Part 3: Three-candle formations · Chapter 9 of 14 · 9 min
Three soldiers, three crows and continuations
Three soldiers, three crows and three methods are rare: 38 and 15 times on H1 gold, as often as in chance. Continuation formations go in the expected direction more often than reversal ones (soldiers and crows 54%), but with 22 and 13 trades the error is ±0.34-0.52 R. Three consecutive candles of the same colour: +0.22 ± 0.23 R over 57 trades, above 97% of the random series.
The rules
- Three white soldiers: three consecutive bullish candles, each with a body of at least 0.5 ATR and an upper wick of at most 50% of the body, that close higher and higher, each opened inside the body of the previous one (with a tolerance of 0.1 ATR). Expected direction: up.
- Three black crows: mirror image. Expected direction: down.
- Three methods (rising/falling three methods): a long candle of at least 1 ATR, three small candles (body at most half) inside its range, and a fifth candle in the same direction that closes beyond the first. Expected direction: the run continues.
Stop beyond the extreme of the formation plus 0.25 ATR (risk 0.5-4 ATR), target equal to the risk, 24 candles, cost 0.30 $. With three large candles the stop is far away: many trades are skipped because the risk exceeds 4 ATR.
How often they appear
| Formation | Found (chance, 5th-95th) | In expected direction | Trades (skipped) | Mean R | Random (median) |
|---|---|---|---|---|---|
| Three soldiers | 28 (27, 20-35) | 63% | 17 (11) | +0.27 ± 0.38 | +0.03 |
| Three crows | 10 (13, 8-18) | 30% | 5 (5) | −0.11 ± 0.77 | −0.10 |
| Soldiers and crows | 38 (40, 30-48) | 54% | 22 (16) | +0.18 ± 0.34 | +0.02 |
| Falling three methods | 11 (16, 11-22) | 45% | 9 (2) | +0.39 ± 0.60 | −0.08 |
| Three methods (with the 4 rising) | 15 (29, 22-35) | 33% | 13 (2) | +0.26 ± 0.52 | −0.03 |
The control is 37% up, 36% down, 27% no threshold. The formations are rare (about 4 a month for soldiers and crows, fewer than 2 for the three methods) and sit in the range of chance. 36% of candles have a body of at least 0.5 ATR and 77% of these have the wick on the closing side at most 50% of the body: the wick rule is not severe, the rarity comes from the demand for three large candles in a row. On H4 the same rules find 6 soldiers, 3 crows and 6 three methods; on D1 none.
The expected direction comes in 54% of cases for soldiers and crows (±16 points) against 37% for the control and 39% for the random series. The 22 trades give +0.18 ± 0.34 R (9 targets, 5 stops, 8 time exits), better than 84% of the random series (+0.02); for the three methods 13 trades at +0.26 ± 0.52 R, better than 93% (−0.03). With these errors it cannot be told from chance. First half of the period for soldiers and crows −0.01 R, second +0.26 R.
A generalisation found afterwards
Soldiers and crows are rare because they demand three large candles in a row. The book therefore tries a wider version, three consecutive candles of the same colour with a body of at least 0.5 ATR (a run of exactly three), with the same entry and stop rules (on the extreme of the three) and continuation as the expected direction. It appears 94 times (102 in chance, from 91 to 114): 50 bullish and 44 bearish. The run continues in 43% of cases (±10 points) and reverses in 46%; the 57 valid trades give +0.22 ± 0.23 R (27 targets, 16 stops, 14 time exits, median risk 3.0 ATR). The random series give +0.02 R (from −0.20 to +0.20): only 3% do as well. First half of the period +0.17 R (30 trades), second +0.27 R (27 trades): the two halves agree. Four consecutive candles are 16 (6 trades, +0.77 ± 0.30 R): an anecdote.
It is the highest result of the chapter and has the same sign in both halves, but it has three flaws to declare. The book tried it after seeing that soldiers were rare: it is a definition chosen after the data. It is one among about 35 formations and variants tried, so some result above the 95th percentile is expected. And the error (±0.23 R) is almost as large as the value. It has to be retested on new data with the rule as it is, before trusting it.
Three examples
Figure 1 shows winning three soldiers: on 1 July 2026 at 09:00, 10:00 and 11:00 (3,976.76 → 3,987.01 $; 3,987.00 → 4,000.57 $; 4,000.56 → 4,031.95 $), after a fall of 2.9 ATR. Entry at 4,031.95 $, stop at 3,971.50 $ (60.45 $, 3.9 ATR), target at 4,092.40 $ reached at 13:00: +1.00 R. Figure 2 shows three soldiers that fail: on 2 June at 02:00, 03:00 and 04:00 (4,476.56 → 4,488.38 $; 4,488.38 → 4,501.67 $; 4,501.67 → 4,519.54 $); entry at 4,519.54 $, stop at 4,471.25 $ (48.29 $, 3.4 ATR), target at 4,567.83 $; stop hit at 23:00: −1.01 R.
