🔄 CCI — Commodity Channel Index
CCI measures how far the current "typical price" has strayed from its own recent average, expressed in units of mean absolute deviation rather than standard deviation. Despite the name, it is used across every asset class, not just commodities.
💡 Financial Meaning
CCI was designed to flag the start of new cycles: readings beyond +100 suggest price is unusually strong relative to its own recent typical range, while readings below −100 suggest unusual weakness. Unlike RSI, CCI is unbounded — it can push far past ±100 during strong trends, so extreme readings should be read as "strength" rather than an automatic reversal signal.
🔢 Mathematical Formulas
-
Typical Price for each bar:
\[ TP_t = \frac{H_t + L_t + C_t}{3} \] -
Simple moving average of the typical price, and its mean absolute deviation:
\[ \overline{TP}_t = SMA_N(TP), \qquad MD_t = \frac{1}{N}\sum_{i=0}^{N-1} \left| TP_{t-i} - \overline{TP}_t \right| \] -
CCI, scaled by the conventional constant \(0.015\) so that roughly 70–80% of values fall inside \(\pm 100\):
\[ CCI_t = \frac{TP_t - \overline{TP}_t}{0.015 \cdot MD_t} \]
⚙️ Parameters
| Parameter | Key | Default | Description |
|---|---|---|---|
| Period (\(N\)) | period |
14 | Window for the typical-price average and mean deviation, in sessions. |
🎛️ Signal Processing Equivalent — Deviation Normalised by Mean Absolute Error
CCI is structurally similar to a Bollinger Band \(z\)-score, but it normalises by mean absolute deviation (MAD) instead of standard deviation. MAD is a more robust (less outlier-sensitive) estimate of dispersion than \(\sigma\), which is why CCI tends to react less violently to single extreme bars than a Bollinger-style normalisation would.
flowchart LR
HLC["📈 High / Low / Close"] --> TP["🎯 Typical Price = (H+L+C)/3"]
TP --> AVG["📏 SMA(N) of TP"]
TP --> DIFF["➖ TP − avg"]
AVG --> DIFF
TP --> MD["📐 Mean Absolute Deviation"]
DIFF --> DIV["➗ ÷ (0.015 × MD)"]
MD --> DIV
DIV --> CCI["🔄 CCI(t)"]±100 is a convention, not a law
The constant \(0.015\) was chosen by Donald Lambert so that, empirically, 70–80% of CCI values land between −100 and +100 for typical instruments. It is a heuristic calibration, not a statistical guarantee — unlike RSI's mathematically fixed 0–100 bound.