Must a belt hold candlestick open exactly at the high or low?

The strict shape has no opening-side shadow: low for bullish, high for bearish. A tolerant scanner may allow a small difference, but its limit must be explicit. This example caps that difference at both one tick and 2% of the range.

Is every marubozu a belt hold pattern?

No. A marubozu identifies candle geometry. This belt hold setup also requires sufficient relative body length and a preceding move opposite the candle direction. A full marubozu can meet both definitions, but the name alone does not establish that context.

Does belt hold require a gap from the previous candle?

Not in the specification on this page. Definitions differ: the thinkorswim BeltHold documentation adds opening-gap conditions. Record whether your scanner requires a gap instead of assuming that the same pattern name means identical rules.

Does a bullish belt hold confirm a reversal?

No. It creates a candidate under the stated rules. The optional next-bar closing test supplies a separate confirmation, and even a confirmed event can later fail. These examples establish neither a success rate nor profitability.

TradingView candlestick patterns

Belt hold candlesticks: open-at-extreme rules and scanner tests

A belt hold candlestick is a long-bodied candle that opens at one extreme of its range after a move in the opposite direction. The bullish version opens at the low after a decline; the bearish version opens at the high after a rise. A small shadow at the closing end is allowed. The opening-side shadow and prior trend distinguish this reversal setup from simply labeling a large candle.

See the decision guide

Translate the belt hold shape into measurable rules

CandleScanner describes a long opening marubozu with a short closing-side shadow, after a decline for bullish belt hold or a rise for bearish belt hold. It explicitly treats its shadow cutoff as an implementation choice. The following custom specification uses ordinary OHLC candles; it is not a TradingView default or a reproduction of another scanner.

O, H, L and C are open, high, low and close. R = H - L and B = abs(C - O). The 20-bar mean uses only completed bars before the candidate, excluding the candidate itself.
Strict mode sets tolerance to zero. Near-extreme mode sets it to the smaller of one minimum tick and 2% of R. Both limits must pass; a large range must not permit a large opening wick.
Reject missing prices, H <= L, an open or close outside the high-low range, insufficient prior history, or a zero prior mean body. Do not divide by a zero range or silently replace missing history.
The 20%, 75%, 1.5 multiplier and three-close trend filter are illustrative research settings, not optimized thresholds. With a 2% opening cap and 20% closing cap, the body is already at least 78%; the 75% body test becomes relevant if the shadow settings are widened.

Test	Bullish candidate	Bearish candidate
Direction	C > O	C < O
Opening-side shadow	O - L <= tolerance	H - O <= tolerance
Closing-side shadow	(H - C) / R <= 20%	(C - L) / R <= 20%
Body share	B / R >= 75%	B / R >= 75%
Relative body length	B >= 1.5 × prior 20-bar mean body	Same requirement
Prior trend	Three preceding closes strictly decreasing	Three preceding closes strictly increasing

Recalculate accepted and rejected candles

These independent hypothetical cases use arbitrary price units, a 0.01 minimum tick and a prior 20-bar mean body of 0.50. The minimum long body is therefore 0.75. Assume sufficient valid history and the required prior trend unless stated otherwise. Tuples are (O, H, L, C).

For the first row, preceding closes of 102, 101 and 100.50, in chronological order, pass the decline filter. Reversing their order passes the rise filter instead.
A 0.01 opening wick on a 0.40 range is 2.5% and fails near-extreme mode even though it is only one tick. Tick and percentage tolerances answer different questions.
A body exactly 0.75 passes the length threshold, and a closing shadow exactly 20% passes its limit. Equal preceding closes fail the strictly directional trend test. Preserve those equality decisions in any implementation.

Case	OHLC	Decision
Strict bullish	(100, 104, 100, 103.60)	Pass: body 3.60 / 4 = 90%; opening shadow 0; closing shadow 10%
Strict bearish	(104, 104, 100, 100.40)	Pass: body 90%; opening shadow 0; closing shadow 10%
One-tick opening shadow	(100.01, 101, 100, 100.81)	Near-extreme pass: opening 1%; body 80%; closing 19%; body 0.80 >= 0.75. Strict mode fails
Two-tick opening shadow	(100.02, 101, 100, 100.82)	Fail: opening 0.02 exceeds the one-tick cap, despite meeting the 2% cap
Closing shadow too large	(100, 104, 100, 102.80)	Fail: closing shadow 30% and body 70%
Small absolute body	(100, 100.10, 100, 100.09)	Fail: 90% body share, but body 0.09 is below 0.75
Correct shape, wrong context	(100, 104, 100, 103.60), after rising closes	Bullish geometry passes; bullish reversal setup fails the prior-decline test
No range	(100, 100, 100, 100)	Reject before computing ratios

Belt hold versus full and opening marubozu

TradingView defines a full marubozu as having no shadow at either end. Belt hold emphasizes the opening end and its reversal context. An opening marubozu can therefore supply the candle geometry for a belt hold, while a full marubozu can also qualify if the length and prior-trend rules pass.

