Part 2: One- and two-candle formations · Chapter 7 of 14 · 9 min

Piercing, dark cloud and tweezers

Piercing and dark cloud appear 106 times on H1 gold (113 in chance) and go in the expected direction in only 29% of cases: the 103 trades give −0.22 ± 0.18 R, worse than every random series. Tweezers are 804, more than the 563 of chance, but the 754 trades give −0.01 ± 0.07 R: in line with chance.

The rules

  • Piercing: the first candle is bearish with a body of at least 1 ATR; the second is bullish, opens at the close of the first (or up to 0.1 ATR above) and closes beyond the middle of the body of the first but below its open. Expected direction: up. In the 24/7 future there is no opening gap like the one the "textbook" piercing requires: the rule replaces it with an open at the close of the first.
  • Dark cloud cover: mirror image. Expected direction: down.
  • Tweezer bottom: the first candle is bearish, the second bullish, the lows differ by at most 0.1 ATR and both have a range of at least 0.5 ATR. Expected direction: up. Tweezer top: mirror image, expected direction down.

The stop sits beyond the extreme of the two candles plus 0.25 ATR (risk 0.5-4 ATR), the target equals the risk, 24 candles, cost 0.30 $.

How often they appear

FormationFound (chance, 5th-95th)In expected directionTradesMean RRandom (median)
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
Piercing + dark cloud106 (113, 102-127)29%103−0.22 ± 0.18−0.01
Tweezer bottoms399 (268, 257-288)39%3740.00 ± 0.10−0.07
Tweezer tops405 (295, 279-312)39%380−0.03 ± 0.10−0.01
All tweezers804 (563, 541-592)39%754−0.01 ± 0.07−0.04

The control is 37% up, 36% down, 27% no threshold.

Piercing and dark cloud go against the theory

Theory reads them as reversals. On gold price goes in the expected direction in only 29% of cases (±9 points) and in the opposite direction in 48%: after the big candle and the partial recovery, the movement of the first candle resumes. The 103 valid trades give −0.22 ± 0.18 R (37 targets, 57 stops, 9 time exits): the random series give a median −0.01 R (from −0.15 to +0.08) and none does worse. First half −0.27 R, second −0.17 R. With the confirmation of the next candle: 38 confirmations out of 106 and 36 trades at −0.40 ± 0.29 R (random −0.06). With an opposite trend before (31 cases): −0.19 ± 0.33 R. It is a negative result consistent in the two halves, but it comes from one formation among the 29 the book tests: it has to be retested on new data.

Tweezers

Tweezers appear 804 times against the 563 (541-592) of the random series (739 with blocks of 24 candles): more than chance. The reason is not clear. The future opens every candle at the close of the previous one, and when a candle closes near its high and the next turns right away, the two highs coincide: in the real data the alternation of colours is a little more frequent (52.2% against 51.7% in the shuffled series) and the one-step correlation of the bodies is −0.032 (−0.017 in chance, 5th percentile −0.029). It does not explain the excess on its own. The highs also fall near multiples of 5 $ in 12.1% of cases against 10.0% expected, and near multiples of 10 $ in 6.5% against 5.0%: round numbers attract a little.

Price goes in the expected direction in 39% of cases (±3 points), as much as chance (38%) and slightly above the control. The 754 valid trades give −0.01 ± 0.07 R (370 targets, 354 stops, 30 time exits): in line with chance (−0.04, 24% of the random series do at least as well). The tweezer bottoms (+0.00 ± 0.10 R) sit above 90% of the random series, but the error is larger than the edge. First half −0.01 R, second −0.02 R. After an opposite trend (159 cases): 35% in the expected direction, −0.04 ± 0.16 R. With the confirmation of the next candle: 251 confirmations and 246 trades at −0.08 ± 0.12 R (random −0.06). More sample, result zero.

Four examples

Figure 1 shows a winning piercing: on 5 February 2026 at 23:00 (4,778.54 → 4,702.29 $, low 4,700.00 $) and on 6 February at 00:00 (4,702.28 → 4,763.16 $), after a fall of 1.5 ATR. Entry at 4,763.16 $, stop at 4,635.01 $ (128.15 $, 2.1 ATR), target at 4,891.31 $ reached at 05:00: +1.00 R. Figure 2 shows a dark cloud that fails: on 3 May 2026 at 12:00 (4,620.48 → 4,623.82 $) and at 13:00 (4,623.82 → 4,622.10 $), after a rise of 4.1 ATR on a calm Sunday (ATR 1.25 $); entry at 4,622.10 $, stop at 4,625.27 $, target at 4,618.93 $; stop hit at 16:00: −1.09 R.

Figure 3 shows tweezer tops: on 23 December 2025 at 03:00 (4,496.30 → 4,498.16 $, high 4,500.67 $) and at 04:00 (4,498.16 → 4,490.97 $, high 4,500.00 $), after a rise of 4.2 ATR; entry at 4,490.97 $, stop at 4,504.52 $ (13.55 $, 0.9 ATR), target at 4,477.42 $ reached at 06:00: +0.98 R. Figure 4 shows tweezer bottoms: on 4 February at 17:00 (4,906.66 → 4,902.58 $, low 4,879.00 $) and at 18:00 (4,902.56 → 4,929.58 $, low 4,875.53 $), after a fall of 3.7 ATR; entry at 4,929.58 $, stop at 4,863.04 $, target at 4,996.12 $ reached at 23:00: +1.00 R.

