Modified hikkake: the stricter two-inside-bar reversal trap

The modified hikkake is a tighter, rarer relative of the Hikkake pattern: trading the false breakout of an inside bar. Both are built on a false breakout from an inside bar. The modified version demands more before it fires: two inside bars nested one inside the other, a first inside bar that closes at the edge of its range, and then a break of the second inside bar in the direction of that close. The extra conditions are meant to catch a coiled market that lures traders one way just before turning.

The two patterns also differ in intent. The original hikkake can act as either a continuation or a reversal. TA-Lib's own notes describe the modified hikkake as a reversal pattern, to be taken seriously only against an existing trend: a bullish modified hikkake in a downtrend, a bearish one in an uptrend.

The added strictness changes the numbers dramatically. On the same 60 stocks and ETFs where the plain hikkake fired 3,452 times in two years, the modified version fired 36 times. This page explains the four-bar structure, the close-location rule, how TA-Lib's values reach Finabase's scanner and gauge, two bearish cases from 2025, and the small sample of results.

The four bars

Take the bearish version, which TA-Lib's notes place at the top of an advance:

1. Bar 1, the reference. Any bar with a reasonably wide range.
2. Bar 2, the first inside bar. A lower high and a higher low than bar 1, and a close near its own high, within 20% of the average range of the five bars before it. The market contracted, but the close leaned bullish.
3. Bar 3, the second inside bar. A lower high and a higher low than bar 2. The range tightens again.
4. Bar 4, the false push. A higher high and a higher low than bar 3. Traders read the bullish close of bar 2 plus a break higher as a breakout. TA-Lib marks bar 4 −100.
5. Confirmation, if it comes. A close below bar 3's low on any of the next three bars. TA-Lib marks that bar −200.

The bullish version flips it: bar 2 closes near its low, bar 4 makes a lower high and a lower low than bar 3, and a close above bar 3's high within three bars confirms.

The exact rules Finabase uses

Finabase reads TA-Lib's CDLHIKKAKEMOD on daily bars:

How it compares with the plain hikkake:

• Two contractions instead of one. Bar 3 must sit inside bar 2, which already sits inside bar 1. Two consecutive inside bars are themselves uncommon.
• A close-location filter. Bar 2's close must point the same way as the coming false break. The plain hikkake has no such test.
• The confirmation level moves. It is measured against the second inside bar (bar 3), not the first.
• Still no trend test. TA-Lib's comment says the pattern is significant only against a trend, but the function does not measure one. In our sample, 12 of the 26 bearish signals came after a ten-session rise; the rest did not have the textbook backdrop.
• The same overwrite rule. If a new setup forms on the bar that would confirm an older one, only the new setup is reported.

What the extra conditions are trying to capture

Picture the bearish case. The market has been rising. Bar 2 contracts, but closes at its high, so bulls look in charge. Bar 3 contracts further: the market is coiling, and traders who read the bullish close are waiting for an upside break. Bar 4 delivers one, and buy orders above the coil are filled.

If the move had real demand behind it, price would carry on. When it does not, and price instead closes below the second inside bar, the traders who bought the break are trapped above the market with stops underneath. The modified hikkake tries to isolate the cleanest version of that trap: a market compressed twice, with a misleading close pointing traders the wrong way, at the end of a move.

How Finabase handles the ±100 and ±200 values

Finabase passes the TA-Lib values straight through to the chart, but its other features use only the sign:

• Chart markers show both the setup bar and, when it happens, the confirmation bar.
• The candlestick scanner lists every positive value under bullish and every negative value under bearish, so setups and confirmations sit on the same list.
• The Signals-tab gauge (see AAPL's signals) gives each of the 61 patterns one vote based on its latest non-zero value in the past eight bars, Buy for positive and Sell for negative, and combines them as (buy − sell) / (buy + sell + min(neutral, max(buy, sell))). Unlike the plain hikkake, which holds a vote on most days, the modified hikkake had a value in that window on just 0.7% of stock-days in our sample, so it rarely affects the gauge at all.

