Using AI for Pine Script VWAP Indicators

To get a working VWAP indicator out of an AI assistant, your prompt has to name Pine Script v6, the exact built-ins, the anchor behavior, the inputs and a confirmed-bar rule. Without those constraints, an assistant may invent functions, mix language versions or produce signals that change before the realtime bar closes. Start with the broader Best AI for Pine Script guide if you need a general prompting and review workflow.

VWAP, volume profile, and footprint requests are especially error-prone because they combine cumulative calculations, intrabar volume, newer Pine features, and realtime data. This guide provides a reusable prompt, a working VWAP band example, the correct footprint API, and an audit process for turning reviewed output into alerts.

What This Indicator Category Covers

VWAP and deviation bands

VWAP is cumulative typical price multiplied by volume, divided by cumulative volume. Typical price is normally calculated as high plus low plus close, divided by three. Pine exposes that traditional input as hlc3.¹

TradingView's VWAP supports Session, Week, Month, Quarter, Year, Decade, Century, Earnings, Dividends, and Splits anchors. Standard deviation bands measure dispersion since the latest anchor and multiply that value by a selected factor.¹ An AI prompt must distinguish this anchored calculation from a rolling moving average.

Volume profile and POC

A volume profile aggregates activity by price level over a selected period. Its point of control, or POC, is the price level with the highest traded volume. The value area is the range containing a chosen share of total volume, typically 70%. Value area high and value area low identify its upper and lower boundaries.²

This differs from VWAP because the profile organizes volume across price rows rather than reducing the period to one volume-weighted average. For implementation details, see the POC and value area Pine Script guide.

Footprints and order flow

A footprint distributes seller and buyer volume across price levels inside each chart bar. TradingView classifies an intrabar as buy volume when its close exceeds its open and sell volume when its close is below its open. If they are equal, classification falls back to movement relative to the previous intrabar close.³

Imbalance detection compares the buy volume at one price level with the sell volume at the level below it, and the reverse for sell imbalances. The default threshold is 300%, meaning one side must be three times the opposing volume.³

Why AI Gets This Wrong

Invented functions and mixed versions

TradingView warns that AI can hallucinate Pine language features, borrow constructs from general-purpose languages, and combine syntax from different Pine versions. It also notes that code can run without doing what the request intended.⁴

Treat generated code as an unreviewed draft. A plausible name such as ta.vwap_band() is not evidence that a function exists. Check unfamiliar identifiers against the Pine Script v6 reference before debugging the surrounding logic.

Repainting from fluid data

Historical bars contain fixed OHLC values, but high, low, and close remain fluid while a realtime bar is open. Calculations using those values can produce intrabar results that cannot be reproduced after the bar becomes historical.⁵

For alerts, require barstate.isconfirmed or deliberately use values from the preceding bar. Either choice delays the signal, but it prevents a temporary cross or footprint condition from being treated as permanent.

Footprint is a newer feature

TradingView added request.footprint() and the footprint and volume_row types in January 2026.⁶ An assistant trained before that release may not recognize the API and may attempt to recreate it with invented functions or incomplete lower-timeframe logic.

A Prompt Template That Pins Specifics

Name the version and function

State //@version=6 explicitly. Then name the required built-in, such as ta.vwap(source, anchor, stdev_mult), instead of asking broadly for VWAP bands. This prevents the assistant from selecting a legacy pattern or fabricating a separate band function.

Specify inputs and anchors

List every permitted anchor value and identify hlc3 as the source. Also define the default band multiplier, plot behavior, and whether event-based anchors must use request.earnings(), request.dividends(), and request.splits().

Require confirmed-bar logic

Tell the assistant that alert conditions may only fire on confirmed bars. Ask it to identify every use of fluid high, low, close, requested data, or intrabar data that could change before confirmation. This review request is as important as the code request.

The reusable template

Write a Pine Script indicator with these requirements:

Version: Pine Script v6 using //@version=6
Indicator type: Overlay VWAP with standard deviation bands
Required function: ta.vwap(source, anchor, stdev_mult)
Source: hlc3
Anchor input options: Session, Week, Month, Quarter, Year, Decade, Century, Earnings, Dividends, Splits
Default anchor: Session
Band multiplier input: float, default 2.0
Alert: Trigger when close crosses either band
Confirmation: Alert conditions must require barstate.isconfirmed
Output: Plot VWAP, upper band, and lower band

After the code:
1. Explain which values can change on an open realtime bar.
2. List every built-in function used.
3. Identify any function not present in the Pine Script v6 reference.
4. Explain how each anchor resets ta.vwap().

