Part 3: What happens to levels · Chapter 9 of 14 · 10 min
Trendlines and channels
Trendlines (lines through two consecutive swings) hold 47-48% (random levels 46-48%); the "third touch" holds less (42%), but random lines at first contact give 45% too. The channel's parallels hold 50% (random 47-50%). Deleting the line at the first break drops the hold to 41%, and that of random lines too (44%).
Diagonal levels
A horizontal level has a fixed price. A trendline is a level that changes price over time: a straight line through two points. It is the oldest tool of technical analysis: you join two rising lows (the support line of an uptrend) or two falling highs (the resistance line of a downtrend) and extend the line into the future. The channel is the same line with a parallel on the other side.
The book's rules, without looking at the future:
- The two points are two consecutive swings of the same type, confirmed (the price has moved away by at least 1.5 ATR from each). For a support, two lows with the second higher; for a resistance, two highs with the second lower.
- The line must not already be violated: no candle between the two points passes it by more than 0.25 ATR.
- The line is born when the second point is confirmed and is valid for 230 market-open H1 candles (about ten days).
- The line's price at each candle is that of the straight line: the zone is ±0.25 ATR around it, as for the other levels.
- Contact, hold and break are measured as in chapter 1.
The typical slope of lines on H4 (1.5 ATR) is 18.9 $ per day as a median (between 9.8 and 33.6 $ from the 25th to the 75th percentile).
How many are born and how well they hold
| Trendline | Lines | Contacts | Hold | Break | Random lines (touches, hold) | Lines never touched |
|---|---|---|---|---|---|---|
| H4 swings, 1 ATR | 140 (69 supports, 71 resistances) | 457 | 47% (± 5) | 52% | 403, 47% (42-51) | 17% |
| H4 swings, 1.5 ATR | 113 (56, 57) | 365 | 47% (± 5) | 52% | 331, 48% (43-53) | 14% |
| H4 swings, 3 ATR | 41 (20, 21) | 138 | 46% (± 8) | 54% | 108, 46% (39-56) | 22% |
| H1 swings, 1.5 ATR | 512 (256, 256) | 1,588 | 48% (± 2) | 51% | 1,134, 48% (46-50) | 13% |
Random lines are the same lines shifted in parallel by 2-6 ATR (the band in brackets is the 5th-95th percentile of 30 trials). The result is the same as in the rest of the book: the hold of trendlines is equal to that of lines shifted at random, between 46% and 48%.
By direction: supports (180 contacts, arrival from above) hold 44% (± 7), resistances (185, arrival from below) 51% (± 7). Inside the error.
The "third touch"
A common rule says the trendline is valid from the third touch: two points define the line, the third confirms it. In our data the first contact after the two points, that is the third touch, is the weakest: on H4 (1.5 ATR) it holds 41% (± 10, 97 contacts), the second contact 51% (± 11, 82), from the third on 49% (± 7, 186). On H1: 42% (± 5, 444 contacts), 50% (± 5, 333), 50% (± 3, 811).
It looks like an effect, but it is not all the line's. Random lines, shifted in parallel, give 45% at the first contact (against 42% for real lines), 51% at the second, 49% from the third. The first time the price returns near a newly born line, real or fake, it holds less than 50%: the price is still moving away from the confirmation point. The "third touch" holds a little less than a random line (42% against 45%, inside the error), not more.
Read the chart in words
H4 gold chart from 1 to 12 April 2026 with a support trendline, the two anchor points and the contacts marked as hold or break.
| When | What | Price |
|---|---|---|
| 2 Apr 2026 04:00 UTC | trendline | 4,557.60 → 4,687.39 |
| 2 Apr 2026 04:00 UTC | 1 | 4,557.60 |
| 6 Apr 2026 00:00 UTC | 2 | 4,601.50 |
| 7 Apr 2026 12:00 UTC | 3: hold | 4,612.13 |
| 12 Apr 2026 20:00 UTC | 4: break | 4,647.18 |
| 12 Apr 2026 20:00 UTC | 5: break | 4,676.98 |
Times in UTC.
