Write a strategy¶
Start with the complete EMA crossover strategy. A strategy subclasses Strategy; you provide init() and next(), and Backtest supplies the account and data.
Decide what belongs where¶
| Task | Location |
|---|---|
| Declare tunable values | Class attributes or inputs |
| Register EMA, RSI, or custom calculations | init() |
| Read the current bar and indicator values | next() |
| Place or cancel orders; update stops | next() |
| React to actual fills | Optional on_fill(fill) |
| Inspect order changes | Optional on_order(event); read event.snapshot |
| Report termination | Optional on_finish(summary) |
init() sees complete history, so use trailing formulas and keep indicator handles. next() sees the current prefix and is called even when an indicator is invalid. Check valid[-1] and check length before reading an earlier bar.
Work with state¶
self.position is exposure, self.orders contains pending orders, and self.trades contains open lots. self.close is a price series; use self.close_position() to exit. Store your own flags or counters on other instance attributes. Input fields and framework data are read-only.
Log decisions with self.log.info("signal detected") and read them later with bt.get_logs(). Logging a signal does not mean an order filled; inspect fills or on_fill for execution facts.
All Strategy methods and attributes · Orders · Composable strategies