Worked Example: Prompting VWAP Bands

The flawed first response

A vague request for “VWAP with standard deviation bands” may produce code calling ta.vwap_band(). That function does not exist. It is a representative example of the invented Pine features TradingView warns users to expect from AI output.

Correcting it with ta.vwap

ta.vwap(source, anchor, stdev_mult) returns a tuple containing VWAP, the upper band, and the lower band. The anchor is a Boolean series that resets the calculation when true. TradingView added the anchor and stdev_mult parameters to ta.vwap() in June 2022, so a separate band function is unnecessary.⁶

The corrected v6 code

//@version=6
indicator("Confirmed VWAP Bands", overlay = true)

string anchorPeriod = input.string("Session", "Anchor",
     options = ["Session", "Week", "Month", "Quarter", "Year", "Decade", "Century", "Earnings", "Dividends", "Splits"])

float bandMultiplier = input.float(2.0, "Band multiplier", minval = 0.1, step = 0.1)

bool earningsEvent = not na(request.earnings(syminfo.tickerid, earnings.actual, barmerge.gaps_on, barmerge.lookahead_off, ignore_invalid_symbol = true))
bool dividendEvent = not na(request.dividends(syminfo.tickerid, dividends.gross, barmerge.gaps_on, barmerge.lookahead_off, ignore_invalid_symbol = true))
bool splitEvent = not na(request.splits(syminfo.tickerid, splits.denominator, barmerge.gaps_on, barmerge.lookahead_off, ignore_invalid_symbol = true))

bool anchor = switch anchorPeriod
    "Session" => timeframe.change("D")
    "Week" => timeframe.change("W")
    "Month" => timeframe.change("M")
    "Quarter" => timeframe.change("3M")
    "Year" => timeframe.change("12M")
    "Decade" => timeframe.change("12M") and year % 10 == 0
    "Century" => timeframe.change("12M") and year % 100 == 0
    "Earnings" => earningsEvent
    "Dividends" => dividendEvent
    "Splits" => splitEvent
    => false

[vwapValue, upperBand, lowerBand] = ta.vwap(hlc3, anchor, bandMultiplier)

float confirmedVwap = barstate.isconfirmed ? vwapValue : vwapValue[1]
float confirmedUpper = barstate.isconfirmed ? upperBand : upperBand[1]
float confirmedLower = barstate.isconfirmed ? lowerBand : lowerBand[1]

vwapPlot = plot(confirmedVwap, "VWAP", color.blue, 2)
upperPlot = plot(confirmedUpper, "Upper band", color.orange)
lowerPlot = plot(confirmedLower, "Lower band", color.orange)
fill(upperPlot, lowerPlot, color.new(color.orange, 90))

bool crossAbove = ta.crossover(close, upperBand) and barstate.isconfirmed
bool crossBelow = ta.crossunder(close, lowerBand) and barstate.isconfirmed

alertcondition(crossAbove, "Cross above upper VWAP band", "Price crossed above the upper VWAP band.")
alertcondition(crossBelow, "Cross below lower VWAP band", "Price crossed below the lower VWAP band.")

Paste generated code into the Pine Editor and compile it before publishing it or attaching an alert. Compare its output with TradingView's built-in VWAP using the same source, anchor, and multiplier.

Mistakes Specific to Footprint Requests

Check the signature

The documented signature is request.footprint(ticks_per_row, va_percent, imbalance_percent), returning a series footprint.⁷ ticks_per_row has no default and must be a positive simple integer. The optional value area percentage defaults to 70, while the imbalance percentage defaults to 300.

One call and possible na

A script can make only one request.footprint() call. Calls nested inside request.security() or request.security_lower_tf() count toward that limit. The result is na when footprint data is unavailable, so check na() before calling methods on the object.⁷

Plan access matters

Footprint requests are available to TradingView Premium or Ultimate users.⁶ A script may therefore be structurally correct but unavailable to the person testing it on another plan.

