SPY-Filtered Entries

A two-symbol strategy that trades anything subscribers pick — but only when SPY confirms a market-wide uptrend. Demonstrates data feeds, regime filtering, and Compounding sizing.

class SPYFiltered(Strategy):
    """Only buy when SPY confirms uptrend."""
    sma_period = Param(50, min=20, max=200)
    rsi_period = Param(14, min=5, max=30)
    trade = Symbol(session=USEquityDaily, trade=True)
    spy = Symbol(ticker="SPY")
    sizing = Compounding(percent=Param(95, min=50, max=100))

    def execute(self):
        spy_sma = SMA(self.sma_period, on=self.spy)
        rsi = RSI(self.rsi_period, on=self.trade)

        spy_bullish = self.spy.close > spy_sma.value

        if spy_bullish and rsi.crossed_above(30) and self.trade.is_flat:
            self.buy(self.trade, reason="SPY uptrend + RSI oversold")

        if not spy_bullish and self.trade.is_long:
            self.sell(self.trade, self.trade.position, reason="SPY lost uptrend")

Walkthrough

Two Symbols

trade = Symbol(session=USEquityDaily, trade=True)   # the trading slot
spy = Symbol(ticker="SPY")                            # data feed (read-only)

When more than one Symbol is declared, exactly one must be marked trade=True. The other is a data feed — its OHLCV is readable, but you can’t place orders against it (self.buy(self.spy, ...) is RC2_SIG_DATAFEED).

The trade slot is session-locked but ticker-open — subscribers pick the actual instrument they want filtered by SPY.

Indicators on different symbols

spy_sma = SMA(self.sma_period, on=self.spy)     # SMA of SPY's close
rsi = RSI(self.rsi_period, on=self.trade)        # RSI of the trading instrument

Each indicator picks its own source. The strategy sees SPY’s trend as a regime filter while reading RSI of whatever the subscriber bound to trade.

Two-condition entry

if spy_bullish and rsi.crossed_above(30) and self.trade.is_flat:

Both must be true: SPY trading above its 50-day SMA and the trading instrument’s RSI just exited oversold. Either condition alone wouldn’t be enough.

Regime-driven exit

if not spy_bullish and self.trade.is_long:
    self.sell(self.trade, self.trade.position, reason="SPY lost uptrend")

When SPY drops below its trend line, exit immediately — even if the instrument’s own RSI says nothing’s wrong. The thesis is “I only want exposure during market uptrends,” and this enforces it.

Compounding sizing

sizing = Compounding(percent=Param(95, min=50, max=100))

Compounding means “use 95% of current equity” — winners grow the position over time. Combined with the Param, the optimizer can find the right risk fraction.

In walk-forward, Compounding chains equity across windows (unlike FixedCapital, which resets to initial per window). For long-horizon backtests where compound growth matters, this is the right model.

What to tune

Param	Typical range	Effect
sma_period	20..200	Lower = more responsive regime; higher = smoother.
rsi_period	5..30	Lower = more entries; higher = fewer but cleaner.
sizing percent	50..100	Risk control — lower = less drawdown, less upside.

Common modifications

VIX as a second filter

Add VIX as a third symbol and skip entries when fear is high:

class SpyAndVixFiltered(Strategy):
    """Trade only when SPY is trending and VIX is calm."""
    trade = Symbol(session=USEquityDaily, trade=True)
    spy = Symbol(ticker="SPY")
    vix = Symbol(ticker="VIX")
    sizing = Compounding()

    def execute(self):
        spy_sma = SMA(50, on=self.spy)

        spy_bullish = self.spy.close > spy_sma.value
        low_fear = self.vix.close < 20

        if spy_bullish and low_fear and self.trade.is_flat:
            self.buy(self.trade)

        if not spy_bullish and self.trade.is_long:
            self.sell(self.trade, self.trade.position, reason="SPY turned")

Sector rotation pattern

Pair a sector ETF as the trading symbol with SPY as the filter — buy the sector when both it and SPY are trending:

trade = Symbol(session=USEquityDaily, trade=True)   # subscriber binds XLK, XLF, etc.
spy = Symbol(ticker="SPY")

Each subscriber gets a different sector — but the SPY filter applies identically.