A gold chart organizes observations; it does not reveal the future. Memorizing candle names is therefore different from building a decision process. This guide starts with verifiable numbers, separates interpretation from evidence, and shows how a chart observation becomes a rule that can be reviewed without hindsight.

All prices and costs are hypothetical, not current quotations or trade signals. Examples do not establish profitable patterns. We mainly discuss ordinary time-based candles. Identify the product, quotation unit and data feed before applying any calculation, because charts carrying the word gold need not represent interchangeable contracts.

1. Identify the instrument before interpreting the picture

Record the product name, data feed, currency, weight unit and chart time zone. A margined XAUUSD CFD is not every delivery month of gold futures or a local dealer's bullion buyback quote. Differences can reflect contract terms, session boundaries or quote sides rather than an error.

Also distinguish ordinary candles from transformed displays such as Heikin-Ashi, whose calculated values should not automatically be treated as executable prices. Keep the dataset name in your worksheet. Switching feeds halfway through an exercise can change the apparent results while leaving the written rule unchanged.

2. Read the four values before the color

OHLC means open, high, low and close. The body spans open to close and the wicks extend to the extremes. Color conventions are configurable.

A close displayed on a still-forming candle may simply be the latest observation. Fifteen minutes into an hourly interval, the final shape is unknown. A rule requiring a completed close above a level must wait for the interval to finish; a temporary move above it is a different event. Write the distinction into the rule rather than deciding after seeing the next candle.

3. Calculate the anatomy of one example

Assume O = 3,000, H = 3,018, L = 2,994 and C = 3,012, in USD per troy ounce. The range is 24, the body 12 and each wick 6. The body occupies half the range. These calculations describe the interval without assigning a probability to the next move.

Components of a hypothetical candle
ComponentCalculationResult
RangeH − L = 3,018 − 2,99424
Body|C − O| = |3,012 − 3,000|12
Upper wickH − max(O,C)6
Lower wickmin(O,C) − L6

For ordinary OHLC data, both open and close should lie inside the high-low range. If H equals L, a body-to-range ratio needs an explicit zero-range case. An impossible number may indicate a data or formula problem, not a special trading opportunity.

4. One candle does not reveal the path

The same four values can arise from a move to the high before the low, or to the low before the high, followed by the same close. OHLC alone does not identify which extreme occurred first or how many times a level was revisited.

This matters when both a stop and target lie within one historical candle. Assuming the target was reached first invents an event order that the dataset does not establish. Use suitably detailed, time-ordered observations or disclose a cautious rule for ambiguous cases. Never count every unknown sequence as a winner to improve a simulated result.

5. Aggregate timeframes consistently

Four M15 candles can form one H1 candle when boundaries, price type and coverage match. The example includes both rising and falling subperiods, although the hour closes above its open. One falling M15 candle does not, by itself, establish a reversal in a daily description.

Four complete M15 intervals aggregated to H1
IntervalOHLC
13,0003,0082,9963,006
23,0063,0183,0023,014
33,0143,0162,9942,998
42,9983,0152,9973,012
H13,0003,0182,9943,012

Assign each timeframe a purpose in advance, such as context versus event recording. Do not cycle through intervals until one agrees with an existing opinion. Check session boundaries, especially for daily and four-hour charts; different boundaries can produce different candles from overlapping observations.

6. Define trend structure reproducibly

Higher highs and higher lows require a rule for selecting pivots. One practice definition could require a high to exceed the highs of two candles on each side. That pivot is only confirmed after the two right-hand candles finish, not when the high first occurs.

Using the confirmed pivot as though it were immediately known introduces future information. In a simulation, delay recognition accordingly and define how ties are treated. Uptrend, downtrend and range are descriptions under a chosen method; different methods need not classify every period identically. Record uncertainty rather than drawing only the pivots that fit the eventual result.

7. Give support and resistance explicit boundaries

Instead of drawing a line through whichever historical turns look convenient, specify how a zone is selected from information already available. Define the lookback window, width and qualifying observations. Adding a new line whenever a prior interpretation fails makes the method difficult to evaluate.

