Part 1: What a level is · Chapter 2 of 14 · 11 min
How to draw a level
A level is drawn with a price (the high or low of a swing) and a zone around it. We tried three timeframes and three thresholds: swing levels hold 48-49% (random levels 47-49%), and the width of the zone changes the number of contacts but not the outcome. There is always a level nearby: within 1 ATR of the price there are on average 2.8, and a 50 $ round number in 74% of candles.
Three ways of choosing levels
In the book levels come from three sources.
- Time. Highs, lows, closes and opens of a finished period: the previous day, the week, the month, Asia. They are prices anyone can compute at the same time (chapters 3 and 4).
- Round numbers. 4,000 $, 4,500 $, 4,550 $ (chapter 5).
- Swings. Highs and lows where the price really turned. It is the source most used in technical-analysis courses, and the one that needs the most care in drawing, because a swing is recognised only afterwards.
This chapter is about swings, and the drawing rules apply to the other sources too.
How to recognise a swing without looking at the future
A swing is a high (or low) higher (lower) than the neighbouring candles. The trouble is that on the day the price reaches that point you do not yet know whether it is a swing: it could be just a stage. A high is confirmed when the price, afterwards, falls far enough to show the high was there. The book's engine uses a fixed criterion: a swing exists after the price has moved away from the pivot by at least 1.5 ATR. The level is born at the moment of confirmation, not before, and is valid for about ten days (230 market-open H1 candles).
Drawing a level on a swing not yet confirmed, or drawing it afterwards, is the most common source of error (chapter 1): the line drawn afterwards already knows where the price turned.
How many swing levels are born, depending on timeframe and threshold? On market-open candles between December 2025 and October 2026:
| Timeframe | Move-away threshold | Levels | Contacts | Hold | Random levels |
|---|---|---|---|---|---|
| H1 | 1 ATR | 1,406 | 6,798 | 48% (± 1) | 48% |
| H1 | 1.5 ATR | 1,078 | 5,070 | 48% (± 1) | 49% |
| H1 | 3 ATR | 392 | 1,514 | 48% (± 3) | 48% |
| H4 | 1 ATR | 305 | 1,126 | 49% (± 3) | 48% |
| H4 | 1.5 ATR | 250 | 870 | 48% (± 3) | 48% |
| H4 | 3 ATR | 108 | 280 | 49% (± 6) | 47% |
| D1 | 1 ATR | 51 | 90 | 49% (± 10) | 47% |
| D1 | 1.5 ATR | 47 | 77 | 40% (± 11) | 48% |
| D1 | 3 ATR | 16 | 23 | 43% (± 20) | 52% |
The result for the first six rows is always the same: hold of 48-49% and random levels (shifted by 2-6 ATR, 20 trials) at 47-49%. Changing the timeframe or the threshold changes the number of levels (from 16 to 1,406), not how well they hold.
One row stands apart: D1 swings at 1.5 ATR hold 40% (± 11) over 77 contacts, below the band of random levels (43-53%). In a table of nine tests, one row outside the random band is normal; it is a question to retest, not a result.
Read the chart in words
H4 gold chart from 3 to 13 March 2026 with swing highs and lows marked and their level lines.
The zone, not the line
The precise line is a fiction. We counted the contacts with the previous day's high and low changing the width of the zone, from zero (the candle must reach the exact line) to ±1 ATR:
| Zone (per side) | Previous day, contacts | Hold | H4 swings, contacts | Hold |
|---|---|---|---|---|
| ±0 ATR (line only) | 310 | 46% (± 6) | 726 | 47% (± 4) |
| ±0.1 ATR | 326 | 48% (± 5) | 777 | 47% (± 4) |
| ±0.25 ATR | 347 | 48% (± 5) | 870 | 48% (± 3) |
| ±0.5 ATR | 436 | 52% (± 5) | 1,108 | 48% (± 3) |
| ±1 ATR | 428 | 47% (± 5) | 1,137 | 48% (± 3) |
Widening the zone from zero to half an ATR adds 40% more contacts on the previous day and 53% on swings, but the hold stays between 46% and 52%. The zone width decides how many contacts you count, and usually not how it ends. The book uses ±0.25 ATR (±4.7 $ as a median) because it is wide enough not to miss obvious contacts and narrow enough to remain a level.
