RTTADocumentation Built by Goblin Reactor ↗

InverseFisherRSI

Incremental, causal technical analysis documentation

Summary

InverseFisherRSI is RTTA's streaming Ehlers inverse Fisher transform applied to RSI. RSI is mapped to a roughly symmetric domain, smoothed with a WMA, then passed through the inverse Fisher map to produce a sharp oscillator typically in \((-1, 1)\).

Update API

value = rtta.InverseFisherRSI(rsi_window=5, wma_window=9, fillna=True).update(close)

With fillna=False, output is NaN until rsi_window + wma_window samples have been processed.

Theory Of Operation

John Ehlers' inverse Fisher transform amplifies values near zero and compresses extremes, turning a smooth oscillator into one with clearer turning points. The common recipe for RSI is:

  1. Compute RSI.
  2. Scale RSI so mid-scale (50) maps near 0 and extremes map near \(\pm 5\).
  3. Smooth with a short weighted moving average.
  4. Apply \(\tanh\)-equivalent inverse Fisher on the smoothed value.

RTTA uses its streaming RSI and WMA primitives for steps 1–3.

Recurrence

Let \(c_t\) be close, \(n_r\) be rsi_window (default \(5\)), and \(n_w\) be wma_window (default \(9\)).

\[RSI_t = \operatorname{RSI}_{n_r}(c_t)\]
\[x_t = 0.1\,(RSI_t - 50)\]
\[y_t = \operatorname{WMA}_{n_w}(x_t)\]
\[e_t = \exp(2 y_t), \qquad IFR_t = \frac{e_t - 1}{e_t + 1} \quad\text{(safe divide)}\]

Note \(\frac{e^{2y}-1}{e^{2y}+1} = \tanh(y)\). Nested RSI/WMA are constructed with fillna=True; the outer warm count is \(n_r + n_w\).

Implementation Notes

The recurrence is implemented in src/rtta/indicator.cpp in class InverseFisherRSI. See also RSI.

Reference