Anchored VWAP

What anchored VWAP is

VWAP stands for volume weighted average price. It answers one question: what was the average price paid so far, counting busy bars more than quiet ones? The plain version is explained in the VWAP indicator guide. It usually restarts at the start of each trading session.

Anchored VWAP changes one thing. You choose the bar where the count starts. Everything from that bar to now goes into the average, and everything before it is ignored. The line begins at the anchor and then moves with each new bar. The further it runs from the anchor, the more bars it holds and the slower it moves.

The idea behind it is simple. Some bars are events that people may care about: a swing low, a gap, a big news bar, the day a company reported. The anchored line shows the average price paid by everyone who traded since that event. It is a record of the cost of that group, nothing more.

The formula

For each bar, take a typical price. Many tools use the high plus the low plus the close, divided by 3. Some use the close alone. Multiply it by the bar volume. Then add up from the anchor bar:

Because the sums are running totals, a bot does not need to loop over all the bars each time. It keeps two numbers, the running price x volume and the running volume, and updates them on each bar. Guard the division: if the running volume is zero, there is no VWAP yet.

A worked example

These are made-up bars. Each has a typical price and a volume, shown in the table. We anchor at bar 5, the lowest price in the set. The table shows the first five bars from the anchor.

At bar 5 the VWAP is just the price, 47.50, because only one bar has traded. At bar 6 the running price x volume is 31125 and the running volume is 650, so VWAP = 31125 / 650 = 47.88. By bar 9 it is 49.02. By bar 12 it has settled at 49.35, after 8 bars. The last price is 50.50.

Volume weighting matters. A plain average of the same 8 prices is 49.63, while the VWAP is 49.35. The difference comes from the bars with large volume, which pull the VWAP toward their prices.

Which anchor to pick, and why it is a judgement

Traders use many anchor points. None of them is the right one in any measured sense. Each answers a different question.

• A swing low or swing high. The average price paid since the turn. It depends on which turn you call the swing, and that is only clear later.
• A gap or a news bar. The average since the event. The event is easy to see, but how much it mattered is a guess.
• The start of the week, month or year. A calendar anchor that needs no judgement about the chart. It also has no link to anything that happened in the market.
• A large-volume bar. You can find it with the volume indicator. A bar can be big for many reasons, and volume alone does not say who was buying.
• The start of a trade. For a bot that holds a position, the average price weighted by size since entry is a cost basis. This is closest to arithmetic and furthest from opinion.

The example shows the effect. With the same data, the end values are 49.16, 49.35 and 50.18. A person looking at the chart in hindsight can always find an anchor where price seemed to respect the line. With enough anchors to try, some line will look good. That is the same trap described in overfitting: the anchor is a setting, and a setting tuned on past data fits noise.

Limits for a bot

• The anchor must be chosen by a rule, not by eye. A bot cannot say "the obvious low". It needs a written rule, for example "the lowest low of the last 20 bars", and that rule is a setting to test. A swing low is only known after later bars confirm it, so a test that anchors at it on the bar it happened uses information the bot did not have. This is look-ahead bias.
• It slows down. After many bars the line barely moves, since one new bar is a small part of a large total. A rule that waits for price to cross a flat line behaves differently from one tested earlier, when the line was young.
• Volume differs by market. Some markets report real traded volume. Others report tick counts or the volume of one venue, which is only a part of the whole. The line is only as good as the volume behind it.
• A crossing is not an edge. Price spends time on both sides of any average. A bot that buys above and sells below pays the spread on each trade, and the line gives no reason the next move continues.
• Many anchors, one result. If a bot tries dozens of anchors and keeps the best, it has done a search, and the best result of a search is optimistic.
• Gaps and sessions. If data skips overnight or has thin bars, the sums include them. Check how your data source handles them.

How to check a rule that uses it

Write the anchor rule down before testing. Run it on data where the anchor can only use bars that had closed. Test in a range, an uptrend and a downtrend. Count the trades, the cost per trade and the largest losing run. Then keep a part of the data back and test once, as in backtesting. If the result needs one particular anchor to work, it is probably fitted.

Compared with a volume profile

Anchored VWAP gives one price, the average. A volume profile shows how much traded at each price level over a period. Both use volume, but they answer different questions: one gives an average, the other a shape. Neither says what happens next.

What it does not tell you

It does not say that price will return to the line, bounce from it or break it. It summarises prices already traded, from a start you chose. This page is education about a calculation, not a signal.