Part 4: Context, system and reference · Chapter 12 of 14 · 10 min

All the formations together

The 29 formations together give −0.06 ± 0.04 R over 2,346 trades, as much as the random series (−0.05). By number, 14 types out of 29 sit outside the range of chance (2.9 expected), because real candles are made differently from shuffled ones; by R no type with at least 10 trades sits above the 95th percentile.

The single table

The rules are those of the previous chapters: formations recognised in ATR at the close of the last candle, entry at that close, stop beyond the extreme plus 0.25 ATR (risk 0.5-4 ATR), target equal to the risk, 24 candles, cost 0.30 $. Data: H1 gold, December 2025 - October 2026. On the right the comparison with 100 random series obtained by shuffling blocks of 6 candles. The "expected direction" is that of the theory: for dojis and spinning tops the reversal, for inside bars the continuation.

FormationFound (chance, 5th-95th)In expected directionTradesMean R ± errorRandom (median)
Bullish marubozu101 (76, 70-82)41%92−0.17 ± 0.20−0.09
Bearish marubozu78 (49, 44-54)32%72−0.18 ± 0.22−0.01
Standard doji151 (158, 147-173)34%132−0.19 ± 0.17−0.10
Dragonfly29 (24, 20-29)38%28−0.19 ± 0.37−0.07
Gravestone26 (21, 17-24)54%26+0.21 ± 0.38+0.09
Spinning top94 (101, 90-114)37%93−0.08 ± 0.20−0.06
Hammer28 (25, 18-32)39%28−0.38 ± 0.35+0.08
Hanging man23 (26, 18-31)43%10−0.08 ± 0.66−0.04
Shooting star24 (25, 19-32)42%24+0.04 ± 0.40−0.02
Inverted hammer25 (21, 16-28)36%10−0.05 ± 0.66−0.27
Bullish engulfing108 (147, 138-159)41%103−0.06 ± 0.19−0.02
Bearish engulfing101 (144, 132-158)42%96+0.09 ± 0.190.00
Bullish harami92 (136, 126-145)39%91−0.15 ± 0.20−0.16
Bearish harami93 (152, 142-163)40%91−0.05 ± 0.20−0.09
Inside bar after a trend439 (470, 447-496)45%398−0.06 ± 0.10−0.05
Piercing57 (58, 51-67)25%55−0.17 ± 0.25−0.08
Dark cloud49 (55, 47-64)35%48−0.29 ± 0.27+0.05
Tweezer bottoms399 (268, 257-288)39%374+0.00 ± 0.10−0.07
Tweezer tops405 (295, 279-312)39%380−0.03 ± 0.10−0.01
Morning star31 (25, 19-31)45%27+0.12 ± 0.35−0.16
Evening star28 (25, 20-31)46%28−0.22 ± 0.36−0.09
Bullish three inside3 (6, 3-10)33%3−0.42 ± 1.38−0.01
Bearish three inside4 (9, 5-13)50%4+0.06 ± 1.07−0.11
Bullish three outside51 (67, 59-76)49%48−0.09 ± 0.27−0.01
Bearish three outside55 (72, 64-82)43%50−0.12 ± 0.25−0.04
Three soldiers28 (27, 20-35)63%17+0.27 ± 0.38+0.03
Three crows10 (13, 8-18)30%5−0.11 ± 0.77−0.10
Rising three methods4 (12, 7-17)0%4−0.04 ± 1.10+0.01
Falling three methods11 (16, 11-22)45%9+0.39 ± 0.60−0.08

How many types leave the range of chance

By number of formations: 14 types out of 29 sit outside the 5th-95th interval of the random series (2.9 are expected). Five have more: the two marubozu, the gravestone and the tweezer bottoms and tops. Nine have fewer: the two engulfings, the two haramis, the inside bar, the bearish three inside, the two three outside and the rising three methods. The total, instead, is inside the range of chance (2,547 against 2,526).

The difference has a recognisable cause. Formations that demand a size contrast with the previous candle (a body that covers another, a small one inside a large one, a range inside a range) are less frequent in the real data; those that demand an extreme candle (the marubozu) or two equal highs are more frequent. The reason is that volatility clusters: the correlation between the bodies of two consecutive candles (in logarithm) is 0.43 on gold, 0.35 in the series shuffled in blocks of 6 candles and 0.41 with blocks of 24. Nearby candles have more similar sizes than a shuffle suggests: the number of formations speaks of how the candles are made, not of what price does afterwards.

By R: with at least 10 trades the types are 24. None sits above the 95th percentile of the random series; two sit below the 5th (hammer and dark cloud, already seen): about 1.2 per side are expected.

