`CCI`#

Description#

CCI (Commodity Channel Index, Donald Lambert) measures how far the current bar's typical price has moved from its rolling mean, normalised by the mean absolute deviation of the same window:

\[\begin{split} \begin{aligned} \text{TP}[t] &= (\text{high} + \text{low} + \text{close}) / 3 \\ \overline{\text{TP}} &= \text{SMA}(\text{TP},\ n) \\ \text{MAD} &= \text{mean}\big(\ |\text{TP} - \overline{\text{TP}}|\ \big) \quad \text{over the same window} \\ \text{CCI}[t] &= \frac{\text{TP}[t] - \overline{\text{TP}}[t]}{0.015 \cdot \text{MAD}[t]} \end{aligned} \end{split}\]

The 0.015 constant is a Lambert convention: roughly 70-80% of CCI readings fall in [-100, +100] for a normal-distributed input.

3-input, 1-output (FunctorBase<_, 3, 1>) on (high, low, close).

Parameters:

window_size (int, default 14).

Warmup: NaN for the first window_size - 1 samples.

Range-zero: returns 0 when the MAD is 0 (a perfectly flat window).

NaN handling: NaN inputs poison the rolling sum.

NaN handling#

Policy: ignore. A NaN in any input at index t causes the function to skip that step: output at t is NaN and internal state is unchanged. Subsequent finite samples are processed as if step t had not occurred.

Examples#

Usage example#

import numpy as np
import plotly.graph_objects as go
from plotly.subplots import make_subplots
from screamer import CCI

np.random.seed(0)
close = 100*np.exp(np.cumsum(np.random.normal(0, 0.01, size=300)))
open_ = np.concatenate([[close[0]], close[:-1]])
wick = np.abs(np.random.normal(0, 0.4, size=300))
high = np.maximum(open_, close) + wick
low  = np.minimum(open_, close) - wick
out = CCI(20)(high, low, close)

fig = make_subplots(rows=2, cols=1, shared_xaxes=True,
                    row_heights=[0.55, 0.45], vertical_spacing=0.08)
fig.add_trace(go.Scatter(y=high, name="high", line=dict(color="#888")), row=1, col=1)
fig.add_trace(go.Scatter(y=low, name="low", line=dict(color="#bbb")), row=1, col=1)
fig.add_trace(go.Scatter(y=close, name="close", line=dict(color="royalblue")), row=1, col=1)
fig.add_trace(go.Scatter(y=out, name="CCI(20)", line=dict(color="red")), row=2, col=1)
fig.update_layout(title="Commodity channel index (CCI)",
                  margin=dict(l=20, r=20, t=60, b=20),
                  legend=dict(orientation="h", yanchor="bottom", y=1.02, xanchor="right", x=1))
fig.update_yaxes(title_text="price", row=1, col=1)
fig.update_yaxes(title_text="CCI", row=2, col=1)
fig.show()

Implementation Details#

Holds one circular buffer of TP values with a rolling sum (the SMA part is incremental O(1)); the per-step MAD sweep over the window is O(window_size). Same trade-off as RollingMad, and the same reason: there is no closed-form O(1) MAD when the mean shifts each step.

Reference#

Bit-exact match to talib.CCI(high, low, close, timeperiod) post-warmup (verified to ~1e-11 in tests/test_third_party_alignment.py).

CCI#