SqueezeMomentum

Summary

SqueezeMomentum is RTTA's streaming TTM-style squeeze indicator: a binary flag for Bollinger bands inside Keltner channels (volatility compression) plus a linear-regression momentum of price relative to a hybrid mid-line.

Update API

result = rtta.SqueezeMomentum(window=20, bb_mult=2.0, kc_mult=1.5, fillna=True).update(close, high, low)

Parameter	Default	Meaning
`window`	`20`	Lookback for BB, ATR/KC, Donchian, and linreg
`bb_mult`	`2.0`	Bollinger standard-deviation multiplier
`kc_mult`	`1.5`	Keltner ATR multiplier
`fillna`	`True`	If `False`, NaN until `window` samples

update(...) returns:

on — 1.0 if squeeze is on (BB inside KC), else 0.0
momentum — linear-regression value of the de-meaned series

advance(...) updates state; last() returns the cached result.

Theory Of Operation

Squeeze on means Bollinger bands (mean ± \(k\cdot\sigma\)) lie entirely inside Keltner channels (mean ± \(m\cdot ATR\)). That is a classic TTM Squeeze compression regime: realized volatility (BB) is low relative to average true range width (KC). When the squeeze "fires" (on flips from 1 to 0), energy is often released as a directional move.

Momentum measures where close sits versus a basis that averages the Donchian midpoint and the SMA of close, then fits a rolling linear regression (as in LazyBear / TTM-style histograms). Positive momentum favors upside release; negative favors downside.

Recurrence

Let \(n\) be the window, \(k=\texttt{bb\_mult}\), \(m=\texttt{kc\_mult}\).

Rolling mean and population stdev of close:

\[\mu_t = \operatorname{mean}_n(c),\qquad \sigma_t = \operatorname{stddev}_n(c)\]

\[BB^{hi}_t = \mu_t + k\sigma_t,\qquad BB^{lo}_t = \mu_t - k\sigma_t\]

With ATR of length \(n\):

\[KC^{hi}_t = \mu_t + m\cdot ATR_t,\qquad KC^{lo}_t = \mu_t - m\cdot ATR_t\]

Squeeze flag:

\[on_t = \begin{cases} 1 & BB^{lo}_t > KC^{lo}_t \;\land\; BB^{hi}_t < KC^{hi}_t \\ 0 & \text{otherwise} \end{cases}\]

Donchian mid and hybrid basis:

\[HH_t = \max_{0\le i<n} h_{t-i},\qquad LL_t = \min_{0\le i<n} l_{t-i}\]

\[basis_t = \frac{1}{2}\left(\frac{HH_t+LL_t}{2} + \mu_t\right)\]

\[\delta_t = c_t - basis_t\]

\[momentum_t = \operatorname{LinReg}_n(\delta)_t\]

(value of the regression fit at the current bar, via LinearRegressionCore).

When fillna=False and fewer than \(n\) samples have been seen, both fields are NaN.

Implementation Notes

Implemented in src/rtta/indicator.cpp in class SqueezeMomentum.
Uses rolling sum/sum2 for mean/variance, RollingExtreme for HH/LL, ATR, and LinearRegressionCore linreg_.
Result type: SqueezeMomentumResult (on, momentum).
Batch helper: batch_squeeze_momentum.
Note: Keltner mid uses the SMA of close (mean), not a separate EMA of typical price—matching this C++ implementation.

Summary

Update API

Theory Of Operation

Recurrence

Implementation Notes

Reference