Use indicators in a strategy¶
Register formulas in init() through self.ta. Consume their current values in next() only after checking the selected channel's validity. The complete indicator strategy demonstrates EMA, MACD, and composition.
Common recipes¶
| Goal | In init() |
|---|---|
| EMA of Close | self.average = self.ta.ema(self.close, length=20) |
| RSI | self.rsi = self.ta.rsi(length=14) |
| MACD channels | self.macd = self.ta.macd(fast=12, slow=26, signal=9) |
| Smooth a derived channel | self.smooth = self.ta.sma(self.macd.histogram, length=3) |
| Custom rolling calculation | self.mean = self.ind(lambda s: s.rolling(20).mean(), self.close.s) |
| Historical crossing signals | self.cross = self.ta.crossover(self.fast, self.slow) |
During next(), self.average[-1] is the latest value and self.average.valid[-1] says whether it can be consumed. Multi-output indicators require a channel, e.g. self.macd.histogram. Use self.ta.is_crossover(a, b) for a current boolean rather than registering a series inside next().
Avoid future information¶
The full input is available in init(). Use causal calculations; negative shifts or saving a future scalar can leak data. Indicator handles constrain visible history during next() but cannot repair a noncausal function. Arrays created by arithmetic or .s conversions are snapshots; retain handles for dynamic reads.
Confirmed pivots are published after the right-hand window finishes. Anchored formulas require an explicit 0/1 anchor series. Pattern score zero can mean a valid non-event; validity is separate from direction or strength.
Find a formula¶
Every catalog page includes a complete script, parameter descriptions, and output names. Start with EMA, RSI, ATR, or Bollinger Bands.
All shared arguments and handle operations · Custom indicators