Real footprint object methods

Audit generated footprint code against these documented method families:

• footprint.total_volume(), buy_volume(), sell_volume(), delta(), poc(), vah(), val(), and rows()
• volume_row.up_price(), down_price(), buy_volume(), sell_volume(), total_volume(), delta(), has_buy_imbalance(), and has_sell_imbalance()

Treat other footprint-related names as suspect until verified. The footprint request reference and volume footprint build guide provide more focused implementation examples.

A Review Checklist For This Category

Version and function existence

• Confirm the first line declares //@version=6.
• Look up every unfamiliar function in the v6 reference.
• Verify tuple assignments match the built-in return values.
• Check that anchor conditions become true on the intended bars.

Confirmation and repainting

Inspect every condition using high, low, close, requested data, or footprint rows. Determine whether it can fire while the current bar remains open. TradingView specifically recommends asking whether alerts wait for the end of the realtime bar.⁵

Compile and compare visually

Resolve every compiler and runtime error in the Pine Editor. Then compare the custom VWAP against the built-in indicator on the same symbol, timeframe, source, and anchor. A visual mismatch does not identify the exact bug, but it quickly exposes reset, source, or band calculation errors.

Turning the Indicator Into an Alert

Define the trigger

Choose one precise event: a confirmed price cross of a VWAP band, a footprint buy imbalance, or a break of a profile value area. TradingView footprint charts support new and stacked buy or sell imbalance alerts, plus conditions based on delta and total buy or sell volume.³

Send a webhook payload

A TradersPost webhook signal can include ticker, action, quantity, orderType, and optional stopLoss or takeProfit objects. After receiving the JSON, the platform processes it under the configured strategy and subscription settings, then submits the resulting trade to the connected broker or exchange. Paste a message like this into the alert's message field; TradingView fills in the placeholders when the alert fires.

{
  "ticker": "{{ticker}}",
  "action": "buy",
  "quantity": 1,
  "orderType": "market",
  "price": {{close}}
}

Use action deliberately. For example, buy and sell express directional signals, while resize targets an ending position quantity and can flatten the position when its quantity is zero.

Test before live deployment

First verify that the alert appears in TradingView's Alert Log. Then compare it with the destination's signal history. TradersPost's documentation warns that alert webhooks from third parties such as TradingView can be delayed or may not be sent at all, so automation still requires monitoring.

Avoid back-to-back entry and exit signals. TradersPost does not retry an order that a broker rejects or that fails during execution, is not designed for high-frequency trading, and recommends leaving at least 1 to 5 minutes between signals to avoid conflicting orders.

Bottom Line

• Prompt for Pine Script v6 and name the exact built-ins.
• Use ta.vwap(source, anchor, stdev_mult) for VWAP bands.
• Verify the footprint signature, plan requirement, call limit, and na handling.
• Require confirmed-bar alert logic and test every script in the Pine Editor.
• Compare generated output with TradingView's built-in studies before trusting it.

Use the broader Best AI for Pine Script guide for related workflows, including trend and moving average indicators, momentum indicators, volatility indicators and support, resistance and pivot indicators.

Once your indicator compiles, matches the expected chart behavior, and waits for confirmed bars, connect its alert to a TradersPost webhook. Test the complete signal and order workflow with paper trading before enabling live execution.

Frequently Asked Questions

How do VWAP bands work?

Use ta.vwap(source, anchor, stdev_mult). It returns [vwap, upper_band, lower_band]. There is no separate VWAP band function because the multiplier and tuple return are part of ta.vwap().

Which plan supports request.footprint()?

TradingView footprint requests require a Premium or Ultimate plan. The function and its related object types were introduced in January 2026.

Why does generated volume code repaint?

High, low, and close remain fluid until a realtime bar closes. Add barstate.isconfirmed to signal conditions or use a one-bar offset when delayed values are acceptable.

Is footprint the same as profile?

No. A footprint separates buy and sell volume by price level inside one bar. A volume profile aggregates volume by price level across a selected period and identifies levels such as POC, VAH, and VAL.

References

¹ Volume Weighted Average Price
² Volume Profile Indicators: Basic Concepts
³ Volume Footprint Charts: A Complete Guide
⁴ Will AI Help Me Write Pine Script Indicators?
⁵ Repainting
⁶ Pine Script Release Notes
⁷ Other Timeframes and Data