Part 2: Where levels go · Chapter 3 of 14 · 11 min
The previous-day levels
Yesterday's high, low, close and open are the most used levels. On gold they hold the same (48% over 830 contacts, random levels 47-49%). The next day touches yesterday's high or low in 50% of cases and the close in 84%; the hold changes with the hour (57% in the American morning, 45% in the evening), but it changes almost the same way for random levels too: it is the clock, not the level.
Yesterday's four levels
Of a COMEX trading day (from 18:00 to 17:00 New York time) four prices are known at the close: the high, the low, the close and the open. At 18:00 New York time, when the next day starts, they are known to everyone and are valid for the whole new trading day. They are the most used levels because they are the easiest to compute and are the same for everyone (chapter 6 of the book on hours for the definition of trading day).
What they represent:
- High and low. The two extremes: the highest and lowest points where trading took place. Above the high, whoever was selling was wrong; below the low, whoever was buying. Stops and orders sit there.
- Close. The last price of the day, a reference for contracts and for whoever measures the next day's "gap".
- Open. The first price of the day, the point the market started from.
In our data (210 days, chapter 1 rules: zone ±0.25 ATR, hold and break at 1 ATR from the contact point, 24 candles):
| Yesterday's level | Levels | Contacts | Hold | Break | Hold of random levels | Move after 12 candles (ATR) |
|---|---|---|---|---|---|---|
| High | 210 | 166 | 48% (± 8) | 52% | 49% | +0.31 ± 0.42 |
| Low | 210 | 181 | 48% (± 7) | 50% | 49% | −0.24 ± 0.38 |
| Close | 210 | 297 | 48% (± 6) | 50% | 49% | +0.11 ± 0.32 |
| Open | 210 | 186 | 48% (± 7) | 51% | 47% | −0.14 ± 0.38 |
The four levels give the same result: 48% hold. Random levels shifted by 2-6 ATR give 47-49%. The last column measures, in ATR, where the price is 12 candles after the contact in the direction of the expected bounce: the numbers are close to zero, with an error (± 0.3-0.4) larger than the value. The level produces no measurable bounce.
How many are touched
Knowing what happens after the contact is of little use if there is no contact. The next day touches the zone of yesterday's high on 105 days out of 210 (50%), that of the low on 107 (51%), that of the open on 111 (53%) and that of the close on 176 (84%). The close is in the middle of the field and the price almost always passes through it; the two extremes are far away and half the time the price does not get there.
Another way of seeing the same thing. In a trading day, compared with yesterday's, the price breaks only the high on 38% of days, only the low on 37%, both (an outside day) on 11%, neither (an inside day) on 14%. On 86 days out of 100 at least one of yesterday's extremes is exceeded; on three days out of four only one, never both.
These numbers are more useful than the hold rate, because they are base rates for setting a target: if your target is yesterday's high, the price reaches it one time in two.
When: does the session matter?
An analysis worth doing is splitting the contacts by time of day: is the hold the same in Asia, London and New York? The right comparison, in each band, is with random levels that have a contact in the same band.
| Time of contact (New York) | Contacts | Hold | Break | Hold of random levels (same band) |
|---|---|---|---|---|
| Evening and Asia (18:00-02:00) | 432 | 45% (± 5) | 54% | 44% |
| London (02:00-08:00) | 137 | 45% (± 8) | 54% | 55% |
| American morning (08:00-12:00) | 175 | 57% (± 7) | 41% | 50% |
| American afternoon (12:00-17:00) | 86 | 52% (± 11) | 48% | 46% |
The hold changes with the hour, from 45% to 57%, but it also changes for random levels: in the evening and Asia an ordinary price "holds" 44%, in London 55%. The reason lies in the market and not in the level: in the evening, after a contact, the price tends to carry on (in the previous books: the quiet hours move in one direction without bounces), in London to turn. With the effect of the hour removed, real levels are 1 point above random ones in the evening, 10 below in London (45% against 55%, just outside the 48-65% band), 7 above in the American morning (57% against 50%, inside the 42-60% band) and 6 above in the afternoon. No clear difference.
It is a lesson in method: a comparison by time band without a control by time band leads to wrong conclusions. The 57% of the American morning looked like a result; compared with the 50% of random levels at the same hour it is inside the error.
Distance, touch number, size
We looked at three other things, because they are the questions traders ask.
Distance from the open. Levels 0-1 ATR from the day's open hold 46% (500 contacts), those 1-2 ATR 53% (166), 2-4 ATR 47% (139), over 4 ATR 68% (± 18, on only 25 contacts). No pattern: the 68% has an error of 18 points.
The touch number. The first touch of the same level holds 47% (499), the second 49% (248), from the third on 51% (± 11, 83). A level touched several times does not behave differently (chapter 6).
Yesterday's size. After a wide day (at least 98 $ of range) its levels hold 50% (369 contacts), after a narrow one 47% (461). Inside the error.
