Key takeaways
- Fading the band works. Both 2-sigma signals beat their controls at both targets, at the highest significance level we report.
- The cross does not. Reclaiming session VWAP is flat at a 1:1 target across more than three hundred thousand trades.
- Losing VWAP is worse than random. A significantly negative edge - the trade taught as a short trigger.
- Certain is not the same as large. These z-scores come from 200,000-plus samples, not from big effects. The edges are still under two points.
What it does
VWAP is the average price paid over a session, weighted by volume: every trade counts in proportion to its size, so a large print moves the line more than a small one. It resets at the start of each session, which is what separates it from a moving average — VWAP answers "what has the average participant paid today", not "what has price averaged over the last N bars".
That anchoring is why institutions use it as an execution benchmark, and why it is the one indicator here that genuinely requires volume and an intraday session boundary. We measure it on hourly bars anchored to the UTC day, with bands drawn at two volume-weighted standard deviations either side.
How this was measured
Every figure below comes from our own tick-derived archive, not from a vendor summary. The method is identical to the one behind our indicator signal study, so the numbers here are directly comparable to every other signal we publish.
- Data. Dukascopy bid-side OHLC for 93 instruments — major and cross FX pairs, gold, silver, stock indices and crypto — from 2010 to August 2026.
- The trade. Enter at the open of the next bar. Stop at 1.5 × ATR(14), targets at 1:1 and 1:2, walking forward up to 40 bars.
- Ties go against you. A bar holding both stop and target is scored a loss every time.
- Real costs. Per-instrument median spreads from our own bid/ask measurement.
- The control. For every signal, a trade with the same direction and the same stop distance at a random bar — five draws each. The gap between the two columns is all the signal is worth.
The result
Hourly bars, VWAP re-anchored each UTC day. Breakeven is a 50% win rate at a 1:1 target and 33.3% at 1:2. "Matched random" is a trade in the same direction with the same stop distance, opened at a random bar.
| Signal | Target | Trades | Win rate | Matched random | Edge | Significance | Beats breakeven? |
|---|---|---|---|---|---|---|---|
| Close above VWAP +2 sigma | 1:1 | 210,087 | 48.9% | 47.5% | +1.3 pts | *** | No |
| Close below VWAP -2 sigma | 1:1 | 204,496 | 48.8% | 47.5% | +1.3 pts | *** | No |
| Price reclaims session VWAP | 1:1 | 335,834 | 47.4% | 47.6% | -0.2 pts | not significant | No |
| Price loses session VWAP | 1:1 | 340,867 | 47.2% | 47.6% | -0.4 pts | *** | No |
| Close above VWAP +2 sigma | 1:2 | 202,809 | 31.9% | 30.3% | +1.6 pts | *** | No |
| Close below VWAP -2 sigma | 1:2 | 196,506 | 31.9% | 30.4% | +1.5 pts | *** | No |
| Price loses session VWAP | 1:2 | 330,141 | 30.9% | 30.3% | +0.6 pts | *** | No |
| Price reclaims session VWAP | 1:2 | 322,551 | 31.0% | 30.4% | +0.6 pts | *** | No |
The split here is unusually clean. The two mean-reversion signals — fading a close beyond either 2-sigma band — beat their controls at both targets with the strongest statistics in the study. The two trend signals, price reclaiming or losing session VWAP, are flat or negative at a 1:1 target. The trade that tutorials describe as the VWAP entry is the one that does not work.
A caution that matters more here than anywhere else on this site: these significance markers come from samples of two hundred thousand trades and upward. That makes the direction of the effect close to certain. It does not make the effect big. A +1.6 point edge is statistically overwhelming and economically thin, and on hourly bars you are paying the spread often enough for that distinction to decide the outcome.
How to use it
- Fade the stretch, do not chase the cross. The measured edge is in trades taken against a 2-sigma extension from the session anchor, not in the reclaim.
- Respect the session boundary. VWAP means nothing carried across sessions; the anchor is the whole idea. We drop the first few bars of each session, when a one-bar VWAP is just that bar's own price.
- Count your costs first. At this edge size and this trade frequency, the spread is a bigger lever than the signal. If your execution is poor, none of this survives.
- Use the bands as context for other setups. Knowing price is two sigma from the day's anchor is useful information for a trade you were already considering.
What does not work
- "Price reclaimed VWAP, go long." Flat against its control at a 1:1 target across more than three hundred thousand trades.
- "Price lost VWAP, go short." Significantly worse than a random short with the same stop.
- Treating a huge z-score as a big edge. The samples here are enormous. That makes small effects measurable, not profitable.
- Applying VWAP to daily bars. It is a session tool. Without an intraday anchor it is not VWAP.
Limitations
- Default parameters only. Tuning them on this same data would be curve-fitting.
- One stop rule (1.5 × ATR) and two fixed targets. Trailing stops, time exits and scaling are untested.
- Signals are taken in isolation, with no trend filter or confluence — which is how the isolated value of each is measured, but not how most people trade them.
- Bid data with modelled ask. Spread is a measured median, not a replayed tick-level ask feed.
- Spread only — no commission or swap. Both would push results further down.
- A statistically significant edge of one point is still only one point. Large samples make small effects certain; they do not make them large.
Cite or republish this data
Free to quote, screenshot, or republish — in an article, a video, a newsletter, or a course — as long as you credit FxBacktest with a link back to this page.
Frequently asked
Does VWAP work?
The mean-reversion reading does; the trend reading does not. Across more than a million hourly signals on 93 instruments, fading a close beyond a 2-sigma band from session VWAP beat a matched random entry by up to +1.6 points at both targets. Price reclaiming session VWAP was flat at a 1:1 target, and price losing session VWAP was significantly worse than a random short with the same stop distance.
Should I buy when price crosses back above VWAP?
Not on this evidence. That is the most commonly taught VWAP entry and it produced no edge over a random long at a 1:1 target across more than three hundred thousand trades. The signal that did produce an edge was the opposite shape - fading a stretch away from the anchor rather than joining the move back through it.
What are VWAP bands?
Lines drawn a set number of volume-weighted standard deviations either side of session VWAP, measuring how far price has stretched from the average price paid so far that session. We tested the 2-sigma bands, and closes beyond them were the only VWAP signals to beat their controls at both targets.
Why does VWAP reset every session?
Because it answers what the average participant has paid during that session, which makes it an execution benchmark rather than a trend line. Carrying it across sessions would destroy the property that makes it useful. We anchor to the UTC day, the only boundary present across every instrument in the archive, and skip the first few bars of each session where a VWAP calculated from one or two bars is meaningless.
How were these VWAP numbers produced?
Hourly bars with tick volume, VWAP re-anchored at each UTC day. Enter at the open of the bar after the signal, stop 1.5 x ATR(14) away, targets at 1:1 and 1:2, walking forward up to 40 bars. Bars containing both stop and target are scored as losses. Real per-instrument median spreads are applied. Every signal is compared against five random-entry draws in the same direction with the same stop distance. Data is Dukascopy bid OHLC for 93 instruments, 2010 to August 2026.