A concrete case. With the ±0.25 ATR zone, of 347 contacts with the previous day's high and low, in 114 (33%) the candle does not reach the exact line: it stops a few dollars below (or above) and the price turns. With a thin line these would not be contacts, and a trader would call them "the level held" only if they had already decided to put the line there. By construction they hold more than the others (60% against 42%): the price is still on the starting side. It is an effect of the definition, not of the market: remembering it avoids comparing results obtained with different zones.
Read the chart in words
H1 gold chart of 5 and 6 April 2026 with the level's zone, the exact line and the contact point.
| When | What | Price |
|---|---|---|
| 5 Apr 2026 15:00 UTC | zone ±0.25 ATR | 4,674.13 – 4,683.01 |
| 5 Apr 2026 15:00 UTC | exact line 4678.57 | 4,678.57 |
| 6 Apr 2026 05:00 UTC | touch of the zone, not of the line | 4,676.09 |
Times in UTC.
The age of the level
A swing level is valid for about ten days: if the price returns after one or two days, does it hold more than if it returns after a week? On H4 swings at 1.5 ATR: contacts between 0 and 2 days from the birth of the level, hold of 45% (± 6, 249 contacts); between 2 and 5 days 51% (± 6, 261); between 5 and 10 days 47% (± 5, 360). No pattern: an old or recent level behaves the same way.
Too many levels
The practical problem of all level analysis is that there is always one nearby. On every candle we counted the "active" levels (previous day, week, month, Asia, H4 and D1 swings) within 1 ATR (18.6 $ as a median) of the price: on average 2.8, and on 91% of candles there is at least one. If you add the 50 $ round numbers, the share with no level nearby falls to 2%: round numbers alone are within 1 ATR on average 0.87 times per candle and there is at least one in 74% of candles.
With so many levels, every move has its own: after a turn you always find a level a few dollars away. That does not mean the level caused the turn; it means that, with 3 levels within 18 $ of the price, the turn is almost always near something. That is why the test must fix the list of levels and the contact rules beforehand: it is the only way to say that a level counts.
Read the chart in words
H4 gold chart from 3 to 10 March 2026 with four candles marked with the letters A, B, C and D.
| When | What | Price |
|---|---|---|
| 5 Mar 2026 04:00 UTC | Point A | 5,182.95 |
| 5 Mar 2026 16:00 UTC | Point B | 5,102.94 |
| 6 Mar 2026 12:00 UTC | Point C | 5,167.24 |
| 7 Mar 2026 08:00 UTC | Point D | 5,172.32 |
Times in UTC.
What to do with it
- Write the list of levels first. Types, timeframe, duration: all before looking at a contact. With too many possible levels, choosing afterwards always finds the right one.
- Draw swings only after confirmation. A pivot is a level when the price has moved away by at least 1.5 ATR (or the threshold you choose), not before.
- Use a zone of ±0.25 ATR. With the ATR of your timeframe: ±4.7 $ on H1 (median), wider on H4. Do not change it to make the numbers work.
- Give it an expiry. Ten days for H1 and H4 swings is enough: after that the level does not behave differently.
- Do not think a higher timeframe holds better. In our data H1, H4 and D1 give the same result; D1 has too few contacts to say.
Procedure in five steps
- Compute the ATR(14) of the timeframe you use and set the zone at ±0.25 ATR.
- Identify the pivots (local highs and lows) without marking them as levels yet.
- When the price moves away from the pivot by at least 1.5 ATR, the pivot becomes a level, with the date of confirmation.
- Keep the level for ten days; then delete it.
- Write everything in a log and do not touch the rules until you have at least a few hundred contacts.
Exercises
Chapter exercises
1. On the H4 chart the letters A-D are on four candles around a swing low. Which letter is the low that will become a level (the pivot)?
Read the chart in words
| When | What | Price |
|---|---|---|
| 5 Mar 2026 04:00 UTC | Point A | 5,182.95 |
| 5 Mar 2026 16:00 UTC | Point B | 5,102.94 |
| 6 Mar 2026 12:00 UTC | Point C | 5,167.24 |
| 7 Mar 2026 08:00 UTC | Point D | 5,172.32 |
Times in UTC.
2. With a ±0.25 ATR zone, of 347 contacts with the previous day's high and low, in 114 the candle does not reach the exact line. What does it mean?
3. Within 1 ATR of the price there are on average 2.8 levels (previous day, week, month, Asia, swings) and a 50 $ round number in 74% of candles. Why does it matter to whoever draws levels?
4. D1 swings at 1.5 ATR hold 40% (± 11) over 77 contacts, while random levels give 48%. How do you treat 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.