A touch does not guarantee a reversal, and a temporary crossing does not guarantee continuation. Compare the zone width with observed movement and transaction costs. If the zone is so broad that you cannot say where the interpretation fails, it is not yet a usable entry-and-exit rule.

8. Treat pattern names as descriptions

A doji describes similar open and close values under a chosen tolerance; it need not have a small high-low range. A long wick describes prices outside the body. It does not identify a particular participant or prove that someone deliberately targeted an individual trader's stop.

For an engulfing rule, state whether the body or full range must be covered and whether equality qualifies. Identical pattern names can hide different definitions. A label supplies no automatic success probability. Calling a pattern confirmed is not a substitute for evaluating a precisely specified rule after costs.

9. Separate a touch, an intrabar crossing and a closing break

Suppose a zone ends at 3,010. A candle with high 3,018 and close 3,006 crossed the boundary during the interval but did not close above it. A close of 3,012 qualifies under this particular closing rule, yet says nothing certain about whether the next interval stays above.

Choose beforehand whether to wait for a close, a retest or neither. Waiting changes the available price and can miss moves; it does not eliminate false signals. Once the candle finishes, do not assume an order based on that completed information can always execute at the historical closing price.

10. Identify what volume and news timestamps mean

Check whether a volume series counts traded contracts in a specified market or represents tick activity in a feed. A provider's series is not automatically total worldwide gold turnover. Combining unlike units or venues without understanding them can create false confidence rather than independent confirmation.

Record announcements, holidays and missing observations separately. A large candle near a release does not isolate that release as its sole cause. Revised economic data may also differ from what was known at the time. When the cause is uncertain, say so instead of inventing a narrative for every bar.

11. Convert the observation into a cash plan

Specify the entry condition, invalidation, exit method and quantity separately. At a hypothetical entry of 3,012 and planned exit of 2,994, the distance is USD 18 per ounce. Two ounces produce a USD 36 price loss; an assumed USD 4 total additional cost makes USD 40 if execution occurs at the planned level.

If the actual exit is 2,990, the result becomes (3,012 − 2,990) × 2 + 4 = USD 48. This does not recommend placing every stop below a candle low. It shows that the same picture has different cash consequences depending on quantity and execution. Visual appeal cannot replace a risk calculation.

12. Practice without revealing the future

Advance one completed candle at a time and record the observation before revealing the next. Allow decisions of trade, no trade and insufficient information. Keep every qualifying case, including failures, instead of collecting successful screenshots. Choosing not to trade is part of the process, not missing data to discard.

When changing a rule, label the new version and reserve a separate period that was not used to tune it. Account for spread, fees and unavailable fills. A small sample's win rate is not a permanent property of a pattern. Simulated results also do not guarantee live outcomes, particularly when the model ignores execution uncertainty.

13. Match the chart to your real-world exposure

A USD-per-troy-ounce series is not automatically a local-currency bullion quote. Local products can involve different weights, purity, conversion costs and buyback terms. The same candle therefore need not create an equivalent opportunity in a physical purchase or another derivative.

Record both the event time zone and your viewing time zone, including the date when crossing midnight. A daily candle does not necessarily end at your local midnight. An English interface identifies neither a jurisdiction nor a session boundary. Keep units beside prices so that later calculations remain meaningful.

14. Review checklist and common questions

  1. Identify the product, units, price side and time zone.
  2. Use completed intervals for close-based rules.
  3. Separate observations from interpretations.
  4. Define patterns and lookback windows reproducibly.
  5. Avoid future data and invented intrabar sequences.
  6. Calculate quantity, costs and adverse execution.
  7. Retain all outcomes and check a separate data period.

Does a green candle mean buy?

No. It describes an open-close relationship under the color settings. It is not a complete trading decision.

Which timeframe is most accurate?

No interval guarantees an outcome. Suitability depends on available time, costs and a clearly defined decision process.

Why do two applications show different candles?

Compare the contract, data feed, time zone and chart transformation first. These can change OHLC without necessarily indicating an error.

Educational information. Background documentation checked 28 September 2026. All worked examples are exercises, not signals or promised returns.