MACD above 0.0003 on EURUSD, USDJPY and GBPJPY: why one number cannot be shared

[Fairman]

MACD above 0.0003 on EURUSD, USDJPY and GBPJPY: why one number cannot be shared

The MACD line is the difference of two EMAs, so it comes out in price units. Suppose a rule says: take longs when MACD(12,26,9) is above 0.0003.

On EURUSD, 0.0003 is 3 pips. On USDJPY it is 0.0003 yen, which is 0.03 pips, so the condition is almost always true. On GBPJPY the same.

A fix is to express the threshold in pips. On EURUSD a reading of 0.00031 is 3.1 pips. On USDJPY a reading of 0.031 is also 3.1 pips.

Still not enough. 3.1 pips on a quiet hour with an M5 ATR of 2.2 pips is 1.4 ATR, a strong move. The same 3.1 pips at the news hour with ATR 7 pips is 0.44 ATR, background noise.

Dividing the MACD line by the ATR of the same timeframe gives a unit free number. A rule such as MACD above 0.5 ATR means the same on every pair and in every hour.

[FXS HongKong Scalper]

Normalising by ATR is the cleanest fix. I use MACD divided by ATR(14) on the same timeframe as a dimensionless value. Then one threshold works across pairs and sessions, and in quieter hours like late Asia a modest move can still pass, which matches what the market is actually doing.

[Fairman]

In code the division is one line, and it avoids pip conversions altogether. With handles for MACD and ATR on the same symbol and timeframe, read the main buffer of MACD at shift 1 and the ATR at shift 1, then ratio = macd / atr. The shift matters: both from the last closed bar. If you use shift 0 for one and 1 for the other, the ratio mixes two bars. For display, a pip is 10 times _Point on 3 and 5 digit symbols, so pips = difference / (10 * _Point), but the ratio itself needs no conversion, which removes the 3 digit and 5 digit trap when you move between USDJPY and EURUSD.

[Fairman]

Rather than choosing 0.5 ATR by feel, take a distribution. Over 500 M5 bars, calculate the absolute value of MACD divided by ATR for each bar and sort them. The 80th percentile might be 0.62, meaning one bar in five is stronger than that. A rule at that level flags about 20% of bars whatever the volatility regime, so the threshold adapts as markets calm or speed up. The 0.62 is an invented example, and your figure will differ with pair and timeframe. What the process gives is a threshold with a known frequency, so you can see before trading whether the rule fires 3 times a day or 30.

[Fairman]

Timeframe changes the scale too. An H1 MACD reading of 0.0009 is 9 pips, three times the M5 figure, but H1 ATR might be 14 pips, so the ratio is 0.64. M5 at 3.1 pips with ATR 7 gave 0.44. The ratios are comparable even though the pip values are not. Using a pip threshold you would need a different number per timeframe, per pair, and maybe per session. With the ratio you need one number per idea. It also makes a multi timeframe rule easy to state: both M5 and H1 ratios above 0.5, which has the same meaning on each chart.

[Fairman]

The histogram, MACD line minus signal line, has the same scaling issue, in smaller numbers. On EURUSD a histogram of 0.00008 is 0.8 pips, on USDJPY it is 0.008 yen, and so on. A rule based on the sign alone, such as histogram above zero, or on the MACD crossing zero, has no scaling problem at all. The problem only appears once you add a minimum size. If you only use signs and crossings, your rules already travel across pairs, and the only thing left to consider is that the default 12, 26, 9 settings mean different clock times on different timeframes.