Anchored VWAP: Choose the Starting Point Without Hindsight

A volume-weighted average price (VWAP) gives more influence to prices associated with more reported trading volume. An anchored version starts that calculation at a user-selected bar or timestamp and accumulates through the current observation. TradingView describes its drawing tool as VWAP beginning at a point the user selects on the chart ([anchored VWAP guide]({source:tradingViewAnchoredVwapSupport})). The line therefore answers a bounded question: what is the average of the chosen price inputs, weighted by the included volume, since this specified start? It does not reveal every holder’s cost basis or predict what price will do next.

For observations i from anchor a through time t, write AVWAP(t) = Σ(Pᵢ × Vᵢ) ÷ ΣVᵢ. Pᵢ is the price input and Vᵢ is the associated volume. If the data are individual trades, Pᵢ can be the execution price and Vᵢ the executed quantity. With bars, a platform may use a representative price such as the bar’s typical price, then weight it by bar volume. Those are not identical calculations. The [Pine Script VWAP reference]({source:tradingViewPineVwapReference}) exposes source and anchor/reset inputs, illustrating why you should record the implementation rather than assume every chart uses the same series.

Possible anchors include a regular-session open, an earnings or macro release, a gap, a known swing point, or the first trading bar after a major event. Each defines a different sample. Prefer a timestamp or event whose meaning can be stated independently of the later outcome: “the first regular-session bar after the scheduled release,” for example, is easier to reproduce than “the low from which price rallied.” A swing high or low can still be a useful descriptive reference, but the pivot is not confirmed at the instant it forms. Specify the event, instrument, venue/session, timezone, and whether the anchor bar is included.

A chart can place a line at an earlier low after later candles reveal that the low mattered. A trader viewing the finished chart sees a clean anchor; a historical simulation that acts from the low may be using future confirmation. For a live rule, define a confirmation condition and the first eligible decision time. If a pivot requires two later bars to confirm, the test cannot act at the pivot timestamp unless the strategy could know it then. For a scheduled release, state whether the anchor is the release timestamp, the first eligible trade, or the next regular-session bar; each choice handles gaps and off-hours volume differently.

Assume three successive bar-price inputs and volumes: 100.00 with 2,000 units, 102.00 with 1,000, and 101.00 with 3,000. The weighted numerator is 200,000 + 102,000 + 303,000 = 605,000 price-units; total volume is 6,000, so AVWAP is 605,000 ÷ 6,000 = 100.8333. The plain average of the three inputs is 101.00, a different answer because the 101.00 bar carries the most volume. If the anchor is moved to the second bar, the calculation becomes (102,000 + 303,000) ÷ 4,000 = 101.25. Both calculations are correct for their stated windows; the 0.4167 difference comes from changing the sample, not from a new market fact. The inputs are invented and the price-unit arithmetic is illustrative, not observed data.

If the current quote is above the chosen AVWAP, it is above that sample’s volume-weighted average input price. That may help compare price with a defined event window, but it does not mean all buyers since the event are profitable: participants traded at many prices, may have opened and closed positions, and may not belong to the data feed. Nor does a touch or cross establish support, resistance, institutional buying, or a reversal. A line can rise while price falls if the accumulated high-volume observations are expensive; it can flatten when fresh volume is small. A chart label compresses the sample, it does not explain the cause of subsequent price action.

If you try many anchor dates, event definitions, bar sizes, price inputs, and exit rules, then keep the best-looking combination, the result has been selected from a search. Bailey and coauthors discuss the probability of overfitting investment backtests ([paper]({source:baileyProbabilityBacktestOverfitting})); White’s Reality Check addresses data snooping across a family of rules ([paper]({source:whiteRealityCheck})). These methods do not validate an anchored-VWAP strategy automatically. Write the anchor and entry/exit rule first, include every tried variant in the search record, use chronological evaluation, and compare net results after spread, fees, slippage, and incomplete fills. A line that appears to work on a handpicked chart is a hypothesis, not evidence of a repeatable edge.

Record symbol and contract, exchange or feed, session calendar, timezone, bar interval, price source, volume unit, adjustment policy, anchor rule, and confirmation delay. Equity charts may differ on premarket prints and split-adjusted history; futures trade across sessions and roll between contracts; spot FX volume can be provider-specific; crypto venues can report different activity. A data correction, missing bar, or session reset can change the cumulative denominator and move the curve. Use the same convention in the chart, alert, and backtest, and inspect a second data source if the result depends on a small set of prints. The VWAP versus TWAP guide separates chart references from execution benchmarks; order flow imbalance explains a different order-book measure.