Summary
Alligator is RTTA's streaming Bill Williams Alligator: three shifted smoothed
moving averages of median price—jaw, teeth, and lips—that open and
close like an alligator's mouth as trend and consolidation alternate.
Update API
result = rtta.Alligator(
jaw_window=13, teeth_window=8, lips_window=5,
jaw_shift=8, teeth_shift=5, lips_shift=3,
fillna=True,
).update(high, low)
| Parameter | Default | Meaning |
|---|---|---|
jaw_window |
13 |
SMMA length for jaw (blue) |
teeth_window |
8 |
SMMA length for teeth (red) |
lips_window |
5 |
SMMA length for lips (green) |
jaw_shift |
8 |
Forward shift (bars of delay) for jaw |
teeth_shift |
5 |
Delay for teeth |
lips_shift |
3 |
Delay for lips |
fillna |
True |
If False, NaN while any line is still NaN from shift warm-up |
update(high, low) returns jaw, teeth, lips.
advance(...) updates state; last() returns the cached result.
Theory Of Operation
Median price \(m_t=(h_t+l_t)/2\) is smoothed with three independent Wilder SMMA
(RMA) windows. On classical charts the lines are drawn forward (future
shifted). In a causal streaming API that future shift is implemented as a
delay buffer of the same length: the value reported today is the SMMA from
shift bars ago, matching how a completed chart would have shown that SMMA
aligned with the current bar after the forward offset.
- Lips (fastest) react first when a trend starts.
- Teeth form the middle balance line.
- Jaw (slowest) is the sleeping / deep trend line.
When the lines are intertwined (mouth closed), the "alligator sleeps"
(range/chop). When they fan out in order, the mouth is open and a trend is
active. GatorOscillator quantifies the gaps between
these lines.
Recurrence
Let \(\operatorname{SMMA}_n\) be SmoothedMovingAverage
(SMA seed, then Wilder recursion with \(\alpha=1/n\)). Let \(\Delta_k\) be a
causal delay of \(k\) bars (ShiftBuffer: returns NaN until \(k\) samples have
entered, then the value from \(k\) bars ago).
Defaults: jaw \(13/8\), teeth \(8/5\), lips \(5/3\).
When fillna=False and any of the three delayed lines is still NaN, the result
struct is all-NaN.
Implementation Notes
- Implemented in
src/rtta/indicator.cppinclass Alligator. - SMMA instances always use
fillna=Trueso partial SMAs exist; shifts produce NaN until full. - Result type:
AlligatorResult(jaw,teeth,lips). - Batch helper:
batch_alligator.