VWAP, and why the anchor is the whole story

The calculation is the least interesting part, so it goes first and briefly. For every bar, multiply a representative price — usually the average of high, low and close — by the volume that traded in it. Keep a running total of that. Keep a running total of the volume. Divide one by the other.

What you get is the average price at which business was actually done, with busy prices counting for more than quiet ones. That is genuinely different from a moving average, which treats a bar where nothing traded exactly like a bar where everything did.

It matters more than most indicators for a reason that has nothing to do with chart analysis: it is a benchmark used to judge execution. Institutions filling large orders over a session are measured against the day's VWAP — buying below it is a good fill, above it is a poor one. So a lot of real activity is organised around this number. It is one of the few lines on a chart that participants are being paid to care about.

The anchor decides everything

VWAP is cumulative. It has to start somewhere, and where it starts is not a detail — it is the whole measurement. «Average price since when» is a different question for every value of when.

This produces a trap that is almost never mentioned, and it is the reason this section exists:

Your session VWAP resets when your platform thinks the session begins — and that is decided by the trading hours template you selected, not by the exchange and not by the chart. Change the template and the line moves. Two traders, same instrument, same platform, different templates: different VWAP, same name.

If you use it as a reference others are watching, that matters a great deal — the value of a shared level is that it is shared. The article on sessions and trading hours covers where those boundaries come from, and candle types flags the same template quietly deciding where your first bar of the day begins.

Anchored VWAP

Once you accept that the anchor is a choice, the obvious next step is to choose it deliberately: start the calculation at an event rather than at a clock time. The session high. A capitulation low. The minute a release landed.

The idea is older than its current popularity. It comes out of the MIDAS work of Paul Levine, a physicist who built it in the mid-1990s by launching volume-weighted curves from significant turning points, and it was brought to a much wider audience later by Brian Shannon.

What it gives you is specific and modest: the average price paid by everyone who has transacted since that event. If the event was a sharp low and price is now above that line, the people who bought into the move are, on average, holding a gain. That is a statement about position, not a forecast, and it is most of what the tool is for.

The deviation bands

Most platforms will draw bands at one, two and three standard deviations from the VWAP. These behave like Bollinger bands with a different centre, and they inherit the same warning: a band is not a barrier. A trending session spends its time outside the first band, and that is what trending looks like here, not a reason to fade it.

It goes deaf as the day goes on

This is the practical limitation nobody puts in the tooltip, and it follows directly from the arithmetic.

VWAP is a running total. In the first fifteen minutes it is computed from very little volume, so each new bar moves it noticeably. By late afternoon it carries the whole session's volume, and a single bar — even a violent one — barely shifts it at all.

So the same line is two different instruments at two times of day: jumpy and unreliable early, stable and unresponsive late. Neither is a defect. But «price crossed VWAP» means something quite different at 09:35 than it does at 15:30, and a rule written for one will behave strangely in the other.

VWAP and the volume profile are not the same thing

Both are described as «where the volume was», which hides a real difference.

VWAP — an average

One number, pulled by every trade in proportion to its size. It can land at a price where very little actually traded, in the same way that the average of two crowds standing far apart is the empty space between them.

Point of control — the busiest price

Not an average at all: the single price where the most volume changed hands. By definition, business happened there. Volume profile basics is the longer version.

They frequently sit near each other, which is why the difference goes unnoticed. When they separate, the separation is the information: the average is being dragged by activity that the busiest price is not near.

What VWAP does not tell you

• Who is above or below it. «The average buyer is offside» is a summary of a crowd, and a crowd contains people with completely different positions, timeframes and intentions.
• Anything about tomorrow. Reset it at the next session and the whole measurement starts again from nothing.
• Whether the volume was aggressive or passive. It weights by size, not by who crossed the spread. That distinction is delta, and it is a different measurement entirely.