Two bearish cases from 2025

Walmart, February 12, 2025. After a 6.3% rise over ten sessions, WMT's February 7 bar spanned $101.11 to $103.11. February 10 was inside it and closed at $102.92, a cent below its high. February 11 was inside again ($101.85 to $102.85). On February 12 the stock pushed to a higher high and higher low, $102.04 to $103.90, and TA-Lib printed −100. A confirmation needed a close below $101.85 within three sessions; the closes were $105.05, $104.04 and $103.78, so none came. Then, on February 20, the day Walmart reported quarterly results, the stock opened about 5% lower. Ten sessions after the signal it was down 6.6%. The direction was right, but the pattern's own trigger never fired, and the move came from news.

iShares Russell 2000 ETF, April 28, 2025. During the rebound from the April 2025 lows, IWM's April 23 bar ran from $189.84 to $195.51. April 24 was inside it and closed at $194.06, near its $194.37 high, and April 25 was inside that ($191.55 to $194.26). April 28 broke higher ($192.53 to $196.15) for a −100. The confirmation level was a close below $191.55. On April 30 IWM traded down to $190.72 intraday but closed at $194.86, so it never confirmed. Ten sessions later the ETF was 6.6% higher. Exxon Mobil and the Energy Select Sector SPDR (XLE) printed the same −100 on the same day, and both did confirm with −200 on April 30; Exxon was 4.9% lower five sessions after its setup, but back slightly above it after ten.

What the data says

CDLHIKKAKEMOD produced 36 values on 22 of the 60 large US stocks and ETFs between October 30, 2024 and October 29, 2026: 7 bullish setups, 3 bullish confirmations, 18 bearish setups and 8 bearish confirmations.

The bearish side looks strong over five sessions and then fades: the ten-session average is slightly positive. With 26 cases, a 65.4% hit rate is about 1.9 standard errors above the baseline, which is suggestive at best.

Where it sits among inside-bar patterns

An educational outline

A structure some traders use for this pattern, offered for study, not as a recommendation:

1. Insist on the trend. Only consider a bearish modified hikkake after a clear advance, and a bullish one after a clear decline.
2. Wait for the ±200. In both 2025 examples the setup alone would have been an ambiguous entry.
3. Use the false break as the stop. For a bearish signal, the high of the −100 bar is where the trap would be disproved.
4. Watch the clock. If three bars pass without confirmation, the setup has expired in TA-Lib's terms.

Frequently asked questions

What is the difference between the hikkake and the modified hikkake?

The modified version needs two nested inside bars instead of one, requires the first inside bar to close near the end of its range that points toward the coming false break, and measures confirmation against the second inside bar. It fires about one-hundredth as often.

Is the modified hikkake a reversal or a continuation pattern?

TA-Lib's documentation treats it as a reversal: bullish signals matter in a downtrend and bearish signals in an uptrend. The function itself does not check the trend.

What do −100 and −200 mean?

−100 marks the bar that broke higher out of the second inside bar (the setup). −200 marks a later bar, within three bars, that closed below the second inside bar's low (the confirmation).

Why did the bearish side outnumber the bullish side?

In our sample there were 26 bearish and 10 bullish values. With numbers this small, the split mainly reflects which stocks happened to coil and break higher during the volatile spring of 2025.

Summary

The modified hikkake narrows the hikkake to its most deliberate form: a market that contracts twice, closes one way, breaks that way, and then reverses. TA-Lib marks the setup with ±100 and the confirmation with ±200; Finabase's scanner and gauge read only the sign, and the pattern is too rare to sway the gauge often. Its 36 cases in our sample leaned in the bearish signals' favor over five sessions but faded by ten, and the two 2025 examples show both the appeal and the ambiguity. Respect the trend context, wait for the confirmation and keep the stop at the false break.