VWAP & anchored VWAP
Learning objectives
- Compute session VWAP and its volume-weighted σ bands, and say which session start (18:00 ET Globex or 09:30 ET RTH) your platform uses.
- Use above/below VWAP as a description of who has controlled the session so far, not as an entry signal.
- Read the evidence grade of "price reverts to VWAP from its bands" and "anchored VWAPs act as support/resistance", and treat a band touch as context only.
- Anchor VWAP to an event, read it as the average price traded since that moment, and turn it into a written if-then plan with a stop where the read is plainly wrong.
It is 10:00 ET (22:00 ICT, winter) on a one-sided morning. NQ opened at 21,000.00 and now trades 10 points above its RTH VWAP, just past the +2σ band. The tempting read is “stretched, fade it back to VWAP”. This lesson is about what the VWAP line and its bands tell you for certain (an average, and how far price sits from it), what they do not (where price turns), and what the tests say about the popular uses.
Key terms
VWAP (giá bình quân theo khối lượng): the volume-weighted average price of every contract traded since a start time Definition. VWAP = Σ(price × volume) ÷ Σ(volume). It is the average price paid by every buyer, and received by every seller, since the start. Above it, the average buyer since the start is ahead; below it, the average seller is.
VWAP bands (dải độ lệch chuẩn VWAP): VWAP ± k × σ, where σ is the volume-weighted standard deviation of the traded prices around VWAP Definition. σ = √(Σ(price² × volume) ÷ Σ(volume) − VWAP²). The bands say how dispersed today’s trading has been; ±2σ marks prices far from where most of the day’s volume traded.
Anchored VWAP (VWAP neo, AVWAP): the same calculation started from a moment you choose, such as a news release, a swing high or the week’s open Definition. It answers a narrower question: “since that moment, is the average buyer or the average seller ahead?”
Term you need: session start. “Session VWAP” depends on where the sum starts. A 24-hour (ETH) VWAP starts at the 18:00 ET Globex open (06:00 ICT, winter) and includes the thin overnight trade; an RTH VWAP starts at 09:30 ET (21:30 ICT, winter). They answer different questions and can sit many points apart. Check which one your platform draws before you compare notes with anyone.
Chart first: a one-sided day along its bands
Price rides the bands instead of returning to VWAP
LoreOne trader's studyHow it works
Why VWAP is the line everyone draws
VWAP is the standard execution benchmark: a fund that buys over a day is often judged by whether its average fill beat the day’s VWAP Definition. A VWAP execution algorithm slices a large order into small ones and trades them roughly in proportion to the market’s usual volume through the day, so its average fill lands near VWAP. Notice what that does not mean. The algorithm tracks volume over time; it does not stand at the VWAP price buying every dip to it.
Myth / hypothesis: “Institutions defend VWAP, so price reacts there, and it works because everyone watches it” Lore. A tidy story, but our lab has no test of a plain VWAP touch: not yet measured. The two related claims that were tested on NQ, band reversion and anchored-VWAP support/resistance, were not supported (boxes below).
Above or below: a description of control so far
Price holding above session VWAP means the average buyer since the start is ahead and the auction has been accepting higher prices Context. Below it, the reverse. That is a fact about the session so far, not a forecast. In the public NQ data, trend days do hold one side of VWAP after 10:30 ET, but “trend day” is a label given at the close, so that result describes trend days; it does not tell you this morning is one PlausibleOne trader's study. On balance days price crosses VWAP many times and the line is simply the middle of the rotation.
Bands: stretch, not a turning point
The bands turn “far from average” into a number. In the figure price spent most of the day between +1σ and +2σ: on a one-sided day the bands widen behind price, and a +2σ touch can repeat for hours.
Myth / hypothesis: “Price reverts to VWAP from the ±1σ and ±2σ bands” Lore. Tested on NQ: band touches reverted slightly less often than the baseline (box below). Use the bands to describe stretch and to size risk, not as a fade signal.
Anchored VWAP: who is ahead since the event
An anchor turns a vague idea (“FOMC resistance”) into one number with a precise meaning: the average price traded since that bar. If price is below an AVWAP anchored at a spike high, the average buyer since the spike is behind, by exactly the distance to the line Definition. The arithmetic does not tell you what those buyers will do next. The popular reading, “they defend their breakeven, so the line is support or resistance”, was tested for anchors at major swings and not supported Lore. Anchors at news releases have not been tested separately: not yet measured.
What the data says
Why it is still worth knowing
You will still use VWAP every day. It is the shared reference that other traders, algorithms and your own journal talk about. It describes the auction in one number: who is ahead since the open or since the news. It gives you a place where a read is plainly wrong, which is where a stop belongs. And “band touches revert” or “the anchor holds” are exactly the kind of claim you can test yourself with the drill below.
Worked example: the CPI anchor, long side
Synthetic session, not a real date. NQ, 1-minute chart, tick 0.25 points, $5 per tick per contract.
CPI prints at 08:30 ET (20:30 ICT, winter). NQ drops to 20,940.00 in the 08:31 bar, then recovers. You anchor VWAP at that bar. By 10:00 ET:
- AVWAP from the release: 20,978.50. Price: 20,991.25.
- The 24-hour VWAP from 18:00 ET: 20,962.75. It still averages in the overnight trade from before the news.
- RTH VWAP from 09:30 ET: 20,986.00.
