Summary
HilbertTrendline is RTTA's streaming period-adaptive instantaneous trendline
(TA-Lib HT_TRENDLINE). It averages raw price over the current dominant-cycle
length, then applies a 4-bar weighted moving average of that average.
Update API
result = rtta.HilbertTrendline(fillna=True).update(value)
| Parameter | Default | Meaning |
|---|---|---|
fillna |
True |
If False, NaN until lookback of 63 samples is exceeded |
update(value) returns the current adaptive trendline as a scalar.
Theory Of Operation
Where a fixed-period SMA always lags the same amount, the Hilbert trendline shortens its averaging window when the dominant cycle is short and lengthens it when the cycle is long. TA-Lib averages raw prices over \(N=\operatorname{clip}(\lfloor\overline{P}+0.5\rfloor,1,50)\) bars (not the WMA smooth), then smooths that average with the same 4-tap weights used on the input price:
This is the adaptive counterpart to
EhlersInstantaneousTrendline, which uses
a fixed critical period and a two-pole recursion instead of a Hilbert period.
Recurrence
Let \(\overline{P}_t\) be the smoothed dominant period and
Average of the last \(N\) raw prices (from a retained price history of up to 64 samples):
(missing history treated as zero contribution in early bars when fewer than \(N\) prices exist).
Four-bar WMA of the averages:
Returned value: \(TL_t\).
With fillna=False, output is NaN until more than 63 updates have completed.
Implementation Notes
- Thin wrapper around
HilbertCycleEngine::trendline()(class HilbertTrendline). - Engine keeps
prices_(max 64), andi_trend1_,i_trend2_,i_trend3_as lags of \(A_t\). - Lookback:
lookback_phase_ = 63.