Direct answer
Mean Reversion Range is calculated by picking a central value (the midpoint) from recent data, then expanding outward by a dispersion measure that represents how much the price typically varies. A common, checkable form is:
Mean Reversion Range (upper/lower bands)
- Midpoint: (M_t)
- Dispersion: (D_t)
- Lower bound: (L_t = M_t - k,D_t)
- Upper bound: (U_t = M_t + k,D_t)
Here, (k) is a multiplier that scales the band width. To calculate (M_t) and (D_t), you must define:
- what “recent data” means (the lookback window),
- which price series you use, and
- how you measure dispersion.
Mechanism and definition: what the terms mean
To calculate Mean Reversion Range, treat it as a statistical “band” rather than a trade signal.
1) Choose the midpoint (M_t)
The midpoint is a center-of-gravity estimate of price over a window. Common choices include:
- Simple moving average (SMA): the arithmetic mean of the last (N) observations.
- Exponential moving average (EMA): a weighted average that reacts faster to recent changes.
- Median (less sensitive to outliers): the middle value of the last (N) observations.
All of these are valid midpoint definitions, but they produce different band behavior.
2) Choose the dispersion (D_t)
Dispersion represents typical deviation around the midpoint. A widely used dispersion choice is standard deviation:
- Given observations (x_{t-N+1}, \dots, x_t),
- Standard deviation: (\sigma_t = \sqrt{\frac{1}{N-1}\sum_{i=0}^{N-1}(x_{t-i}-M_t)^2})
Then you set (D_t = \sigma_t).
Other dispersion definitions can be used if your formula states them clearly, for example:
- Average true range (ATR)-style dispersion (uses high/low/close relationships, not just a single price series),
- Mean absolute deviation (MAD) (uses average absolute distance, often less sensitive than squared deviations).
3) Choose (k), the band width multiplier
The multiplier (k) controls how wide the upper and lower bounds are.
- Higher (k) generally produces wider bands.
- Lower (k) generally produces tighter bands.
There is no single universal value: the “right” value depends on how your dispersion measure behaves and what your window length does.
4) Define the data series (x_t)
Many calculations depend on which price you feed into (M_t) and (D_t). Examples of input choices:
- Close price
- Typical price (often defined as average of high/low/close)
- Mid price (often based on bid/ask, if you are using quote-based data)
Your result is only comparable to other people’s results if they use the same input definition.
Example of a full, explicit calculation
Assume you define:
- (x_t) = close price at time (t)
- (M_t) = SMA over last (N) closes
- (D_t) = standard deviation over the same window
- (k) = chosen multiplier
Then:
- Compute (M_t = \frac{1}{N}\sum_{i=0}^{N-1}x_{t-i})
- Compute (\sigma_t = \sqrt{\frac{1}{N-1}\sum_{i=0}^{N-1}(x_{t-i}-M_t)^2})
- Set (L_t = M_t - k\sigma_t) and (U_t = M_t + k\sigma_t)
This is fully deterministic once (N), (k), and the input series are fixed.
Evidence or example you can verify independently
Because the calculation is formula-based, you can verify it using your own dataset and spreadsheet or code.
Verification checklist
To independently confirm the result, you need to ensure all of these match:
- Same window length (N)
- Same price series (x_t) (close vs typical vs another definition)
- Same midpoint method (SMA vs EMA vs median)
- Same dispersion method (standard deviation vs another measure)
- Same multiplier (k)
If any one item differs, the bands will differ, even if the formula structure looks similar.
Why multiple implementations differ
Two platforms can both say “mean reversion range” but implement different practical choices:
- If one uses EMA for (M_t) and another uses SMA, the center moves differently.
- If one uses standard deviation and another uses a volatility-like measure, (D_t) changes.
- If one uses close price and another uses a transformed price (like typical price), the inputs differ.
These differences are not cosmetic; they change the numerical bands.
Limitations and failure modes
Mean Reversion Range is a mathematical construction around a center and a dispersion measure. Its limitations come from those choices.
1) Window length can be unstable
- A short window can react too quickly to noise.
- A long window can lag regime changes.
In either case, the band may stop reflecting the “typical” behavior you intended.
2) Outliers can distort dispersion
Standard deviation squares deviations, so extreme observations can strongly widen (D_t), making bands less informative. Using a more robust dispersion (like MAD) is one mitigation, but that changes the meaning of the range.
3) Dispersion doesn’t guarantee mean-reverting behavior
Even if the band is computed correctly, the market (or dataset) may not show mean-reversion. The range is not a prediction; it is only a descriptive band based on past variability.
4) Costs, spreads, and execution are not part of the formula
Mean reversion range calculations typically ignore trading costs and execution details. If you later apply the numbers to real-world decisions, those factors can dominate outcomes. This is why the calculation itself should be kept separate from any interpretation.
5) Non-stationarity: historical statistics may not carry forward
The idea assumes that “recent variation” is a reasonable proxy for what comes next. If the data-generating process changes, the dispersion measure changes too slowly or too quickly.
Verification or next question
A good next step is to rewrite the exact version you want to compute in the form (L_t = M_t - kD_t) and (U_t = M_t + kD_t), then explicitly list:
- how you define (M_t),
- how you define (D_t),
- the chosen window (N),
- the multiplier (k), and
- the input price (x_t).
If you do that, you can reproduce the range from raw data with no ambiguity.
If you want, tell me your specific definitions for midpoint, dispersion, window length, and input price (for example: “SMA midpoint, standard deviation dispersion, close prices, N=20”), and I can restate the calculation in a fully explicit, step-by-step way.