Session-Anchored VWAP With Honest Bands: Complete Guide

What VWAP actually is

VWAP — volume-weighted average price — is the average price of every unit traded since a chosen starting point:

VWAP = Σ(price × volume) / Σ(volume)

Each bar’s price is weighted by how much volume traded there. A quiet drift on thin volume barely moves it; a heavy-volume push moves it a lot. That’s the entire reason VWAP means something while a simple moving average of the same length doesn’t: VWAP is the market’s actual average transaction price over the period. A trader who bought the session is, on average, in profit when price is above session VWAP and underwater below it. That’s why execution desks benchmark against it, and why price so often reacts around it.

One practical note for Gold traders: on spot XAU/USD, TradingView’s volume is tick volume — a proxy for activity, not true traded volume. It works well in practice as a weighting, but if you want true volume, chart COMEX Gold futures (GC) instead. We’d rather tell you that than let you discover it later.

The anchor is the decision

VWAP is meaningless without a starting point. “Anchored VWAP” just makes that choice explicit — and for intraday trading, the natural anchors are session opens:

• Daily anchor — resets at the daily open. The default institutional benchmark; the reference for the whole day.
• Asia anchor — resets at the Asia open. Useful early, but be aware of what the data says about Asia’s predictive power: in our study of 821 days, Asia’s direction told you nothing about London’s — London continued Asia only 49.3% of the time. Use the Asia VWAP as a location reference, not a directional one.
• London anchor — resets at the London open, giving you the average price of European participation.
• New York anchor — resets at the NY open. On Gold, this is the one to respect most: in our volume-by-hour study across 653 days, the 14:00–16:00 Paris block (the NY open window) carries the heaviest share of daily volume — 15:00 and 16:00 alone account for 8.4% and 8.5% of the day’s volume each, roughly double the London open hour’s 4.5%. The folklore says London is Gold’s big session; the volume data says New York is. A VWAP anchored where the volume actually is describes the positioning that matters.

Running two anchors at once — Daily plus the current session — is the standard setup: the daily VWAP gives the day’s big-picture value area, the session VWAP tracks who’s in control right now. When they’re far apart, that gap itself is information.

Why “honest bands” — the variance identity, in plain terms

Most VWAP indicators draw bands at VWAP ± k standard deviations. Here’s the problem: many compute that standard deviation as a plain, unweighted standard deviation of price — the same math as Bollinger Bands, just centered on VWAP. That’s internally inconsistent. If the average is volume-weighted, the dispersion around it must be too, or the bands don’t describe the same distribution as the line they’re drawn around.

The correct formula is the volume-weighted variance, and it comes from a classical identity:

Var(P) = Σ(v × P²) / Σ(v) − VWAP²

In words: the volume-weighted average of the squared prices, minus the square of the volume-weighted average price. This is just the familiar “mean of squares minus square of the mean” rule, with every average taken volume-weighted. The intuition:

• If lots of volume traded far from VWAP, the first term grows, variance grows, bands widen. Correct — the market genuinely transacted across a wide range.
• If price wicked far away on almost no volume, that excursion barely enters the sum, and the bands correctly ignore it. An unweighted calculation would widen the bands for a move nobody actually traded.

So volume-weighted bands answer the right question — “how far from average has traded volume been distributed?” — while unweighted bands answer a different one (“how far has price wandered, regardless of participation?”). On an instrument like Gold, where news wicks on thin volume are routine, the difference is not academic: unweighted bands inflate after every low-volume spike and then give you ±2σ “extremes” that aren’t extreme at all. This is why our VWAP Session Bands indicator computes the bands this way, and why we put “honest” in the name: it’s the statistically consistent version, implemented cumulatively from the anchor, per anchor.

What the bands actually contain: 830 days of data

Formulas are nice; frequencies are better. We measured daily-anchored VWAP bands on XAU/USD spot (Dukascopy data) from January 2023 to August 2026 — 830 trading days. Two baseline results:

• Price sat inside ±1σ 44.7% of the time.
• Price sat inside ±2σ 86.1% of the time.

Note what that means: this is not a normal distribution (which would put ~95% inside ±2σ). Gold spends about one bar in seven beyond its ±2σ VWAP band. “Beyond 2σ” is uncommon — but it is nowhere near rare enough to fade on sight.

The reversion number everyone quotes — and why it’s an artifact

Here’s the stat that launches a thousand fade strategies: after a touch of the ±2σ band, price came back to VWAP before the end of the day 97.6% of the time.

That number is real — we measured it ourselves. It is also deeply misleading, and we’re going to show you exactly why, because if you day trade Gold in the London or New York session it does not describe what you see on your screen.

The catch: that statistic counts only the day’s first ±2σ touch — and on a daily-anchored VWAP, roughly 98% of first touches happen in the quiet Asian morning (02:00–04:00 Paris), right after the bands have stabilized, when σ is still small and the entire day remains for price to revisit VWAP. Measured that way, 826 of 830 days produce a “touch”. It’s a near-degenerate statistic dressed up as an edge.

