CHARTARMORYChartArmory home →
Free guide

How to read the weekly average candle (Heikin-Ashi)

Daily candles change their mind several times a day. Fold a whole week into one 'average candle' and only the big direction is left. From the calculation to the reverse formula for the deciding price and the traps that change the result, here it is with real Bitcoin numbers.

1. The basic calculation in three lines

  • Average close = (this week's open + high + low + close) ÷ 4
  • Average open = (last week's average open + last week's average close) ÷ 2
  • Body = average close ÷ average open − 1

The average open is recursive: it comes from last week's values. So each week's average candle carries a little of the weeks before it.

2. Work it out: Bitcoin, the week of 5–11 October 2020

OpenHighLowClose
10,670.0111,498.9810,524.0011,374.01
  • Average close = (10,670.01 + 11,498.98 + 10,524.00 + 11,374.01) ÷ 4 = 11,016.75
  • Average open = (10,710.68 + 10,692.50) ÷ 2 = 10,701.59
  • Body = 11,016.75 ÷ 10,701.59 − 1 = +2.95% → green, more than 1%

Coinbase BTC-USD daily data (the same prices as TradingView COINBASE), grouped Monday to Sunday (UTC).

3. Reading the colour: the 1% rule

  • Body above +1%: green — buy next week
  • Body below −1%: red — sell next week
  • Anything in between is a small candle — noise, ignore it (a small red week while holding is only a warning)

4. Five weeks in a row from the same period

Three weeks that went up and down on the daily chart become 'small candles', and the colour turns clear once a direction appears.

WeekBodyReading
2020.9.14–9.20−0.77%small · ignore
2020.9.21–9.27−0.07%small · ignore
2020.9.28–10.4−0.17%small · ignore
2020.10.5–10.11+2.95%green
2020.10.12–10.18+5.47%green
9.149.219.2810.510.12

5. Working the deciding price backwards (with exceptions)

During the week what you want to know is 'at what weekly close does the signal flip?'. Call the target average close T (buy: T = average open × 1.01, sell: T = average open × 0.99):

  • Basic: deciding close = 4 × T − open − high − low
  • Exception ①: if that is below this week's low, the close would become the new low, so deciding close = (4 × T − open − high) ÷ 2
  • Exception ②: if that is above this week's high, the close would become the new high, so deciding close = (4 × T − open − low) ÷ 2

The 5–11 October week above: buy T = 10,808.60 → 4 × T − open − high − low = 10,541.42 (between the low and the high) → a close above 10,541.42 makes it more than 1% green. Sell T = 10,594.57 → the basic formula gives 9,685.29, below the low (10,524.00), so exception ① applies and the deciding close is 10,104.65.

The high and low keep changing during the week, so the deciding price has to be worked out again every day.

6. Three traps that change the result

  • The first-week seed: the data's first average open starts as (open + close) ÷ 2. A different starting point gives different values for the first few weeks.
  • The week boundary: TradingView's crypto weeks run Monday to Sunday (UTC). Grouping by another boundary, such as a Friday close, can change the colour of the same prices.
  • Exchange differences: an almost flat week, like one in November 2018, can flip colour depending on the exchange's prices (Coinbase vs Bitstamp). The 1% filter screens out such weeks.

7. A real case: 28 weeks from October 2020

Buying at the open the day after the green week (12 October 2020) at $11,374, you would sell the day after the first week whose body closed more than 1% red, 19–25 April 2021 (−9.11%), at $49,121. That is +331% after 0.1% trading costs.

+331%held 28 weeks
−39.3%fall within 90 days of the sale

The week just before, 12–18 April, still had a +1.19% green body. Within 90 days of the sale Bitcoin fell to $29,796.

8. Limits

This rule does not always win. On Bitcoin since 2015 only 24 of 52 trades won, and sideways markets brought several small losses (−8.5% on average). One big trend (wins average +53%) makes up for them. The rule is tuned for crypto weekly candles; on stocks it did worse than just holding.

9. The weekly checklist if you do it by hand

  1. When the weekly candle closes at Sunday 24:00 UTC, note that week's open, high, low and close from the same exchange.
  2. Work out this week's average open from last week's average open and close (miss one week and every later value is off).
  3. Work out the average close and the body and compare them with +1% / −1%.
  4. Work the deciding price for next week backwards, and redo it every time the week's high or low changes.
  5. Buy or sell the next day (Monday) only in a signal week; otherwise let the week go by.

Do you want to do this by hand every week?

No spreadsheet, no headache: switch on Calm Candle and it draws the weekly average candles and the deciding price for you, every week. When a week ends clearly green it tells you "buy next week"; when it ends clearly red, "sell next week" — in one plain sentence. All you do is check that one line once a week.

See Calm Candle →

This guide helps with chart analysis. It does not guarantee profits, and investing involves a risk of loss. The tables and the case are calculated on past data and do not guarantee future returns.
Heikin-Ashi (the average candle) is a public calculation. Candlestick charts are often credited to the 18th-century Japanese rice trader Munehisa Homma, though whether he invented them is uncertain.