Summary
MACD is RTTA's multi-output Moving Average Convergence/Divergence oscillator:
fast EMA minus slow EMA, a signal EMA of that difference, and a histogram of
MACD minus signal.
Update API
result = rtta.MACD(a=12, b=26, c=9, fillna=False).update(value)
| Parameter | Default | Meaning |
|---|---|---|
a |
12 |
Fast EMA period |
b |
26 |
Slow EMA period |
c |
9 |
Signal EMA period |
fillna |
False |
If False, NaN until warm-up |
update(value) returns a result with:
macd— fast EMA − slow EMAsignal— EMA ofmacdhistogram—macd−signal
advance(value) updates state; last() returns the cached result.
Related APIs: MACDFix (fixed 12/26), MACDExt
(selectable SMA/EMA types).
Theory Of Operation
MACD measures the gap between a short-horizon and long-horizon exponential average of price. When the fast EMA is above the slow EMA, intermediate momentum is positive; the opposite signals negative momentum. The signal line is a further EMA that smooths MACD so that:
- MACD / signal crosses are common entry triggers.
- Histogram visualizes the gap and its expansion/contraction.
- Zero-line crosses of MACD mark shifts in the fast/slow EMA order.
All three nested EMAs run with internal fillna=True so coefficients and seeds
advance every bar; the outer fillna flag only gates whether incomplete
warm-up samples are returned as NaN.
Recurrence
Let \(x_t\) be the input series and \(a,b,c\) the three periods.
When fillna=False, all three fields are NaN while the sample counter is less
than \(\max(a,b)+c\); otherwise the current MACD triple is returned.
EMA seeding and \(\alpha=2/(n+1)\) follow RTTA's EMA.
Implementation Notes
- Implemented in
src/rtta/indicator.cppinclass MACD. - Members:
EMA a_,EMA b_,EMA c_; resultMACDResult. - Default
fillna=False(unlike many other RTTA indicators that default True). - Batch helpers exist for MACD family series processing.