — the two settings that move your measured fill rate most are not the paid ones, because there are no paid ones.
Does it repaint?
No. Every detector is fed candles strictly up to your replay cursor and never past it, so a gap appears exactly when its third candle closes and not a moment earlier. A backtesting tool that leaks the future is worse than one that draws nothing.
Can I change what counts as a filled gap?
Yes — the fill rule switches between first touch and a complete fill, and mitigated gaps stay visible on the chart rather than disappearing. Those two choices move your measured fill rate more than any other setting.
Can I test FVGs together with liquidity sweeps?
Yes. Concepts stack — run fair value gaps alongside a liquidity sweep and market structure on the same chart, which is how most people actually trade them rather than in isolation.
What timeframe should I backtest fair value gaps on?
The one you trade. Intraday traders usually work a 1–5 minute chart inside higher-timeframe context; the detector runs identically on every timeframe from 1 minute to weekly, so the honest answer is to test yours rather than the one that looks best.