Direct answer: average of ATR in Forex
To “take the average of ATR” in Forex, you average a time series of ATR values over time. ATR itself is a volatility measure calculated from recent price ranges. After ATR is computed for each bar (for example, each candle), you can compute an average of those ATR values using a standard averaging rule such as a simple moving average (SMA) or an exponential moving average (EMA).
In the context of using ADX and moving-average-style smoothing, the key idea is consistent: first compute ATR per period, then apply the same type of averaging you would apply to other time series (like moving averages), but to the ATR series rather than to price.
Mechanics: inputs, steps, and what “average” means
- Compute ATR for each bar
- ATR typically uses the True Range (TR) of each bar, then averages TR over a chosen ATR lookback length. The result is a new series: ATR[t].
- Material assumption: you already have ATR values for each time step (either from a platform calculation or from your own calculation).
- Choose an averaging method for ATR
- Simple moving average (SMA) of ATR: for a window of N bars, compute the mean of ATR over that window.
- Conceptually: ATR_SMA[t] = (ATR[t] + ATR[t-1] + … + ATR[t-N+1]) / N.
- Exponential moving average (EMA) of ATR: compute a weighted average where more recent ATR values carry more weight.
- Conceptually: ATR_EMA[t] = EMA update formula applied to ATR[t].
- Choose parameters consistently
- ATR period: the lookback length used to create ATR.
- Averaging window (N) or EMA smoothing length: the lookback/smoothing used to create the “average of ATR.”
Example and checks you can do independently
Example workflow (no platform needed conceptually):
- Suppose you computed ATR for each bar and you now have ATR[t] values.
- To get an ATR average with a 10-bar SMA, average the most recent 10 ATR values at each time step once enough data exists.
Independent checks:
- Window sensitivity: compute the ATR average with two different window sizes (for example, a shorter and a longer N). If the “average ATR” line becomes much smoother and slower to change, that is expected from increased smoothing.
- Unit consistency: ATR is expressed in price units. The average of ATR keeps the same units.
- Data alignment: ensure you use the same bar indexing for ATR and its average; mixing time steps (for example, using ATR from one bar with price from another) changes results.
- Missing bars: if your data has gaps or different trading sessions, the ATR computation and therefore its average may behave differently. Averaging cannot fix inconsistent inputs.
Limitations and uncertainty
- Averaging does not remove uncertainty; it only smooths a known time series. It can reduce noise, but it introduces lag (especially for SMA).
- “Average ATR” is not a forecast of future volatility by itself. It summarizes past ATR behavior within the chosen windows.
- Results depend on parameter choices (ATR period and averaging window/EMA length). Different settings can produce different “average ATR” levels.
- Any claim about current market conditions would require current, primary data. This explanation describes the method, not current values.
- No guarantee of outcomes: averaging ATR is a mathematical transformation, not a promise of performance.
Related comparison: averaging ATR vs averaging price
A useful comparison is that averaging ATR differs from averaging price:
- Averaging price creates a smoother price path.
- Averaging ATR creates a smoother volatility path. Using ADX and moving-average-style concepts together typically means you are combining “trend strength” information (from ADX) with “smoothed volatility” information (from averaged ATR).