Part 1: What a level is · Chapter 1 of 14 · 13 min

What a level is and how it is tested

A level is a price where the market is thought to react. We put it to the test: over 1,715 contacts with the highs and lows of the previous day, week, month and Asia the price "held" in 48% of cases and broke in 51%, like random levels shifted by a few ATR (49%). Levels are not useless, but they hold like a coin.

What a level is

A level is a price at which, according to whoever draws it, the market has already reacted or will react: a support if it sits below the price and should stop a fall, a resistance if it sits above and should stop a rise. Yesterday's high and low, a round number like 4,500 $, last Friday's close are levels of this kind.

Why would price react? There are three explanations, and it is worth knowing them because the book tests them all:

  • Orders accumulate. Whoever bought below a previous high puts the stop below, whoever wants to sell puts a limit order above, whoever follows the market waits for the break: around certain prices there are many orders, and the price stops or accelerates.
  • Memory. Many traders look at the same points (yesterday's high, the round number), and if everyone acts there the level becomes real through attention.
  • Reference. Even without special reactions, a level gives a point of comparison: you can ask whether the price is above or below it, and how far it is from the point where it already turned.

Each of these reasons makes sense. What remains is to see whether the price, in the data, really behaves differently when it meets a level. That is the book's question.

A level is not a thin line. Gold does not stop at exactly 4,500.00 $: it stops, or does not stop, in a zone of a few dollars. In the book a level is a price with a zone around it, a quarter of an ATR wide on each side. With the median ATR of our candles (18.6 $ on the market-open H1 candles) the zone is ±4.7 $, 9.3 $ in total.

The three words: contact, hold, break

To measure you need definitions that do not change depending on who looks at the chart.

  • Contact. The first H1 candle that enters the level's zone from outside. "From outside" means that the close of the previous candle was outside the zone: if the price arrives from below it is a resistance being tested, if from above a support.
  • Hold. From the contact point (the edge of the zone on the side the price arrives from) the price moves away by 1 ATR towards the starting side, with an H1 close, before doing the same on the other side.
  • Break. From the contact point the price goes 1 ATR the other way, beyond the level, with an H1 close, before coming back 1 ATR.

The check lasts 24 market-open H1 candles (one trading day). With this rule the price has a real "choice": the first of the two targets, at equal distance (18.6 $ as a median), decides the outcome. Only 1% of contacts reach neither within 24 candles, and we set those aside.

Symmetry matters. If we had asked "the price moves away by 1 ATR" against "the price passes the level by a single dollar", the break would almost always be the answer, and we would have measured the rule, not the level.

The calculations use market-open H1 candles, that is COMEX hours (from Sunday 18:00 to Friday 17:00 New York time, without the daily pause): 4,865 candles between 15 December 2025 and 6 October 2026. The perpetual's weekend, with almost flat candles, has no real levels to test (chapters 1 and 7 of the book on hours).

Figure 1. A hold (20-21 April 2026, UTC hours): the previous day's high (4,827.23 $) is a resistance, the price arrives from below and touches the zone at 22:00 (contact), then falls 1 ATR (15.1 $) before rising by the same amount: hold.Data: gold, XAUUSDT perpetual future (real candles) · 1 hour (H1) · from 20 Apr 2026 08:00 UTC to 21 Apr 2026 20:00 UTC.
Read the chart in words

H1 gold chart of 20 and 21 April 2026 with a level's zone, the contact point and the hold and break lines.

WhenWhatPrice
20 Apr 2026 08:00 UTClevel 4827.234,823.34 – 4,831.12
20 Apr 2026 22:00 UTCtouch4,832.36
20 Apr 2026 22:00 UTChold (15.1 $ from the touch)4,808.24
20 Apr 2026 22:00 UTCbreak4,838.43

Times in UTC.

Figure 2. A break (27 May 2026, UTC hours): the previous day's low (4,487.34 $) is a support, the price arrives from above and touches the zone at 06:00, then falls 1 ATR (14.6 $) without bouncing: break.Data: gold, XAUUSDT perpetual future (real candles) · 1 hour (H1) · from 26 May 2026 16:00 UTC to 28 May 2026 04:00 UTC.
Read the chart in words

H1 gold chart of 27 May 2026 with a level's zone, the contact point and the hold and break lines.

WhenWhatPrice
26 May 2026 16:00 UTClevel 4487.344,483.84 – 4,490.84
27 May 2026 06:00 UTCtouch4,483.00
27 May 2026 06:00 UTChold (14.6 $ from the touch)4,505.40
27 May 2026 06:00 UTCbreak4,476.29

Times in UTC.

Why a comparison is needed

If in 1,000 contacts the price holds 480 times, is that a lot or a little? It depends on what happens to prices that are not levels. The price of gold walks, and at every price there are days when it stops and days when it carries on: the market is never still at one point. To understand whether a level does anything special you have to compare it with an ordinary price.

The book's comparison works like this. For each real level we create a fake one: the same period of validity, but the price shifted by a random number of ATR, from 2 to 6, above or below. Then we measure hold and break on the fake levels with exactly the same rule. We repeat 30 times with different shifts. If real levels behave like fake ones, the level adds nothing.

A warning. Fake levels are touched less often than real ones: for the previous day's high and low, 347 contacts against 239 for fake levels, because real ones sit where the price has already been and so close to the current price. This says that levels "attract" the price, but not because they are special: they simply lie on its path. The comparison that counts concerns the outcome after the contact, and that is what we show.

