Gold technical analysis, from scratch: support, resistance, moving averages and patterns

Let us be clear about one thing first: this article does not sell "secret levels" and it does not promise a win rate. What technical analysis can do for gold comes down to three things — writing entry conditions as positions anyone can verify, capping the cost of being wrong at a known number, and sizing every position as a percentage of account risk. Get those three right and you are a systematic trader. Get them wrong and no number of chart patterns will stop you from gambling.

Every figure below comes from today's real market: spot 4,118.01 USD/oz, a 30-session range of 4,073.93–4,402.73, the 20-day moving average at 4,158.21, daily RSI(14) at 34.5 and ATR(14) at 56.0 USD. Open a chart and check each one for yourself — you should not have to take my word for it.

Data captured 2026-10-20 23:04 (Beijing time) · Source: XAUAD live quotes

1. Why technical analysis only half works on gold

A lot of people port equity methods straight onto gold, conclude that technical analysis does not work, and stop there. The method is not the problem — the microstructure of this instrument is genuinely different, and you have to accept three differences before anything else makes sense.

1. No close, so no gaps

Gold is a 24-hour OTC market with London, New York and Asia handing over to each other; there is no real "close" in between. Strategies that depend on gaps and gap-fills are largely useless here. It also means the daily open and close are just arbitrary slices of time — one "big bullish candle" may only be two hours of New York trading. Read a chart with a continuous timeline in mind, not as a row of independent bars.

2. Liquidity is layered by session, so the same pattern is worth different amounts

The Asian session (09:00–15:00 Beijing time) is thin: a 0.50 USD move can print a long wick. The London open (15:00 in summer, 16:00 in winter) brings money in. New York (after 20:20) stacks COMEX futures on top and is the deepest part of the day. So a bearish engulfing in Asia is far less reliable than the same pattern in New York. A signal that shows up in the wrong time window is noise.

3. Slow variables set the direction; the chart only records what already happened

Central bank buying, ETF holdings, real yields and the dollar index are slow variables that move month to month or quarter to quarter. They decide which side gold is standing on, but they will not tell you whether to enter today. Technical analysis contributes exactly what they lack — a sense of location and of volatility, i.e. when to act. The two are not competitors; they are a division of labour.

2. Step one: three lines drawn correctly beat ten indicators learned

1. Prior highs and lows: daily swing points only

A prior high or low is a price where the market previously turned and left an obvious long wick or large body behind. The two cleanest on the board today are 4,073.93 (30-day low) and 4,402.73 (30-day high). How you draw them matters: use the daily chart, not the 5-minute, or you will end up with a dozen lines that contradict each other. A line only counts as valid once it has been tested three times or more — a single wick through it does not qualify.

Why do prior lows matter? Because they are the natural reference for a stop. Go long with the stop a few dollars under the prior low, go short with it above the prior high, and your loss has an objective basis instead of "it felt wrong so I got out".

2. Round numbers: where resting orders and option strikes collect

Prices like 4,000, 4,100 and 4,200 react not because they look tidy but because real orders sit there: option strikes are set on round numbers, institutional limit orders are left on round numbers, and retail mental stops cluster on round numbers too. At 4,118 the market is 18 USD above 4,100 — a "just left the round number" position, the kind that grinds back and forth and punishes chasing from either side.

3. The 20-day moving average: a filter, not a signal

MA20 sits at 4,158.21 with price roughly 40 USD below it, so the short-term structure is weak. The correct use of a moving average is to filter direction, not "buy the cross up and sell the cross down":

• Price above MA20 look for longs only; a short needs an additional reason;
• Price below MA20 look for shorts only; a long is counter-trend and should be half size;
• Price chopping back and forth across MA20 (close to today's situation) this is a range, so trade less often and stop forcing the trend.

The rule forecasts nothing, yet it removes half of your impulse trades. For most people, trading less is a far easier route to profit than trading better.