Read the chart in words
H1 gold chart of 19 February 2026 with the lines of yesterday's high and close and the six contacts marked as hold or break.
| When | What | Price |
|---|---|---|
| 18 Feb 2026 23:00 UTC | yesterday's high 5012.91 | 5,012.91 |
| 18 Feb 2026 23:00 UTC | yesterday's close 4979.25 | 4,979.25 |
| 19 Feb 2026 06:00 UTC | hold | 5,022.38 |
| 19 Feb 2026 14:00 UTC | hold | 5,011.99 |
| 19 Feb 2026 01:00 UTC | break | 4,978.53 |
| 19 Feb 2026 11:00 UTC | hold | 4,981.26 |
| 19 Feb 2026 14:00 UTC | hold | 4,966.28 |
| 19 Feb 2026 18:00 UTC | hold | 4,978.84 |
Times in UTC.
Read the chart in words
H1 gold chart of 20 February 2026 with the lines of yesterday's high, close and open and the six contacts marked as hold or break.
| When | What | Price |
|---|---|---|
| 19 Feb 2026 23:00 UTC | yesterday's high 5024.54 | 5,024.54 |
| 19 Feb 2026 23:00 UTC | yesterday's close 4999.04 | 4,999.04 |
| 19 Feb 2026 23:00 UTC | yesterday's open 4983.79 | 4,983.79 |
| 20 Feb 2026 06:00 UTC | break | 5,031.77 |
| 20 Feb 2026 11:00 UTC | hold | 5,019.30 |
| 20 Feb 2026 15:00 UTC | hold | 4,994.61 |
| 20 Feb 2026 00:00 UTC | break | 5,008.79 |
| 20 Feb 2026 15:00 UTC | hold | 4,994.61 |
| 20 Feb 2026 03:00 UTC | hold | 4,985.60 |
Times in UTC.
Days like these are the ones you remember: in the first, five holds out of six, in the second four out of six. The average over 830 contacts is 48%. A well-chosen chart always shows a nice level.
Read the chart in words
H1 gold chart of 10 March 2026 with the previous day's high and low lines and four candles marked with the letters A, B, C and D.
| When | What | Price |
|---|---|---|
| 9 Mar 2026 22:00 UTC | yesterday's high 5193.82 | 5,193.82 |
| 9 Mar 2026 22:00 UTC | yesterday's low 5019.70 | 5,019.70 |
| 10 Mar 2026 04:00 UTC | Point A | 5,166.47 |
| 10 Mar 2026 08:00 UTC | Point B | 5,192.00 |
| 10 Mar 2026 12:00 UTC | Point C | 5,186.60 |
| 10 Mar 2026 17:00 UTC | Point D | 5,232.32 |
Times in UTC.
What to do with it
- Compute the four levels at 18:00 New York time. High, low, close and open of the trading day just ended, with the ±0.25 ATR zone. Before trading, note how many ATR they are from the price.
- Use them as targets with a base rate. Yesterday's high and low are touched on 50% of days, the close on 84%. If a target is beyond one of yesterday's extremes, the price gets there on less than half the days.
- Do not expect a bounce. Hold of 48%, like random levels.
- Remember inside and outside days. 14% of days stay inside yesterday's range, 11% exceed it on both sides: most break only one extreme.
- Always compare with the same hour. The hold changes with the hour for random levels too (44% in the evening, 55% in London): comparing an evening level with a morning one measures the clock.
Procedure in four steps
- At 18:00 New York time write down high, low, close and open of the day just ended.
- For each compute the distance from the price in ATR and mark the zone (±0.25 ATR).
- If a target coincides with one of them, remember the base rate: 50% for the extremes, 84% for the close.
- At the end of the day log contacts, holds and breaks, to redo the measurement on your own data.
Exercises
Chapter exercises
1. On the chart the lines are yesterday's high (pink) and low (blue), and the letters A-D are on four candles of 10 March 2026 (UTC hours). Which letter is the first contact with yesterday's high?
Read the chart in words
| When | What | Price |
|---|---|---|
| 9 Mar 2026 22:00 UTC | yesterday's high 5193.82 | 5,193.82 |
| 9 Mar 2026 22:00 UTC | yesterday's low 5019.70 | 5,019.70 |
| 10 Mar 2026 04:00 UTC | Point A | 5,166.47 |
| 10 Mar 2026 08:00 UTC | Point B | 5,192.00 |
| 10 Mar 2026 12:00 UTC | Point C | 5,186.60 |
| 10 Mar 2026 17:00 UTC | Point D | 5,232.32 |
Times in UTC.
2. The next day touches yesterday's high on 50% of days and the close on 84%. What does it tell you about a target?
3. In the American morning yesterday's levels hold 57% (± 7) and in the evening 45% (± 5), but random levels with a contact at the same hour give 50% and 44%. How do you read it?
4. In 210 trading days 14% are inside (within yesterday's range) and 11% outside (break high and low); 38% break only the high and 37% only the low. 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.