What you know for certain: the average buyer since the release is 12.75 points ahead and the average seller since the release is 12.75 points behind. What you do not know: whether a dip to 20,978.50 will find buyers. That is the hypothesis, graded Lore, so you write a plan instead of resting a bid:
- If price pulls back to 20,978.50 and a 1-minute bar closes back above it with positive delta, buyers are stepping in at the anchor. Long 1 contract at 20,980.00.
- Stop 20,974.75, below the anchor and the pullback low: 5.25 points, 21 ticks, $105 per contract before costs.
- Target 2R, 10.50 points, at 20,990.50.
- Invalid if two 1-minute bars close below 20,978.50: the average post-news buyer is now behind, and you stay flat.
The anchor did not predict anything. It gave you a precise line to test the idea against and a cheap place to be wrong.
The mirror and the failure
Short side. FOMC day, synthetic. The statement is at 14:00 ET (02:00 ICT the next day, winter). NQ spikes to 21,240.00 in the 14:02 bar and fades. AVWAP anchored at the spike bar reads 21,198.50 at 15:00 ET with price at 21,186.25: the average buyer since the spike is 12.25 points behind. The mirrored plan: short only if a rally to 21,198.50 fails, with a 1-minute close back below it on negative delta; stop 21 ticks above the entry; invalid on two closes above the line.
The failure. Same day, another path: at 15:10 ET price rallies straight through 21,198.50 without a pause and the stop is hit, 21 ticks or $105 per contract plus costs. The “trapped buyers selling at breakeven” never showed up. On the tested swing anchors this is ordinary: the line held no more often than the baseline.
Prop-firm note. Both examples sit next to tier-1 releases. Some evaluation accounts restrict trading around releases such as CPI and FOMC, and the stop you place is the number the firm’s loss limits see. Check your firm’s current rules; the Prop Firm course covers news windows and loss limits.
Do it on the platform
Common mistakes
Common mistake: Trading VWAP crosses as signals: “price crossed above VWAP, buy”. On a balance day VWAP is the middle of the rotation and price crosses it again and again. Above or below describes who has been in control so far; it does not time an entry.
Common mistake: Treating a band touch as a reversal level. The public NQ test found band touches reverted slightly less often than the baseline. A +2σ touch on a one-sided day is often followed by another one.
Common mistake: Mixing up the lines. A 24-hour VWAP, an RTH VWAP and an anchored VWAP are three different averages. Quoting “VWAP” without the start time makes a plan impossible to check.
Drill: score your own band touches
Over ten sessions of NQ replay with RTH VWAP and ±2σ bands, mark the first ±2σ touch of each session. Before looking ahead, write your outcome rule: for example, “reverted” means price trades 20 points back toward VWAP within 60 minutes before it trades 20 points further away.
For a baseline, pick one random minute in the same session (roll a number from 1 to 390 before you look) and score it with the same rule, in the direction of VWAP.
- Log each event: time (ET), price, VWAP, σ and the outcome.
- Score band touches and random minutes with the same rule.
- Compare the two rates.
Solution
Expect the two rates to be close. With ten events per group the 95% interval on each rate is about ±30 percentage points, so ten sessions cannot separate them; the public study needed 2,350 touches to measure a −2.8 point difference.
The point of the drill is the habit: a written outcome rule and a baseline before you trust a line.
What we don’t know yet
- Whether a plain touch of session VWAP (not a band) leads to a reaction more often than a random price: not yet measured in our lab.
- Whether anchors at news releases (CPI, FOMC) behave differently from anchors at swings: not yet measured.
- One combination is being tested: a 2σ stretch plus delta divergence at a pre-marked level. In the public study it showed +5.6 points in development (n = 360) and a lead in the holdout (+15.3, interval +0.1 to +30.5). That is one trader’s unreplicated lead, so it stays Lore until it is confirmed on new data.
Recap
- VWAP is the volume-weighted average price since a start time; its σ bands measure how dispersed the day’s trading has been.
- Know the start: 18:00 ET (24-hour), 09:30 ET (RTH) or an anchor you chose. Each answers a different question.
- Above or below VWAP describes who has been ahead so far; it is context, not an entry signal.
- Band reversion and anchored-VWAP support/resistance were tested on NQ and not supported: use the lines to describe and to place stops, and write if-then plans around them.
Further reading
- Order Flow, Measured: All claims. A public, pre-registered practitioner study on NQ, 2018–2026. Read rows RL-01 (band reversion), RL-02 (trend days and VWAP, descriptive) and RL-06 (anchored VWAP).
- Volume-weighted average price: the definition, its use as an execution benchmark, and anchored VWAP.
- Berkowitz, Logue and Noser (1988), “The Total Cost of Transactions on the NYSE”, Journal of Finance 43(1): the paper that made VWAP the standard benchmark for judging execution.
- Optimal VWAP Execution Under Transient Price Impact: how an execution algorithm tracks VWAP, which is why it follows the volume curve rather than a price.
- Our lab: the VWAP band reversion concept note grades band fades Lore on the public NQ result, and treats “trend days hold one side of VWAP” as a description given at the close.
Next: M3.08 moves from averages to the order book itself. M4.04 applies VWAP in session prep, and M5.08 turns the pullback to VWAP into a written, scorable setup.
Educational content only. Not financial advice. Futures trading involves substantial risk of loss.
Saved in this browser (account sync comes at launch).