Abandoned Baby: doji, full gaps and data checks

Abandoned Baby isolates a central doji from two neighboring candles with full gaps, including shadows. Bulkowski’s published stock-data study reports 293 bullish examples across 4.7 million daily candle lines, with small-sample qualifications. That historical count is not a Forex occurrence rate or a win probability. Check complete OHLC, the price series and recognition timing before evaluating any trade.

What is the Abandoned Baby pattern?

The strict definition uses three candles: a trend-direction first candle, a central doji and an opposite-color third candle. CandleCharts requires the doji’s full High–Low range, including shadows, to be separated from both neighbors. This geometric rule neither proves trend exhaustion nor establishes a strongest-reversal ranking.

After a defined downtrend, the bullish variant has a bearish first candle, doji below it and bullish third candle above the doji. Using numbered highs H and lows L, require H2 < L1 and H2 < L3. The bearish variant after an uptrend is mirrored: L2 > H1 and L2 > H3. Touching boundaries fail this strict gap test. Opens alone cannot verify that later shadows preserve isolation.

Why the gaps matter — and how they look on forex

Check the actual venue and instrument schedule. Neither an exchange closure nor an important headline guarantees a price gap. Retail FX products can have weekend and daily breaks; OANDA Corporation, for example, publishes its own instrument-specific hours and reopening-gap warning. One weekend opening difference does not establish the two separate full gaps required here. Do not replace a gap with a large continuous candle move.

For the strict Abandoned Baby variant, the doji’s full range, including both shadows, is isolated from the first and third candle ranges. A relaxed variant may instead test gaps between real bodies. Name the variant and count its occurrences on the chosen feed and timeframe; a macro release or a sharp movement does not substitute for the required geometry.

Why does this sequence mean the trend is exhausted?

OHLC shows recorded price geometry, not who panicked, liquidated or supplied new capital. A doji has equal or nearby open and close; that does not prove balanced order flow, exhausted sellers or a durable reversal. Keep sentiment explanations as hypotheses requiring other evidence rather than consequences of a candle name.

Detector settings can also change classification. In a separate hypothetical TA-Lib C 0.6.4 control, use first O/H/L/C 1.0900/1.0910/1.0805/1.0810 and a doji 1.0790/1.0798/1.0785/1.0790. Third O/H/L remains 1.0820/1.0890/1.0815; vary its close. Prepend twelve identical bars 1.1000/1.1020/1.0990/1.1010 and keep other default settings. The first body is 90 pips. Penetration thresholds 30% and 50% are 1.0837 and 1.0855, with strict close-above tests. At close 1.0845, the bullish branch gives 100 for 30% and 0 for 50%; at 1.0860, both give 100. Codes are classifications, not probabilities. Both alternatives preserve full gaps. Record penetration and warm-up data; a halfway-close rule and the default 30% rule are different.

How to trade the pattern step by step

For the table, L1 − H2 = 0.0007, or 7 pips, and L3 − H2 = 0.0017, or 17 pips. The first body midpoint is 1.0855, below the third close 1.0880. Now change only L3 to 1.0795: its open still gaps above the doji, but a later shadow overlaps because 1.0795 < 1.0798. The strict pattern fails. RSI is a separate optional tested filter, not part of these geometric inequalities; see the RSI explanation.

The third candle’s final range is known only after it closes. Recognition based on that range cannot justify a retrospective fill at its earlier open. A rule using an unfinished candle is a different model and must allow later invalidation. State doji tolerance, prior-trend rule, confirmation, entry, exit, units and costs before evaluating outcomes. Ordinary stops can slip; no Fibonacci target or stop location follows automatically from the pattern name.

Abandoned Baby versus the morning star and other patterns

The Morning Star definition uses a small middle real body separated from the first body and a bullish third candle closing into the first. Its middle candle need not be doji; body separation is not full shadow-range isolation. Abandoned Baby’s stricter geometry does not prove stronger predictive performance. Bulkowski explicitly distinguishes breakout direction from later movement and notes sample limitations in his stock study.

Compare the engulfing pattern, which checks body coverage, with the harami pattern, which checks containment of a smaller real body. Neither requires the doji isolation used here. These different geometries are recognition rules; they do not by themselves rank the patterns’ predictive strength.

Most common identification mistakes

1. A non-doji middle candle fails Abandoned Baby, but does not automatically establish a Morning or Evening Star: their other conditions must also be checked.
2. Relaxing full-range gaps changes the definition on every market. Name the separate variant rather than calling non-overlapping shadows a softer condition.
3. Ignoring trend context. Inside a consolidation or after a weak move, the pattern carries no exhaustion signal.
4. Using the final third-candle range to authorize an earlier entry. A premature rule and a closed-candle rule are different tests; neither guarantees a future signal outcome.
5. Treating historical, equity-based statistics as a guarantee on forex. The sample is small, and the currency market has a different microstructure.

What to do tomorrow

1. Choose the price series, session boundaries, raw OHLC and doji/trend rules before a chronological scan. Record every strict candidate and rejection without revising gaps after seeing later prices. MT5 OTC charts use Bid; OANDA API smoothing can replace a bar’s first-price open with the previous close.
2. Document any RSI, volume or trend filter separately from the gap conditions. If comparing detectors, also record their version, penetration and warm-up settings. Do not treat recognition codes as win percentages or a scan as a profitability test.
3. If testing trades, specify when the signal becomes available, a subsequent executable entry, exits, size and all costs. Record ordinary-stop adverse fills and unfilled orders. A chart touch or hypothetical historical outcome does not establish future profitability.
4. Keep full-range and relaxed body-gap tests separate on the same data, with rules fixed beforehand. OANDA distinguishes Bid, Ask and midpoint OHLC; real intrabar spread variation requires separate series. A difference in classification is not proof of a better trading method.