CCI: Typical Price and Mean Deviation

Keep the denominator specific

A common CCI uses typical price, (high + low + close) / 3. Subtract its simple average over n observations, then divide by 0.015 times the mean absolute deviation of those typical prices from that same average. Fidelity publishes this CCI construction.

Mean absolute deviation is not standard deviation. It averages absolute distances rather than squared distances followed by a square root. Substituting one for the other changes the scale. The constant 0.015 is a conventional scaling choice, not a confidence level or an estimate of future price risk.

An original three-observation example

Assume typical prices of 98, 100 and 102. Their simple average is 100. Absolute deviations are 2, 0 and 2, whose mean is 4 / 3, approximately 1.3333. Using the latest typical price of 102, CCI is (102 − 100) / (0.015 × 1.3333) = 100.

Now consider a separate window of 100, 100 and 106. The average is 102. Absolute deviations are 2, 2 and 4, averaging 2.6667. CCI is (106 − 102) / (0.015 × 2.6667) = 100 again. The price-unit movement is larger, but its scale relative to the window's dispersion produces the same result.

These small windows illustrate arithmetic only. They are not recommended live settings. A longer lookback can generate values beyond +100 or −100, and the exact sequence depends on the rolling input. If every typical price is identical, mean deviation is zero and an explicit flat-window policy is required.

Interpret distance without assuming a reversal

A positive CCI means the latest typical price exceeds its selected average. A negative value means it falls below that average. The magnitude depends on the denominator as well as the difference. A small deviation estimate can amplify modest price changes; a volatile window can damp a much larger absolute move.

Thresholds such as +100 and −100 are conventions. They do not guarantee that a fixed percentage of future readings will lie between them. They also do not establish whether a large positive distance should be followed as momentum or opposed as mean reversion. Those are different trading hypotheses.

Audit the rolling window

When a reading changes unexpectedly, check whether an old extreme left the window. Recompute the mean and every absolute deviation using the same current-window average. A common spreadsheet mistake averages deviations that were each calculated against a different historical mean, which does not reproduce the stated formula.

Compare standard-deviation envelopes and range-based volatility before treating all normalized indicators as equivalent. Use the overfitting reference when evaluating many CCI thresholds: selecting the best result from repeated trials can exaggerate how informative the final setting actually is.