Figure 1. A winning piercing: 5-6 February 2026, after a fall of 1.5 ATR; entry at 4,763.16, stop at 4,635.01, target at 4,891.31 reached at 05:00 (+1.00 R).Data: gold, XAUUSDT perpetual future (real candles) · 1 hour (H1) · from 5 Feb 2026 15:00 UTC to 6 Feb 2026 11:00 UTC.
Read the chart in words

H1 gold chart of 5-6 February 2026 with the two numbered candles and the lines of entry, stop and target.

WhenWhatPrice
5 Feb 2026 23:00 UTCPoint 14,700.00
6 Feb 2026 00:00 UTCPoint 24,650.00
6 Feb 2026 00:00 UTCEntry4,763.16
6 Feb 2026 00:00 UTCStop4,635.01
6 Feb 2026 00:00 UTCTarget4,891.31
6 Feb 2026 05:00 UTCtarget reached4,897.95

Times in UTC.

Figure 2. A dark cloud that fails: 3 May 2026, 12:00-13:00, after a rise of 4.1 ATR; entry at 4,622.10, stop at 4,625.27 hit at 16:00 (−1.09 R).Data: gold, XAUUSDT perpetual future (real candles) · 1 hour (H1) · from 3 May 2026 04:00 UTC to 3 May 2026 22:00 UTC.
Read the chart in words

H1 gold chart of 3 May 2026 with the two numbered candles and the lines of entry, stop and target.

WhenWhatPrice
3 May 2026 12:00 UTCPoint 14,624.96
3 May 2026 13:00 UTCPoint 24,623.83
3 May 2026 13:00 UTCEntry4,622.10
3 May 2026 13:00 UTCStop4,625.27
3 May 2026 13:00 UTCTarget4,618.93
3 May 2026 16:00 UTCstop hit4,628.87

Times in UTC.

Figure 3. Tweezer tops after a rise of 4.2 ATR (23 December 2025, 03:00-04:00): entry at 4,490.97, stop at 4,504.52, target at 4,477.42 reached at 06:00 (+0.98 R).Data: gold, XAUUSDT perpetual future (real candles) · 1 hour (H1) · from 22 Dec 2025 19:00 UTC to 23 Dec 2025 12:00 UTC.
Read the chart in words

H1 gold chart of 22-23 December 2025 with the two numbered candles and the lines of entry, stop and target.

WhenWhatPrice
23 Dec 2025 03:00 UTCPoint 14,500.67
23 Dec 2025 04:00 UTCPoint 24,500.00
23 Dec 2025 04:00 UTCEntry4,490.97
23 Dec 2025 04:00 UTCStop4,504.52
23 Dec 2025 04:00 UTCTarget4,477.42
23 Dec 2025 06:00 UTCtarget reached4,466.17

Times in UTC.

Figure 4. Tweezer bottoms after a fall of 3.7 ATR (4 February 2026, 17:00-18:00): entry at 4,929.58, stop at 4,863.04, target at 4,996.12 reached at 23:00 (+1.00 R).Data: gold, XAUUSDT perpetual future (real candles) · 1 hour (H1) · from 4 Feb 2026 09:00 UTC to 5 Feb 2026 05:00 UTC.
Read the chart in words

H1 gold chart of 4-5 February 2026 with the two numbered candles and the lines of entry, stop and target.

WhenWhatPrice
4 Feb 2026 17:00 UTCPoint 14,879.00
4 Feb 2026 18:00 UTCPoint 24,875.53
4 Feb 2026 18:00 UTCEntry4,929.58
4 Feb 2026 18:00 UTCStop4,863.04
4 Feb 2026 18:00 UTCTarget4,996.12
4 Feb 2026 23:00 UTCtarget reached5,027.80

Times in UTC.

Procedure in four steps

  1. Write the rules (bodies, opens, tolerance of the highs or lows) before looking for the formations.
  2. Count how often they appear and compare with more than one random shuffle.
  3. Report the results that go against the theory too.
  4. With high frequency the sample is useful: look at the error, not only the mean.

Exercises

Chapter exercises

1. On the chart the letters A-D are on four candles of 23 December 2025. Which letter is the second candle of the tweezer tops (bearish, with a high almost equal to that of the first)?

Chart exercise. Data: gold, XAUUSDT perpetual future (real candles) · 1 hour (H1) · from 22 Dec 2025 18:00 UTC to 23 Dec 2025 13:00 UTC.
Read the chart in words
WhenWhatPrice
23 Dec 2025 00:00 UTCPoint A4,488.00
23 Dec 2025 03:00 UTCPoint B4,500.67
23 Dec 2025 04:00 UTCPoint C4,500.00
23 Dec 2025 07:00 UTCPoint D4,490.00

Times in UTC.

2. Piercing and dark cloud go in the expected direction in 29% of cases and in the opposite direction in 48%. What does it mean?

3. Tweezers appear 804 times against 563 in the random series, but the trades give −0.01 ± 0.07 R. What does it mean?

4. A high falls within 0.25 $ of a multiple of 5 $ in 12.1% of cases, while without an effect 10.0% is expected. What can be said?

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.