Belt hold candlestick: opening at the extreme, rules and data

A belt hold is a long candle whose open is the session's extreme. A bullish belt hold opens on its low and climbs for the rest of the day; a bearish belt hold opens on its high and falls. The closing end may have a shadow. The opening end has essentially none, which means that from the very first trade, the market moved only one way.

The Japanese name, yorikiri, is a sumo term for a winning move in which a wrestler grips the opponent's belt and drives him straight out of the ring. The image fits: no feint, no retreat, just steady pressure from the opening bell. Steve Nison presented the belt hold as a reversal candle when it appears after an extended move, and he added that, traditionally, it opens with a gap in the direction of the old trend before reversing against it.

Finabase's detector, TA-Lib's CDLBELTHOLD, keeps the shape but not the gap. That decision has measurable consequences, which this page walks through: the exact rule, how often it fires (about one bar in seven), two real examples that split the textbook in an unexpected way, and what 4,514 belt holds on 60 large US stocks and ETFs did next.

Bullish and bearish belt holds

Bullish belt hold. The market has been falling. The session opens lower again, often with a gap below the prior close, and that opening print turns out to be the day's low. Buyers lift the price for the rest of the session and close well above the open, possibly a little below the high.

Bearish belt hold. The market has been rising. The session opens higher, and that open is the day's high. Sellers push it down all day, and the close is far below the open.

In both cases the first trade is the extreme. That is the feature that distinguishes the belt hold from other long candles: it shows the reversal began immediately, not after a morning of hesitation.

The exact rules Finabase uses

Finabase uses TA-Lib's CDLBELTHOLD on daily bars:

The output is +100 for a green belt hold and −100 for a red one.

What the code leaves out

Three textbook ingredients are missing, and each one widens the net:

1. The gap. Nison's version opens with a gap in the old trend's direction: a bullish belt hold opens well below the prior close, a bearish one well above. TA-Lib does not look at the previous candle at all. In our sample only 29% of bullish belt holds opened below the prior candle's body, and 35% of bearish ones above it.
2. The trend. 29.6% of bullish belt holds followed a 10-session decline and 39.1% of bearish ones a 10-session rise. Most were not at the end of an extended move.
3. Exceptional length. The textbook calls for a long candle relative to recent action. TA-Lib's threshold is the average body, which in our sample 41% of all daily candles exceeded.

The result is a detector that finds any above-average candle with a clean open, about one trading day in seven.

Belt hold, closing marubozu and marubozu

The belt hold is the mirror of the Closing marubozu: a candle that ends at its extreme, tested: the belt hold's clean end is the open, the closing marubozu's is the close. A candle that is clean at both ends is a full Marubozu candlestick: the shadowless candle, rules and real data, so all 1,227 marubozus in our sample were also belt holds. The other 3,287 belt holds had a visible shadow at the close. 63% of all belt holds were also tagged as a Long line candle: what TA-Lib’s CDLLONGLINE really measures.

The idea behind an open at the extreme

An open at the day's extreme says the opening auction was the most one-sided moment of the session and everything after it went the other way. For a bullish belt hold after a decline, the overnight sellers who pushed the open down found nobody to sell to below that level; buyers took every offer from the first minute. The longer the body, the more the price traveled in that single direction.

The traditional gap adds a second element: it means the open itself extended the trend before the reversal. That is what makes the textbook belt hold a reversal candle rather than just a strong day. A gap-down open that becomes the low traps the sellers who sold the gap.

The shadow at the close is the soft spot. A bullish belt hold that closes well below its high gave back part of the rally late in the day, which is why many practitioners prefer belt holds that close near the far extreme as well.

Confirmation

Typical follow-up checks:

• A close beyond the far extreme on the next session: above the high for a bullish belt hold, below the low for a bearish one. 867 of 2,351 bullish (37%) and 744 of 2,163 bearish (34%) belt holds in our sample confirmed this way.
• The opening extreme holds. If price trades back through the belt hold's open (its low for bullish, high for bearish), the move that started at the bell has been fully undone.
• Context and volume. A belt hold at a prior support or resistance level, on heavy volume, after a long trend, is the version Nison described.