So we re-measured every ±2σ touch — 6,945 episodes across 830 days — and split them by session. Different story:

Two more honest details. First, “reverts to VWAP” means price touched a VWAP that kept moving toward price in the meantime — at the moment of reversion, the VWAP itself had already closed a median 30% of the initial gap. If instead you ask whether price returns to the VWAP level at the moment of the touch — which is closer to what a fade trade needs — the all-touches rate drops to 59%, and to 24% in New York. Second, the New York EOD number is partly runway (less day left), but the fixed 60-minute column makes the fair comparison and the ranking holds: Asia 34%, London 14%, NY 12%.

From reversion to expectancy — the regime table

“Price eventually revisits VWAP” was never a trade anyway. A trade has an entry, a stop, and a path, and the path is where fades die. Here is the close-based expectancy of fading the day’s first touch back to VWAP (stop at 3σ), split by regime classified from the prior day’s daily ADX/DI readings:

Read the last row twice. In downtrends, the reversion still happened 98% of the time — and fading the band still lost money per attempt. How can both be true? Because in a trending market the band touch keeps extending before it reverts: price runs through your entry, hits any reasonable stop, and only then rotates back to a VWAP that has itself been dragged toward price. The revert statistic counts that sequence as a success. Your account does not.

“Reverts eventually” ≠ “profitable fade.” Session and regime decide. The same signal — a ±2σ touch — is a +0.44R proposition in a transition regime and a −0.27R proposition in a downtrend, and a touch during the New York afternoon has less than a one-in-three chance of seeing VWAP again before the close. These are close-based study results, not live trading results (no slippage, fills at closes), so treat the absolute numbers as structure, not promises. But the ordering — across regimes and across sessions — is the lesson, and it’s a big one. Note the scope, too: because first touches cluster in Asia, the expectancy table essentially describes an Asian-morning fade.

Three workflows

1. The band fade (mean reversion) — only in Range/Transition

• Filter first: confirm a Range or Transition regime before the touch matters. This is exactly what a regime classifier is for — our Hurst Regime Classifier’s meta-signal literally switches between “use mean-reversion tools” and “use trend tools,” and this workflow is the mean-reversion branch. In a confirmed trend, especially a downtrend, skip the fade entirely; the table above is the reason.
• Location: price touches or pierces the ±2σ band of the session-relevant anchor.
• Trigger: don’t fade the first touch blind — wait for the stretch to stall and turn. We pair the band with the Kalman Mean Reversion indicator precisely because it requires a flat regime and velocity turning back before it signals; a band touch plus a KMR signal is a far better-defined trade than either alone.
• Management: target VWAP, stop beyond the extension with a volatility buffer — and size your patience by session, not by the folklore number: across all touches the median revert took ~2 hours, and a New York touch gets back to VWAP before the close less than a third of the time.

2. The VWAP pullback (trend continuation)

In a trending regime, invert the logic: VWAP and the ±1σ band become pullback zones in the direction of the trend. Price above a rising VWAP pulling back into the line or the lower ±1σ is the trend trader’s entry, with the trend itself doing the heavy lifting. Remember that price is outside ±1σ more than half the time (55.3%) — trending markets live in that outer region, which is exactly why fading them is expensive and joining them isn’t. Stops go beyond the opposite band or the last swing; the prior table’s Trend rows are your reminder of which side of this trade the math favors.

3. The breakout check (VWAP as truth serum)

When price breaks a level — a session high, a prior-day high, an opening range — the session VWAP tells you whether volume actually endorsed the move. A genuine breakout drags VWAP toward it, with price holding above VWAP and above the level. A break that immediately trades back through the session VWAP was, by definition, not backed by the session’s traded volume — a classic sweep signature. Timing matters too: be especially suspicious of breakouts in quiet pre-release mornings, where our news study found most morning breaks before major releases were traps, and of any structural read during the post-FOMC window, when normal band behavior goes out the window entirely.

Setup summary

• Anchors: Daily always; add the current session’s anchor (Asia/London/NY). On Gold, respect the NY anchor most — that’s where the volume is.
• Bands: ±1σ and ±2σ, volume-weighted variance (Var = Σv·P²/Σv − VWAP²) — if your indicator’s bands aren’t volume-weighted, the ±2σ label doesn’t mean what you think it means.
• Regime before signal: the identical band touch is a good fade in Range/Transition and a losing one in a downtrend. Classify first, act second.
• Statistics are context, not commands: every number above is a historical frequency from one dataset and period, measured on closes. Sample sizes and methodology are published; verify before you rely on anything.

If you want to see session-anchored VWAP with honest bands on your own chart — alongside the free session dashboard we give every trader — start with our free tier.

Delta Terminal ResearchStudies run on 3.5 years of XAU/USD spot data, developed hand-in-hand with a full-time Gold day trader. We publish the negative results too.