Support, Resistance and Supply/Demand Zones: How to Draw Them, Tested on Gold

Support is a price area where price stopped falling and bounced before. Resistance is an area where price rose and got pushed back down. Demand and Supply zones are another way of drawing the same kind of area, based on the short pause price took before a strong move away. This article turns both into clear drawing rules, then looks at what happened when we backtested zones on gold: where they helped and where they did not.

How to draw support and resistance

Support and resistance start from swing points. A swing high is a candle whose High is above the candles on both sides. A swing low is a candle whose Low is below the candles on both sides.

1. Start on a higher timeframe, such as H4 or D1, then move down to the timeframe you trade. Levels from higher timeframes are seen by more traders. More in Multi-timeframe analysis
2. Look for areas where several swings cluster, rather than drawing a line from every swing you see
3. Draw a zone, not a single line. Use the wicks for one edge and the candle bodies for the other. Make it wide enough to cover the small overshoots of several turns, but not so wide that it eats the distance to your target. ATR (the average range per candle, see the ATR article) helps you judge the width
4. Only count real turns. Price simply passing through a zone is not a confirmation
5. Keep only the zones near current price. Two or three on each side are enough. A screen full of lines makes every spot look like a level

Zones matter because price often pokes a little past a level before turning. If you think in single lines, you will see “broken levels” more often than they really happen, and you will tend to place your Stop loss too tight.

Role reversal: resistance becomes support

When price closes clearly above resistance, the old level is often treated as support next time. This is called role reversal. The usual explanation: traders who sold at resistance and are now losing want out when price returns near their entry, and traders who missed the breakout wait to buy the pullback.

A common rule is to wait for price to close beyond the zone, not just poke through with a wick, and to wait for the retest before entering. The price you pay is missing the moves that run away without a pullback.

What are Supply and Demand zones

The idea is that a strong move away from a base shows large buy or sell orders that were not fully filled. When price returns to that base, the remaining orders may push price again.

• Demand zone: a base where price paused before moving sharply up
• Supply zone: a base where price paused before moving sharply down
• Zones are named by the direction before and after the base. DBR (Drop-Base-Rally) and RBR (Rally-Base-Rally) are Demand. RBD and DBD are Supply

There is no single standard definition, and traders draw them differently. In our tests we used these rules:

• Departure candle: body at least 1 ATR(14), body at least 50% of the candle’s full range, and it must close beyond the zone
• Base: 1–3 candles before the departure, each with a body no more than 50% of its range, and the whole base no wider than 1.5 ATR
• Demand zone edges: top = the highest base candle body, bottom = the lowest wick of the base (Supply is the mirror image)
• Broken when price closes beyond the far edge, and dropped after more than 2 touches

Example: finding a Demand zone step by step

Say gold has ATR(14) = 2.00 dollars. Price has been falling, then these candles form:

Candle	Open	High	Low	Close
Base 1	2,403.8	2,404.6	2,402.1	2,403.2
Base 2	2,403.2	2,404.4	2,401.8	2,403.9
Departure	2,403.9	2,408.6	2,403.5	2,408.2

1. Departure: body 2,408.2 − 2,403.9 = 4.3, more than 1 ATR (2.00) ✓. Range 2,408.6 − 2,403.5 = 5.1, so the body is 4.3 ÷ 5.1 = 84% ✓
2. Base 1: body 0.6 out of a 2.5 range = 24% ✓. Base 2: body 0.7 out of 2.6 = 27% ✓
3. Base width 2,404.6 − 2,401.8 = 2.8, not more than 1.5 × 2.00 = 3.0 ✓
4. Top = highest base body = 2,403.9 (Base 2’s close). Bottom = lowest Low = 2,401.8
5. The candle before the base was falling, so this is a DBR zone, 2.1 dollars wide

Price touches lines often, even for no reason

Before looking at test results, you need this. We simulated a pure random walk of 100 candles, each moving about 1 unit, and repeated it 100,000 times. After 100 candles, price was typically about 10 units from where it started.

• A line 5 units above price was touched in 57.9% of runs
• A line 10 units above price was touched in 29.2% of runs
• With support 5 units below and resistance 5 units above, price touched at least one of them in 97.6% of runs

Nobody places orders at any line in a random series, yet price still touches lines almost every time, and after a touch a random price turns back half the time. So “touched and bounced” on a past chart can happen without any real level. The more lines you draw, the more of them will look accurate. More in Why past charts always look easy.

The right question is not “did price touch it and bounce?” but “does it bounce more often than a line placed at random?”

What we found in our backtests

In our tests, point 1 is a count of chart points with no trades opened; the other points are trading systems on gold (XAUUSD) M1, 0.10 lot per trade and a 10,000-dollar balance. 2026 (January to September) used real tick data. Earlier years used data built from 1-minute prices. Backtest results do not predict future results, and the numbers below are not returns you should expect.

1. Support and resistance did not line up with Fibonacci levels more than chance. Many traders look for spots where support or resistance overlaps a Fibonacci level. We drew 8 support and resistance lines from clustered swings (merged into one level when within 0.35 ATR of each other), then checked whether each line sat within 3% of the swing from a Fibonacci level of the latest swing. We compared that with a fake set of levels shifted by 11.8%.