By stability: splitting the period in half, among the 21 types with at least 5 trades per half only three have a positive R in both (gravestone, tweezer bottoms and morning star) and for the tweezers the value is zero; eleven have a negative R in both (the two marubozu, doji, dragonfly, hammer, inside bar, piercing, dark cloud, tweezer tops and the two three outside).

All together

The 29 formations give 2,346 trades (1,074 targets, 1,136 stops, 136 time exits), mean R −0.06 ± 0.04: the random series give −0.05 (from −0.08 to −0.02). With the confirmation of the next candle the trades are 713 and the R is −0.07 ± 0.07 (random −0.07, from −0.13 to +0.01). On H4 (same rules) the formations are 562, the trades 508 and the R −0.01 ± 0.09; on D1 106 formations, 97 trades, −0.06 ± 0.20.

Formations are frequent: 8.6 a day. 71% of candles close no formation.

One candle, several formations

Of 2,040 candles that close at least one formation, 438 (21%) close two or more of different types, and in 182 of these cases the expected directions are opposite. In the random series the share is 23% (from 22% to 23%): it is a property of the definitions, not of the market. A bearish harami after a rise is also an inside bar, and the theory reads the first as a reversal and the second as a continuation.

Figure 1 shows 9 January 2026 at 16:00 (the 15:00 candle from 4,490.40 to 4,507.53 $, the 16:00 one from 4,507.53 to 4,503.29 $, high 4,509.98 $, low 4,496.49 $): a bearish harami (direction down), an inside bar after a rise (up) and tweezer tops (down). At 17:00 the harami and the tweezers reach the target (+0.97 R) and the inside bar the stop (−1.02 R). Figure 2 shows 10 May at 14:00 (candles from 4,727.22 to 4,730.47 $ and from 4,730.47 to 4,730.22 $): a dragonfly (up), a hanging man (down, skipped for risk), a bearish harami (down) and an inside bar (up): the dragonfly and the inside bar win (+0.92 and +0.93 R), the harami loses (−1.12 R).

Figure 1. The same pair of candles closes three formations (9 January 2026, 15:00-16:00): bearish harami (down), inside bar (up) and tweezer tops (down); at 17:00 two reach the target and one takes the stop.Data: gold, XAUUSDT perpetual future (real candles) · 1 hour (H1) · from 9 Jan 2026 07:00 UTC to 10 Jan 2026 06:00 UTC.
Read the chart in words

H1 gold chart from 9 to 10 January 2026 with the two numbered candles and the list of the three formations with the expected direction.

WhenWhatPrice
9 Jan 2026 15:00 UTCPoint 14,487.82
9 Jan 2026 16:00 UTCPoint 24,496.49

Times in UTC.

Figure 2. Four formations on the same pair (10 May 2026, 13:00-14:00): dragonfly (up), hanging man (down), bearish harami (down), inside bar (up); two win, one loses, one is skipped for risk.Data: gold, XAUUSDT perpetual future (real candles) · 1 hour (H1) · from 10 May 2026 05:00 UTC to 11 May 2026 04:00 UTC.
Read the chart in words

H1 gold chart of 10 May 2026 with the two numbered candles and the list of the four formations with the expected direction.

WhenWhatPrice
10 May 2026 13:00 UTCPoint 14,726.94
10 May 2026 14:00 UTCPoint 24,727.33

Times in UTC.

Procedure in four steps

  1. Report all the formations, not only those that worked.
  2. Compare the numbers with more than one shuffle and explain the differences with the shape of the candles.
  3. Look at the R only where there are at least 10 trades and check the two halves of the period.
  4. Count the formations the same candle closes and their directions.

Exercises

Chapter exercises

1. On the chart the letters A-D are on four candles of 9 January 2026. On which letter do a bearish harami, an inside bar and tweezer tops close together?

Chart exercise. Data: gold, XAUUSDT perpetual future (real candles) · 1 hour (H1) · from 9 Jan 2026 06:00 UTC to 10 Jan 2026 01:00 UTC.
Read the chart in words
WhenWhatPrice
9 Jan 2026 12:00 UTCPoint A4,477.00
9 Jan 2026 15:00 UTCPoint B4,510.54
9 Jan 2026 16:00 UTCPoint C4,509.98
9 Jan 2026 19:00 UTCPoint D4,501.46

Times in UTC.

2. Across 29 formation types and a threshold of 10% (5% per side), how many are expected outside the range of chance by chance alone?

3. Engulfings, haramis and inside bars are less frequent in the real data than in the shuffled series. Why?

4. 21% of the candles with a formation close two or more, and in the random series 23%. What does it mean?

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.