3. Candlestick patterns: in the wrong place, a pattern is worthless

Hammers, engulfing bars and inside bars have statistical significance close to zero on their own. The only time they carry information is when they print at a key level. The same long lower wick at 4,074 — a 30-day low — means "someone is absorbing supply here". In the middle of the range at 4,300 it means only "nobody had an order in for those five minutes".

So the correct order is: level first, pattern second, confirmation last. On today's tape, a logical long script looks like this:

1. Price falls to around 4,074 (the 30-day low, in the liquidity pocket under the 4,100 round number);
2. The daily candle prints a long lower wick, or the next day closes back above 4,074 (buyers confirmed);
3. Stop 10–15 USD under the wick low, i.e. around 4,055;
4. First target 4,158 (MA20), then 4,200.

Notice there is no "will it reach 4,074" in that script — only "if it gets there and confirms, this is what I do". The difference is enormous: the first version has you staring at the screen making wishes, the second lets you leave a conditional order and go to sleep.

The honest statistics on patterns

The "70% success rate for hammers" numbers floating around are almost always cherry-picked after the fact. In reality any single pattern bounces around a 45%–55% hit rate, and the only way to make money with it is a reward-to-risk above 1.5 and strict execution. Stop looking for certainty in the pattern; look for it in the payoff.

4. Two indicators you must be able to compute: RSI and ATR

RSI(14) = 34.5: weak, but not a blind buy

A reading of 34.5 means sellers have had the edge over the past 14 days. Plenty of tutorials say "below 30 is oversold, so buy", and on gold that fails constantly: in a trend RSI goes numb for weeks. Through this September 2026 slide from 4,402 to 4,118 the reading wandered between 30 and 45, and every "looks ready to bounce" moment turned into another leg down. The correct use is to treat RSI as one confirmation condition: key support + RSI below 35 + a bullish candle close, all three together, or no trade.

ATR(14) = 56.0: it sets your stop and your lot size

ATR is average true range — in plain terms, how far this instrument normally travels in a day. At 56 USD, a 20 USD stop does not mean you are trading; it means you are donating to the market. Any random 15-minute bar today can take you out.

A workable stop distance is 1.2–1.5 × ATR, i.e. 67–84 USD. Use the low end in a range and the high end around NFP or CPI. Once the stop is fixed, lot size is not something you feel like doing — it is calculated:

The relationship between the third and fifth columns is the whole point: you do not pick a round lot size because it looks neat, you pick the size that makes being stopped out cost about 1% of your capital. A 2,000 USD account is pinned by the 0.01 minimum, so one stop costs 3%–4%; such an account should not be day trading at all, it should be on a demo first.

5. Multi-timeframe alignment: daily for direction, 4-hour for the zone, 15-minute for the trigger

The classic single-timeframe mistake is that "every candle looks like a signal". Three timeframes, three jobs; drop one and the picture falls apart:

Daily: answers only "long or short"

Price at 4,118 is below MA20 (4,158), and since the 4,402 high the peaks have stepped lower — the daily conclusion is bearish. So every plan today defaults to finding shorts; a long is downgraded to "small counter-trend probe".

4-hour: answers only "where"

With a bearish bias, wait for a bounce into resistance. The 4,150–4,160 band, where MA20 overlaps the previous volume shelf, is a zone, not a price. A zone tolerates 10 USD of error; a single price does not. That is exactly why people leave an order at 4,158, miss the fill by 3 USD, and then watch the market run without them.

15-minute: answers only "when to pull the trigger"

Once price is inside 4,150–4,160, wait for a 15-minute structural break — the prior bar's low taken out, or a bearish engulfing — and only then enter. The benefit: if price runs straight through resistance instead, you never got in and never took the hit.