Two real belt holds that broke the rules both ways

IBM, June 3, 2026. IBM had rallied 37.5% in ten sessions. On June 3 it opened at $318.30, which was also the day's high, and fell to close at $305.63, a $12.67 body against an $8.46 average. But it did not gap up: the open was $10.93 below the prior day's $329.23 close. A textbook reading would reject it; TA-Lib accepted it. The next session closed at $301.77, below the $302.53 low, confirming it, and IBM was 10.9% lower after five sessions and 14.2% lower after ten.

Adobe, December 20, 2024. ADBE had dropped 19.1% in ten sessions. On December 20 it opened at $433.02, below the prior close of $437.39, which is the textbook gap-down open. The low was $432.47, 55 cents under the open, and the stock rallied to close at $447.17. Everything Nison asks for was there: a decline, a lower open that became the low, a long white body. The next close, $446.74, did not exceed the $451.42 high, and Adobe drifted lower: −0.3% after five sessions, −5.5% after ten.

The gap-less belt hold worked and the textbook one failed. Two cases prove nothing, but the full sample points in the same direction, as the next section shows.

What 4,514 belt holds did next

We ran CDLBELTHOLD over daily bars from October 30, 2024 to October 29, 2026 for 60 large US stocks and ETFs.

Confirmed outcomes start from the confirmation close. Returns are raw price changes, so prices rose on average after bearish belt holds too.

Did the textbook gap help?

We split the same detections by whether the open gapped beyond the previous candle's body in the traditional direction:

(Counts are slightly lower than the main table because they include only candles with five full sessions of data after them.)

The gapped, textbook-shaped belt holds did worse than the rest on both sides, and worse than a random day. One plausible reading is that in a market of large, heavily traded stocks, a gap followed by a strong reversal is often a one-day event: the gap fills, and the prior trend continues. That is an interpretation, not a tested result.

Summing up:

• No edge from the pattern as Finabase detects it. Bullish belt holds were slightly below the baseline and bearish ones on it.
• Confirmation added a few points on the bullish side (54.2%) and nothing on the bearish side (44.9%).
• The textbook filter was not a fix in this sample.

Belt hold compared with neighboring patterns

An educational outline

Described for learning, not as a recommendation.

1. Add back what TA-Lib leaves out, if you use the textbook version: a real prior trend and a level that matters. Be aware our data did not reward the gap.
2. Watch the opening extreme. For a bullish belt hold, the open is the day's low. A later close below it cancels the idea.
3. Prefer a close near the far end. A long upper shadow on a bullish belt hold means the day's buyers already gave some ground.
4. Illustrative numbers: bullish belt hold opening at $25.00 (low) and closing at $26.80, high $27.00. Entry on a next-day close above $27.00, say $27.20, with invalidation under $25.00: risk about $2.25 a share, so a $450 budget is 200 shares.

How Finabase shows belt holds

• Chart markers. Add Belt Hold (the IBM link above opens it) and every flagged bar is marked. Expect roughly one marker every week or so on a typical large-cap chart, and expect them to share bars with Marubozu and Long Line Candle markers.
• Scanner. Green belt holds go to the Bullish list and red ones to the Bearish list, by the sign of TA-Lib's output. The scan does not check for a gap or a trend.
• Signals tab. The Candlestick gauge takes the most recent belt hold within eight bars as a Buy or Sell vote. In our sample it was voting Buy on 39.4% of days, Sell on 35.0%, and neutral on only 25.7%, so it rarely abstains.

Frequently asked questions

What is a bullish belt hold?

A long green candle that opens at its low, ideally after a decline. Traditionally it also opens with a gap below the prior close. TA-Lib only checks the long body and the clean open.

What does yorikiri mean?

It is a sumo term for forcing an opponent out of the ring while gripping his belt, the image behind the belt hold's name: steady, one-directional pressure from the start.

Is a belt hold the same as a marubozu?

No. A marubozu has no shadow at either end; a belt hold only needs a clean open. Every marubozu is a belt hold, but about 73% of belt holds in our sample were not marubozus.

Are belt holds reliable reversal signals?

Not in our test. Bullish belt holds were slightly worse than a random day over five sessions, bearish ones about equal, and the gapped textbook version did worse than the rest.

Summary

The belt hold is a long candle that opens at its extreme and moves away from it all session. TA-Lib's version tests only the body length and the opening-end shadow, leaves out the gap and the trend, and fires on about 15% of daily bars. In our 60-stock sample it showed no edge, and the gapped subset that best matches the textbook did worse. It is a useful word for a kind of day, and a weak forecast of the next one.