Three Stars in the South is classified as a potential bullish reversal pattern even though its three candles are bearish. Potential matters: smaller bodies or the absence of a new Low do not prove exhausted selling. Recognition depends on exact body, shadow and range rules, which we will test on fictional prices.

What is Three Stars in South?

thinkorswim documents a long first bearish candle with a lower shadow, a smaller second and a small third, together with configurable body, trend and shadow parameters. A platform’s definition explains recognition; it does not establish a universal frequency, win rate or proof that buyers returned.

TA-Lib’s conditions differ in important details: L2 may equal L1, while L3 > L2 and H3 < H2 are strict. It permits C1 < O2 ≤ H1, not necessarily an Open inside body 1. The third body is compared with historical short-body settings, not directly required to be smaller than body 2. The function has no preceding-trend test. Under default settings the first lower shadow must exceed its own body.

A custom detector can specify additional filters, but record them separately from a platform flag. The following exercise uses explicit size limits and a declared previous-price context, not a complete reproduction of thinkorswim or TA-Lib.

Pattern structure — three conditions to meet

Our example requires three strictly bearish bodies. Body 1 must be at least twice the fixed mean of the ten earlier bodies; with an assigned mean 50 pips, its minimum is 100. Lower shadow 1 must exceed body 1. Body 2 must be smaller than body 1, with its lower shadow at least a quarter of body 2. Also require C1 < O2 ≤ H1 and L1 ≤ L2 < C1. The complete range of candle 3 must be strictly inside candle 2, body 3 no more than half body 2, and each third shadow at most 2 pips.

These are chosen exercise thresholds, not optimized strategy parameters. Assign the three preceding closes 1.1000, 1.0970, 1.0940 as a simple falling-context rule, and one pip=0.0001 for this fictional EUR/USD model. Lower shadows, body sizes, range containment and context are distinct tests; missing a Low alone does not express what the market wants.

Fictional EUR/USD: complete OHLC and component lengths
CandleOpen / HighLow / CloseBody / rangeLower / upper shadow
11.0900 / 1.09001.0680 / 1.0800100 /220 pips120 /0 pips
21.0850 / 1.08501.0740 / 1.078070 /110 pips40 /0 pips
31.0800 / 1.08011.0779 / 1.078020 /22 pips1 /1 pip

Body 1 meets 100 ≥ 2 × 50 and its lower shadow 120 >100. Body 2 is 70, below 100, and its lower shadow 40 exceeds 0.25 × 70=17.5. Body 3 is 20 ≤ 35, both shadows 1 ≤ 2, L3 >L2 and H3 <H2. The third candle is not an ideal shadowless marubozu; it passes our explicit small-shadow tolerance.

The lows rise 1.0680, 1.0740, 1.0779, but closes 1.0800, 1.0780, 1.0780 do not. The bodies sum to 190 pips; O1-to-C3 falls 120 pips. Two interbar upward Open changes of 50 and 20 pips account for the difference. This is geometry, not an observed reversal or entry recommendation.

Test boundaries rather than inferred psychology

Change one value at a time. L3=1.0740 makes its lower shadow 40 pips and equals L2, failing both the shadow limit and strict containment. H3=1.0850 creates an upper shadow 50 pips and equals H2, failing those two conditions. O3=C3=1.0780 makes a doji and fails our strictly bearish-body condition. None of these tests requires a story about participants or future prices.

Conversely, replacing only L2 with 1.0680 produces a100-pip lower shadow while retaining L2=L1, which our version allows. The third range remains inside candle 2. A version requiring L2 >L1 would reject it. Record the boundary convention before examining outcomes; an absent classification in one short sample does not establish rarity across the global FX market.

Three Stars in South compared to other reversal patterns

Bullish Engulfing and Morning Star have different recognition conditions. Their relative frequency and trade results require a defined sample, detector and execution model. Three bearish candles alone likewise do not identify Three Black Crows: its body, opening, closing and shadow relationships differ.

The Abandoned Baby is another separate construction. Do not infer that gaps are possible only after weekends or use that assumption to rank how often patterns must occur. Three Stars’ specified body and range tests can be evaluated without assigning a trader identity, exhausted order flow or a subsequent price direction.

Inside candle 2 does not mean inside candle 1

The first body is 10 with lower shadow 12. The second body 4, lower shadow 6 and Low 85 satisfy 78 ≤85 <90; Open 95 lies in the required part of candle 1. Candle 3's body 1 and shadows 0.1 are small; its full range[103.9;105.1] lies inside candle 2[85;110], but entirely above candle 1's High 100.

Containment does not automatically extend backwardThree fictitious falling candles share scale 75–112. Candle 2's range 85–110 exceeds candle 1's High 100. Candle 3's range 103.9–105.1 is inside candle 2, but entirely above candle 1.One scale:75–112H2 =110H1 =1001233 inside 2; entirely above 1
Containment refers to the specified parent. Candle 2's Open condition does not constrain its whole upper wick.

Adding “candle 3 also inside candle 1” rejects this model. That creates a new filter version, not an arithmetic correction. Specify the reference candle for each body, full-range or Open relationship in a scanner. These relationships are not interchangeable.

Assumptions: fictitious hypothetical-unit model: candle 1 O100,H100,L78,C90; candle 2 O95,H110,L85,C91; candle 3 O105,H105.1,L103.9,C104. First-body threshold 8 with mean 4; third-body limit half the second; third shadows at most 0.2. We test the article's explicit geometry variant.

Most common identification mistakes

  1. Replacing exact body/shadow/range inequalities with the label three red candles.
  2. Requiring a smallest third body in every detector: that is an extra filter in our custom version.
  3. Confusing a permitted equal second Low with the strictly higher third Low.
  4. Using a future break above H3 as though it were already part of the three-candle recognition signal.

An RSI threshold or EMA condition defines an added filter, not an essential component or independent proof of reliability. Fix its parameters and evaluate the changed sample separately.

Three Outside Up vs Three Stars in South — what connects them

The Three Outside Up pattern uses a bullish engulfing core and a later closing comparison. Three Stars uses three bearish bodies with different shadow and range tests. This is a construction difference, not a ranking of clarity, speed or profitability.

There is also a source-detector counterexample to steadily shrinking bodies. Place twelve earlier bars(100,112, 92,106) before O/H/L/C(100,100, 78, 90),(95, 95, 89, 93),(94.5, 94.6, 91.4, 91.5). The bodies are 10, 2, 3: the third is larger than the second. Its historical short-body threshold is(8 × 6+10+2)/10=6, so 3 <6; its shadows are 0.1 and its range is strictly inside the second. The default source predicates give+100, while our half-second-body limit 3 ≤ 1 fails. This hypothetical code-condition check is not a fresh market study or new compiled-detector run.

What to do tomorrow

  1. Choose a detector version and record body/shadow thresholds, range parents, equality and prior context.
  2. Check complete OHLC and rejected boundary cases. Keep body sums distinct from the net move across the trio.
  3. Save every candidate in the selected sample, including those with no later reversal or no executable order. A few selected historical examples cannot establish frequency or an edge.
  4. Define entry, exit, volume and costs separately. A break above H3, a Fibonacci target or an indicator filter is a new strategy rule to evaluate, not a guaranteed improvement.