6. Spread and session: the cost floor under any technical view

No pattern survives a spread that eats it. The current spread is 0.66 USD, which looks trivial at 0.016% of the gold price. But eight intraday round trips turn it into 5.3 USD/oz of fixed expense:

This table explains why "high-frequency scalping of small moves" almost never survives in a retail account: you are not losing to direction, you are losing to friction. It also explains why session choice matters — the Asian session often widens to 0.8–1.2 USD while the London–New York overlap compresses to 0.3–0.5 USD. Same strategy, different hours, double the cost.

Three windows to avoid

• 04:00–06:00 Beijing time every day: Europe and the US are both done, the book is at its thinnest and wicks are at their longest;
• Around the London fixes (near 23:00 and 00:00 Beijing time): liquidity shifts instantly and false breakouts cluster;
• The minute NFP or CPI is released: the spread can widen 5–10× in seconds and stops fill at the worst price available.

7. The five most common technical-analysis mistakes

1. Hunting patterns in the middle of a range. In no-man's land like 4,200–4,300 any candle combination is meaningless. Patterns only work at the edges.
2. A stop tighter than 1 × ATR. ATR is 56 right now, so a 20 USD stop voluntarily drops your hit rate below 30%.
3. Using RSI to buy the dip. RSI going numb in a trend is normal; after 34.5 comes 28, then 25, then 22. RSI is a confirmation, never a reason.
4. Ignoring the orders parked at round numbers. Ten dollars under 4,100 is a stop-loss cluster, and it is the first place large players hunt.
5. Treating the spread as a rounding error. 0.66 USD sounds like nothing, yet it can eat 2% of capital in a month — more than most people make in a year.

8. A six-question checklist before every entry

Write these six questions on a sticky note and run through them before every order; they will block about 80% of impulse trades:

1. Which way is the daily trending, and is this trade with it or against it?
2. Am I entering at a key level (prior high/low, round number, moving average) or at some random price?
3. Where is the stop, how many dollars away, and how many ATRs is that?
4. At 1% account risk, what should the lot size be?
5. Where is the target, and is reward-to-risk above 1.3?
6. Is any data due in the next two hours, and will the spread widen?

If you cannot answer any one of the six, it is not an "opportunity", it is an "emotion". Gold trades 24 hours a day; you do not need to catch every swing — you need to act when the conditions line up, and only ever lose the amount you can afford.

9. Frequently asked questions

Q. Does technical analysis actually work on gold?
A. Yes, but not as forecasting. Gold is a 24-hour OTC market with no gaps and no consolidated volume, so patterns are only a probability tilt. Their value is that they give you three quantifiable answers: where to enter, where you were wrong, and how big. Fix those three and the outcome is decided by reward-to-risk and discipline.

Q. Which timeframe should I watch on gold?
A. Daily for direction, 4-hour for the zone, 15-minute for the trigger. Staring at a 1-minute chart means spending most of your day wrestling with the spread and with noise. If you can only look twice a day, daily plus 4-hour is enough; 15-minute charts are for day trading, and then you must sidestep the liquidity shift around the London fixes.

Q. How far away should the stop be?
A. Calculate it from ATR, never from feeling. Daily ATR(14) is about 56 USD right now, so a stop tighter than 1 × ATR is a gift to ordinary noise. 1.2–1.5× is the usual range, i.e. 67–84 USD. Fix the stop first, then back out the lot size from 1% account risk — never enter first and add a stop afterwards.

Q. Can I buy the dip when RSI is oversold?
A. Not on its own. RSI is about 34.5 now: weak, not oversold, and in a trend it can stay numb between 30 and 45 for weeks. Through the September 2026 decline from 4,402 to 4,118 it repeatedly "looked ready to bounce" and kept falling. Use it as a filter, and only alongside a key level and a candle confirmation.

Q. Technicals or fundamentals — which do I trust?
A. They do different jobs. Fundamentals answer "why": real yields, the dollar index, central bank buying and ETF flows set the medium-term direction. Technicals answer "when to act and where I am wrong". Set the bias with fundamentals, then time entries and exits with technicals; on technicals alone you will be stopped out again and again by slow variables.

Related tools