Part 2: Where levels go · Chapter 5 of 14 · 11 min

Round numbers

Multiples of 50 and 100 $ hold 44%, a little less than grids of non-round prices (46-51%, mean 49%): a weak effect, which with ten grids to compare is almost explained by chance. They do not attract the extremes: 20% of the daily highs and lows fall within 5 $ of a multiple of 50, exactly what chance expects.

Why people talk about round numbers

4,000 $, 4,500 $, 5,000 $: in ten months gold has had a path that crossed them several times. Traders watch them because they place orders there (it is more natural to put a limit at 4,500 than at 4,487), because options trade at round prices, because the news "gold passes 5,000" makes a headline. The idea is that at those prices there are more orders than normal, and the price reacts: round numbers as "psychological" levels.

It is an idea that can be measured with a simple test, because round numbers have an advantage over other levels: they do not depend on the price's history. We can make a grid of prices (every 50 $, every 100 $) and shift it: if 4,500 is special and 4,517 is not, shifting the grid by 17 $ must change the hold.

The shifted-grid test

We took the grid of multiples of 50 $ and shifted it by 5, 10, 15... up to 45 $. For each grid we counted the contacts and the hold, with the chapter 1 rules (zone ±0.25 ATR, hold and break at 1 ATR from the contact point, 24 candles). Shift 0 is the grid of true round numbers.

Shift from the multiple of 50 $ContactsHold
0 $ (round number)75244% (± 4)
5 $74446% (± 4)
10 $76848% (± 4)
15 $77748% (± 4)
20 $81850% (± 3)
25 $77349% (± 4)
30 $77051% (± 4)
35 $75451% (± 4)
40 $77149% (± 4)
45 $75947% (± 4)

The nine non-round grids (shifted by 5-45 $) together have 6,934 contacts and a mean hold of 49%, with individual values between 46% and 51%. The true round number is at 44%. With a control like this, the result is:

  • Direction: round numbers hold a little less than an ordinary price, that is they break a little more often. It is the opposite of the common belief that makes them barriers.
  • Size: 5 points below the mean, with an error of ± 4 on the single value. It is a weak effect, and ten random values with a spread of about 2 points have an expected minimum around 46%: the 44% is not far from what chance produces as an extreme.

The same with the 100 $ grid (385 contacts for shift 0): hold 44% (± 5); the other shifts (10-90 $) range from 44% to 52%; the 44% value also appears at the shift of 50 $, which is itself a multiple of 50 and so a smaller round number.

Figure 1. A week of H4 candles (8-14 June 2026) with the seven multiples of 50 $ the price crossed (4,050-4,350; multiples of 100 in pink): in a span of 340 $ there are seven, one every 50.Data: gold, XAUUSDT perpetual future (real candles) · 4 hours (H4) · from 8 Jun 2026 00:00 UTC to 14 Jun 2026 20:00 UTC.
Read the chart in words

H4 gold chart from 8 to 14 June 2026 with the lines of the multiples of 50 dollars.

WhenWhatPrice
8 Jun 2026 00:00 UTC40504,050.00
8 Jun 2026 00:00 UTC41004,100.00
8 Jun 2026 00:00 UTC41504,150.00
8 Jun 2026 00:00 UTC42004,200.00
8 Jun 2026 00:00 UTC42504,250.00
8 Jun 2026 00:00 UTC43004,300.00
8 Jun 2026 00:00 UTC43504,350.00

Times in UTC.

What changes with "roundness"

You can ask whether some round numbers are rounder than others. In our data:

LevelsLevelsContactsHoldRandom levels (5th-95th percentile)
Multiples of 500 $ (four levels, from 4,000 to above 5,000)48544% (± 11)53% (45-62)
Multiples of 100 $ not of 5001330045% (± 6)47% (43-52)
Multiples of 50 $ not of 1001736744% (± 5)48% (44-51)

No clear difference between more and less round. The multiples of 500 are four levels and 85 contacts: the error is ± 11.

Do round numbers attract the extremes?

There is another version of the idea: not only "the price bounces", but "highs and lows form near round numbers". This can be measured without contact rules. For each of the 211 trading days we took the high and low (422 prices) and the distance from the nearest multiple of 50 $. If round numbers do not count, the distance is uniform from 0 to 25 $.