Data	Points measured	On a Fibonacci level	On a fake level
M1, 2026	26,616	35.9%	38.2%
H4, 2022–2026	4,975	33.7%	32.1%

The gap is under 2.5 points and flips sides. The “confluence” you see on a chart happens about as often as with levels we made up.

2. Zones did not work as entries or Stop loss levels. We added Supply/Demand zones to a Fibonacci pullback system (wait for a 61.8% retracement of the latest large swing, Stop loss at 78.6%).

• Stop loss just beyond the zone edge (1 or 3 dollars past it) won only 15–32% of trades and did worse than the 78.6% Stop loss in every version we tried. On tight versus wide stops, see Wide or tight Stop loss
• Only entering when an M15 zone sat at the entry: Profit factor (total profit ÷ total loss) improved in 2026 (1.52 vs 1.40) and 2025 (1.34 vs 1.29), but clearly worsened in 2024 (0.92 vs 1.33), and only about 60 trades a year were left
• Zones from M1 were so many that they filtered nothing. Profit factor was worse than using no zones in all 3 years

The one exception we saw: using M15 zones to filter signals from very small swings turned versions that nearly wiped out the account into positive results in all 3 years, but per unit of risk they were still worse than the large-swing system.

3. Zones worked as Take profit targets. The original system took profit at a near Fibonacci target. We changed it to take profit at the second opposing zone ahead (a Supply zone from M15 for a Buy). If there was no zone ahead, it kept the original target. Entries and Stop loss stayed exactly the same.

Year	Original target: profit / DD	TP at zone: profit / DD
2024	+1,178 / 346	+2,188 / 858
2025	+2,091 / 1,004	+3,242 / 1,485
Jan–Sep 2026	+5,320 / 1,557	+13,322 / 2,984
3-year total	+8,589	+18,752

DD is the largest drop in equity along the way (equity drawdown), in dollars. Total profit rose 18,752 ÷ 8,589 = 2.18 times, but DD also rose 1.9 times (2026), 1.5 times (2025) and 2.5 times (2024), so in 2024 profit per unit of drawdown got worse. The win rate fell from about 56% to about 28–39%, depending on which zone was targeted, because the targets were further away. To use this at the same risk, you would need to roughly halve the lot size.

With the Stop loss moved to 82% and 2023 added, both M15 and M30 zones were positive in all 4 years (2023–2026), though 2023 was very thin.

TP at 2nd zone (SL 82%)	2023	2024	2025	Jan–Sep 2026
Zones from M15	+220	+2,105	+4,911	+14,058
Zones from M30	+648	+2,993	+2,884	+17,810

We did not run the original system on 2023, so there is no comparison for that year. Zones from H4 and D1, which sit further away, lost money in 2025 in every version (−2,333 to −3,487 dollars).

In our tests, zones did not tell us where price would turn. They helped decide how far to hold a trade.

Limits: one market, and one system that originally used a near target. Gold ran hard in 2026, which especially favours far targets, and systems that already use far targets did not benefit from zones this way. For the bigger picture on choosing targets, read Near or far Take profit.

Combining zones with risk management

Using the example zone above, say you buy at the top edge, 2,403.9, and there is a Supply zone ahead at 2,415.0–2,417.2.

1. Stop loss below the bottom edge with a buffer, for example 2,401.8 − 1.0 = 2,400.8. Distance 2,403.9 − 2,400.8 = 3.1 dollars (remember that in our tests, stops this close to a zone did worse than wider ones)
2. Take profit just before the bottom of the Supply zone, for example 2,415.0 − 0.2 = 2,414.8. Distance 2,414.8 − 2,403.9 = 10.9 dollars
3. RR = 10.9 ÷ 3.1 = 3.52. Break-even win rate = 1 ÷ (1 + 3.52) = 22.1%. See Win rate vs break-even
4. Lot size: a 5,000-dollar account risking 1% = 50 dollars. On most accounts 1 lot of gold = 100 ounces (check your account’s contract specification), so a 1-dollar move = 100 dollars per lot. 50 ÷ (3.1 × 100) = 0.161, rounded down to 0.16 lot. Actual risk 0.16 × 100 × 3.1 = 49.6 dollars

The zone sets the Stop loss distance, but the risk per trade always comes from the lot size. See the full method in Position size from Stop loss, and record every zone before price reaches it in your trading journal, including the times the zone breaks.

Common mistakes

• Drawing a line at every swing. The chart fills up until price is always touching something. Keep only the important zones near current price
• Treating levels as exact lines. A small overshoot looks like a break, or the Stop loss sits so close to the edge that it gets swept often
• Drawing zones after seeing the bounce. Every zone drawn in hindsight looks accurate. Draw them before price arrives, then count both bounces and breaks
• Using zones from too small a timeframe. On M1 there are so many zones that they filter nothing. In our tests M15–M30 worked better
• Using confluence as a reason to trade bigger. In our data, overlap with Fibonacci happened about as often as chance. It is not a stronger signal
• Stretching the target to a far zone without cutting lot size. Far targets made drawdown about twice as large in our tests, so the same lot size means more risk