Figure 3 shows falling three methods: on 5 April 2026 a long candle at 12:00 (4,666.68 → 4,659.47 $), three small ones at 13:00-15:00 inside its range (high 4,665.70 $) and a fifth at 16:00 (4,664.01 → 4,656.27 $) that closes below the first; entry at 4,656.27 $, stop at 4,667.60 $ (11.33 $, 3.1 ATR), target at 4,644.94 $ reached at 20:00: +0.97 R.
Read the chart in words
H1 gold chart of 1 July 2026 with the three numbered candles and the lines of entry, stop and target.
| When | What | Price |
|---|---|---|
| 1 Jul 2026 09:00 UTC | Point 1 | 3,975.39 |
| 1 Jul 2026 10:00 UTC | Point 2 | 3,983.78 |
| 1 Jul 2026 11:00 UTC | Point 3 | 4,000.30 |
| 1 Jul 2026 11:00 UTC | Entry | 4,031.95 |
| 1 Jul 2026 11:00 UTC | Stop | 3,971.50 |
| 1 Jul 2026 11:00 UTC | Target | 4,092.40 |
| 1 Jul 2026 13:00 UTC | target reached | 4,109.75 |
Times in UTC.
Read the chart in words
H1 gold chart of 1-3 June 2026 with the three numbered candles and the lines of entry, stop and target.
| When | What | Price |
|---|---|---|
| 2 Jun 2026 02:00 UTC | Point 1 | 4,474.84 |
| 2 Jun 2026 03:00 UTC | Point 2 | 4,488.25 |
| 2 Jun 2026 04:00 UTC | Point 3 | 4,500.67 |
| 2 Jun 2026 04:00 UTC | Entry | 4,519.54 |
| 2 Jun 2026 04:00 UTC | Stop | 4,471.25 |
| 2 Jun 2026 04:00 UTC | Target | 4,567.83 |
| 2 Jun 2026 23:00 UTC | stop hit | 4,463.59 |
Times in UTC.
Read the chart in words
H1 gold chart of 5-6 April 2026 with the five numbered candles and the lines of entry, stop and target.
| When | What | Price |
|---|---|---|
| 5 Apr 2026 12:00 UTC | Point 1 | 4,666.68 |
| 5 Apr 2026 13:00 UTC | Point 2 | 4,663.21 |
| 5 Apr 2026 14:00 UTC | Point 3 | 4,664.24 |
| 5 Apr 2026 15:00 UTC | Point 4 | 4,665.70 |
| 5 Apr 2026 16:00 UTC | Point 5 | 4,664.01 |
| 5 Apr 2026 16:00 UTC | Entry | 4,656.27 |
| 5 Apr 2026 16:00 UTC | Stop | 4,667.60 |
| 5 Apr 2026 16:00 UTC | Target | 4,644.94 |
| 5 Apr 2026 20:00 UTC | target reached | 4,625.11 |
Times in UTC.
Procedure in four steps
- Write the rules before looking, and mark which definitions you changed after seeing the data.
- Count how often they appear: three large candles in a row are rare.
- Report the wide versions too, but as candidates to retest.
- Retest the rule, as it is, on new data or going forward.
Exercises
Chapter exercises
1. On the chart the letters A-D are on four candles of 1 July 2026. Which letter is the third candle of the three soldiers?
Read the chart in words
| When | What | Price |
|---|---|---|
| 1 Jul 2026 06:00 UTC | Point A | 3,980.08 |
| 1 Jul 2026 09:00 UTC | Point B | 3,988.19 |
| 1 Jul 2026 10:00 UTC | Point C | 4,000.80 |
| 1 Jul 2026 11:00 UTC | Point D | 4,037.10 |
Times in UTC.
2. Three bullish candles have bodies of 8, 9 and 12 (ATR 10, hypothetical numbers), each closes higher than the previous one and the upper wicks are 1, 2 and 3. Are they three white soldiers under the rule of the book?
3. The three consecutive candles of the same colour give +0.22 ± 0.23 R, above 97% of the random series, but the definition was chosen after seeing the data. How do you treat it?
4. Why are three soldiers and three crows so rare?
Have a question or want to share your exercise?
Post in the community, or join the free signals room on Telegram.
Educational content, not financial advice. Trading involves risk.