Distance from the multiple of 50 $Extremes observedExpected by chance
Within 2.5 $44 (10%)10%
Within 5 $83 (20%)20%
Within 10 $165 (39%)40%

With multiples of 100: within 2.5 $ 5% (expected 5%), within 5 $ 9% (expected 10%), within 10 $ 18% (expected 20%). Highs are within 5 $ of a multiple of 50 on 21% of days, lows on 18%. A 50 $ grid shifted by 25 $ (which is not a round number) gives 21%. Hourly closes are also within 5 $ of a multiple of 50 in 19% of cases, 20% expected.

The days' extremes do not pile up on round numbers: they sit where they sit by chance. A daily high exactly on a round number is an event you remember; out of 422 extremes, we remember the few that fall on the round number.

Figure 2. Sixty days of daily candles (2 March - 30 April 2026) with, below, the distance of the day's high from the nearest multiple of 50 $: the bars go from 0 to 25 $ irregularly, on average 12.0 $ (expected by chance 12.5 $), with no pile-up near zero.Data: gold, XAUUSDT perpetual future (real candles) · daily (D1) · from 2 Mar 2026 to 30 Apr 2026.
Read the chart in words

Daily gold chart from March to April 2026 with, below, the bars of the distance of the day's high from the nearest multiple of 50 dollars.

Why they seem to matter

Three reasons. The first we saw in chapter 2: the price is within 1 ATR of a multiple of 50 $ in 74% of candles, so a round number is almost always close to every move. The second is that a round number makes the headlines: "gold passes 4,500" we remember, "passes 4,487" we do not. The third is that the distribution of contacts is very uneven: on the multiples of 100 $, contacts range from 2 (the two highest multiples) to 46 (4,400), because the price spent months between 4,200 and 4,800 and only a few hours in the highest zone. Those few heavily visited levels make most of the contacts.

Figure 3. Exercise: four H1 candles around a contact with 4,100 $ (29-30 July 2026, UTC hours), marked with the letters A-D.Data: gold, XAUUSDT perpetual future (real candles) · 1 hour (H1) · from 29 Jul 2026 10:00 UTC to 30 Jul 2026 21:00 UTC.
Read the chart in words

H1 gold chart of 29 and 30 July 2026 with the zone of the round number 4,100 and four candles marked with the letters A, B, C and D.

WhenWhatPrice
29 Jul 2026 10:00 UTCround number 41004,094.40 – 4,105.60
29 Jul 2026 17:00 UTCPoint A4,056.91
29 Jul 2026 22:00 UTCPoint B4,094.99
30 Jul 2026 03:00 UTCPoint C4,077.98
30 Jul 2026 09:00 UTCPoint D4,081.38

Times in UTC.

What to do with it

  1. Do not give round numbers a special value. In our data they hold like an ordinary price, or slightly less (44% against 49%).
  2. Do not expect them to attract highs and lows. 20% of daily extremes are within 5 $ of a multiple of 50, as by chance.
  3. Use them for targets and stops as convenient references, not as barriers. A target at a round number is easy to read and remember; it is no more probable than the others.
  4. If you want to test a round number, shift the grid. It is the only comparison that separates the round number from the rest: same method, same rules, grids at 5, 10, 15... dollars.
  5. Distrust round numbers that "worked" on the chart. Among 422 highs and lows, a few dozen fall near a round number even by chance.

Procedure in four steps

  1. Choose the grid (50 $ or 100 $) and the contact rules (zone, hold, break).
  2. Compute the hold of the true grid and of nine grids shifted by 5-45 $.
  3. Compare the true value with the mean and the range of the shifted grids.
  4. If you want to say they attract the extremes, measure the share of highs and lows within 5 $ of the round number against the expected value (20% for 50 $).

Exercises

Chapter exercises

1. On the chart the grey box is the round number 4,100 $ and the letters A-D are on four H1 candles of 29-30 July 2026 (UTC hours). Which letter is the contact, the first candle that enters the zone from below?

Chart exercise. Data: gold, XAUUSDT perpetual future (real candles) · 1 hour (H1) · from 29 Jul 2026 10:00 UTC to 30 Jul 2026 21:00 UTC.
Read the chart in words
WhenWhatPrice
29 Jul 2026 10:00 UTCround number 41004,094.40 – 4,105.60
29 Jul 2026 17:00 UTCPoint A4,056.91
29 Jul 2026 22:00 UTCPoint B4,094.99
30 Jul 2026 03:00 UTCPoint C4,077.98
30 Jul 2026 09:00 UTCPoint D4,081.38

Times in UTC.

2. How can you tell whether a round number is really special?

3. 20% of the 422 daily highs and lows fall within 5 $ of a multiple of 50, and a grid shifted by 25 $ has 21%. What does it mean?

4. Round numbers hold 44% (± 4), the nine shifted grids on average 49%. 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.