A large green candle after rising prices can be an opening marubozu without satisfying this bullish belt hold reversal setup.
Some implementations add a gap relative to the preceding candle. This page does not require that gap. Compare complete rule sets before treating different scanner labels as errors.

Feature	Belt hold setup	Full marubozu shape
Opening end	At the relevant extreme, or within a declared tolerance	At the relevant extreme
Closing end	Small opposite shadow permitted	No opposite shadow in the strict definition
Prior direction	Bullish after decline; bearish after rise	Geometry alone imposes no prior-trend test
Example overlap	Bullish (100, 104, 100, 104) after the required decline passes	The same candle is also a full bullish marubozu
Example difference	Bullish (100, 104, 100, 103.60) can pass	Its 0.40 upper shadow fails strict full marubozu

Test confirmation without moving the decision backward

For an optional confirmation test, save the candidate high and low when its bar closes. Inspect only the immediately following completed bar. A bullish confirmation requires a close above the saved high without a low below the saved low; a bearish confirmation requires a close below the saved low without a high above the saved high. A breach of the opposite extreme takes precedence if both conditions appear in the same bar.

For the bearish candidate with high 104 and low 100, a next-bar close of 99.80 confirms only if that bar never trades above 104. A later close above 104 is a failure under this review rule.
If neither confirmation nor invalidation occurs on the next bar, expire the candidate. Extending the window after observing later prices would change the experiment.
Confirmation uses later information and may miss reversals that start more slowly. Do not place a confirmed marker or hypothetical entry at the candidate open as though the following close were already known.
These are event labels, not executable orders. A strategy also needs entry timing, exit rules, position sizing, fees and slippage. A later close through a level does not prove a stop could fill at that level.

After the bullish 100-to-104 candidate	Outcome
Next bar H = 104.50, L = 101, C = 104.20	Confirmed at that next close
Next bar H = 104.50, L = 101, C = 104	Expired: equality is not a close above 104
Next bar H = 104.50, L = 99.90, C = 104.20	Invalidated: the low breach overrides the closing breakout
After valid confirmation, a later close is 99.80	Later failure of the reversal hypothesis; retain the earlier confirmation

Review a scanner against prices and screenshots

Validate a detector on one exchange-qualified symbol and timeframe before extending its scope. Save the session, adjustment settings, candle type, tick size and dates. Use ordinary candles: Heikin-Ashi prices are synthetic, so matching their geometry answers a different question.

Evaluate final geometry only after the bar closes. A live candle can develop an opening-side wick or lose enough body to fail. Closing-bar evaluation does not resolve every possible data or repainting issue.
Check raw shape matches, trend-qualified candidates, confirmations and later failures separately. Include rejected near-matches and zero-match samples; a count is not a win rate.
Compare the exact OHLC values in the Data Window with the rules. A compressed screenshot cannot establish whether a wick is zero or one tick long.
Use AI Chart Analysis to review the visible trend and candle context in a clear screenshot. Keep the numeric test alongside that review; image interpretation cannot certify the custom tolerance or recover hidden history.
Give Pineify the direction rules, both shadow caps, body tests, prior-close filter and confirmation precedence to draft a detector. Request distinct candidate and confirmation markers. Review the generated source, request syntax validation, and run the accepted and rejected cases before testing any trading hypothesis.

Where Pineify fits

Pine Script AI Coding Agent

Draft a detector from the explicit OHLC tolerances and confirmation rules. Inspect the generated conditions and compare the test cases on your chart before using alerts.

Also useful: AI Chart Analysis. Upload a chart screenshot for a review of visible trend and candle context. Verify exact opening-wick tolerances against OHLC prices yourself.

Draft belt hold detection rules

Related resources

TradingView candlestick patterns Marubozu candlestick rules Bar-close alert timing

Frequently asked questions

Educational information, not investment advice. Prices and thresholds are hypothetical research examples. No historical performance or predictive accuracy is claimed. Candlestick patterns can fail and trading can lose money.

Sources and verification

Last verified: 2026-10-17

Published by Pineify. Pineify is an independent product and is not affiliated with TradingView. Sources support the technical explanations; examples do not establish trading performance.