The general picture

LevelsContactsHoldBreakHold of random levels
Previous day: high and low34748%51%49%
Previous day: close29748%50%49%
Previous day: open18648%51%47%
Previous week (high, low, close, open)44247%52%48%
Previous month: high and low3551%49%51%
Asia: high and low40850%49%50%
Round numbers every 100 $38544%55%49%
Round numbers every 50 $75244%54%48%

On the six "time-based" types of level (day, week, month, Asia) there are 1,715 contacts, the hold is 48% (± 2 points) and the break 51%. Random levels shifted by a few ATR give 47-51%. The margins of error (± 5-7 points for the types with 200-450 contacts, ± 17 for the month with 35) always contain the random-level result.

Two apparent exceptions. Round numbers hold a little less (44% against 48-49%): if it is a real effect, round numbers break more often than an ordinary price, the opposite of common belief. With ± 4-5 points of error and contacts that cluster around a few levels (17 and 34), it is an indication to retest, not a rule; we discuss it in chapter 5. The other is the difference between resistances and supports: resistances (841 contacts, arrival from below) hold in 47%, supports (874, arrival from above) in 50%. Inside the error.

One thing has to be read together with the others: 50% of these levels, the previous day's high and low, are never touched by the following day. Over 420 levels the contacts are 347, that is 0.83 per level: the price often does not even come back to see them.

Why levels seem to work

If holding is a coin, how come so many traders are convinced that levels work? Three reasons, all measurable or reasonable.

The line is drawn afterwards. On the finished chart you see the points where the price turned, and you draw the line there. A level drawn after the reaction "works" by definition. Our test draws the line before: yesterday's high is known at the 17:00 New York close, the contact comes later.

Bounces are remembered and breaks are forgotten. A bounce gives a story (the level held), a break is a change of scenario that gets filed away. It is not a conscious choice: the brain keeps confirmations better.

Many levels, many readings. On the chart there are always dozens of possible levels: previous day, week, round numbers, old swings. For every move there is at least one nearby. Whoever looks back always finds the level that "held".

What levels are for, then

If they do not predict the bounce, they are good for something: the rest of the book shows it with numbers.

  • As a risk reference. A stop must go at a point that makes sense; a level gives a place from which to measure the distance (chapter 12).
  • As a map of context. Knowing whether the price is half an ATR or four from the nearest level changes the plan for a trade (chapters 3-6).
  • As a home for orders and abrupt moves. Where stops accumulate, the price can snap (chapters 7 and 12).
  • As the basis of a system to test, with written rules and an honest comparison (chapters 11 and 13).
Figure 3. Exercise: four H1 candles around a contact with the previous day's low (23-24 April 2026, UTC hours), marked with the letters A-D.Data: gold, XAUUSDT perpetual future (real candles) · 1 hour (H1) · from 23 Apr 2026 15:00 UTC to 25 Apr 2026 02:00 UTC.
Read the chart in words

H1 gold chart of 23 and 24 April 2026 with the zone of the previous day's low and four candles marked with the letters A, B, C and D.

WhenWhatPrice
23 Apr 2026 15:00 UTClevel (previous-day low)4,663.89 – 4,674.77
23 Apr 2026 22:00 UTCPoint A4,707.57
24 Apr 2026 03:00 UTCPoint B4,680.88
24 Apr 2026 07:00 UTCPoint C4,696.61
24 Apr 2026 13:00 UTCPoint D4,719.18

Times in UTC.

What to do with it

  1. Draw zones, not lines. A level is a price plus a margin: a quarter of an ATR per side is a good starting measure.
  2. Define contact, hold and break before looking at the results. With fixed, symmetric rules that do not change with the chart.
  3. Always compare with fake levels. If your levels' hold is like that of levels shifted at random, the level adds no information.
  4. Count the contacts. With 35 contacts the error is ± 17 points: nothing can be said. You need hundreds.
  5. Do not expect a bounce. Expect a reference point. How far it is, and what to do with it, is the rest of the book.

Procedure in four steps

  1. Choose the type of level and write the rule (which price, for how long it is valid).
  2. Define zone, contact, hold and break with fixed numbers (for example ±0.25 ATR and 1 ATR).
  3. Count the contacts and compute hold and break with their error (± 1.96 times the square root of p(1 − p)/n).
  4. Redo everything on levels shifted at random and compare.

Exercises

Chapter exercises

1. On the chart the grey box is the previous day's low zone and the letters A-D are on four H1 candles. Which letter is the contact, the first candle that enters the zone from above?

Chart exercise. Data: gold, XAUUSDT perpetual future (real candles) · 1 hour (H1) · from 23 Apr 2026 15:00 UTC to 25 Apr 2026 02:00 UTC.
Read the chart in words
WhenWhatPrice
23 Apr 2026 15:00 UTClevel (previous-day low)4,663.89 – 4,674.77
23 Apr 2026 22:00 UTCPoint A4,707.57
24 Apr 2026 03:00 UTCPoint B4,680.88
24 Apr 2026 07:00 UTCPoint C4,696.61
24 Apr 2026 13:00 UTCPoint D4,719.18

Times in UTC.

2. Why does the book draw the level not as a line but as a zone of half an ATR per side?

3. Why do we compare real levels with fake levels (shifted by 2-6 ATR) with the same rule?

4. Over 1,715 contacts the hold is 48% (± 2 points) and random levels give 47-51%. 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.