Read the chart in words
H4 gold chart from 16 to 25 February 2026 with a steep support trendline, the two anchor points and the contacts marked as hold or break.
| When | What | Price |
|---|---|---|
| 17 Feb 2026 12:00 UTC | trendline | 4,845.33 → 5,237.61 |
| 17 Feb 2026 12:00 UTC | 1 | 4,845.33 |
| 23 Feb 2026 08:00 UTC | 2 | 5,125.53 |
| 24 Feb 2026 00:00 UTC | 3: break | 5,148.46 |
| 25 Feb 2026 04:00 UTC | 4: hold | 5,211.46 |
Times in UTC.
The channel
The channel adds a parallel, through the opposite extreme between the two points: for a support, the parallel above, at a distance equal to the largest gap of the highs from the line. Measured as a level:
| Channel (parallel) | Lines | Contacts | Hold | Random lines (touches, hold) |
|---|---|---|---|---|
| H4 swings, 1.5 ATR | 113 | 298 | 50% (± 6) | 239, 50% (44-56) |
| H1 swings, 1.5 ATR | 512 | 1,329 | 50% (± 3) | 1,040, 47% (46-50) |
The parallel holds 50%, like random lines. The channel gives no more information than the trendline: two lines, twice the chance.
The deletion artifact
Whoever draws lines by hand deletes them when they break. It makes sense for someone who trades, but it dirties the measurement. If we delete the line at the first break (a close at 1 ATR beyond), the hold of trendlines falls: on H4 to 41% (± 8, 146 contacts), on H1 to 41% (± 4, 667). It looks like proof that trendlines hold little. But every deleted line ends, by construction, with a break. We do the same to random lines: on H4 they give 43% (36-52) and on H1 44% (41-48). The difference between 41% and 44% is a few points and inside the error: the rest is the effect of deletion.
It is a common error of method: a rule that removes lines after the outcome worsens the sample. To judge trendlines you must keep the same constraints for both groups (real lines and comparison lines).
Read the chart in words
H4 gold chart from 18 to 26 May 2026 with a support trendline and four candles marked with the letters A, B, C and D.
| When | What | Price |
|---|---|---|
| 20 May 2026 00:00 UTC | trendline | 4,457.27 → 4,615.89 |
| 20 May 2026 00:00 UTC | Point A | 4,457.27 |
| 21 May 2026 12:00 UTC | Point B | 4,492.09 |
| 22 May 2026 04:00 UTC | Point C | 4,510.71 |
| 23 May 2026 04:00 UTC | Point D | 4,506.60 |
Times in UTC.
What to do with it
- Draw the trendline only with the two confirmed points. The line is born when the second point is confirmed (1.5 ATR of moving away), not before.
- Do not treat the "third touch" as confirmation. It holds 42%, like a random line at first contact.
- Do not delete the line at the break in measurements. If you want to count its hold, give the same treatment to comparison lines.
- Do not add the channel to believe more. The parallel holds 50%, like chance.
- Use the trendline as a slope reference. A line of 19 $ a day is a measure of the trend, not a barrier.
Procedure in five steps
- Identify two consecutive swings of the same type, confirmed (at least 1.5 ATR of moving away).
- Check that no candle between the two points passes the line by more than 0.25 ATR.
- Draw the line from the second point, with a ±0.25 ATR zone, validity of ten days.
- Measure contacts, holds and breaks, without deleting the line at the first outcome.
- Redo it on lines shifted in parallel and compare.
Exercises
Chapter exercises
1. On the H4 chart the dashed line is a support trendline and the letters A-D are on four candles of 18-26 May 2026 (UTC hours). Which letter is the second point, the one that confirms the line?
Read the chart in words
| When | What | Price |
|---|---|---|
| 20 May 2026 00:00 UTC | trendline | 4,457.27 → 4,615.89 |
| 20 May 2026 00:00 UTC | Point A | 4,457.27 |
| 21 May 2026 12:00 UTC | Point B | 4,492.09 |
| 22 May 2026 04:00 UTC | Point C | 4,510.71 |
| 23 May 2026 04:00 UTC | Point D | 4,506.60 |
Times in UTC.
2. The "third touch" of a trendline holds 42%, and random lines shifted in parallel give 45% at first contact. What does it mean?
3. If you delete the trendline at the first break, the hold of real lines falls to 41%; that of random lines treated the same way to 44%. Why?
4. The channel's parallel holds 50% (± 6) on H4 and random lines 50%. How do you read it?
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.