golden cross backtest

What the numbers mean

This is a rule set with very few decisions in it. A 50-day average has to climb above a 200-day average, which is a slow event, and the only way back out is the same pair crossing the other way. There is no stop loss, no profit target and no maximum holding period, so a position that opens can stay open for years. What you get is a record built from a handful of long holds rather than a distribution of many trades.

That structure sets the shape of everything rendered beside this text. The win rate is high by construction: the trend has to be established before the entry fires, so the signal arrives late and mostly confirms something already underway. The flip side is that the exit arrives late too. A death cross needs months of falling prices to form, so the strategy sits through the whole first half of a decline before the rule says anything. Read the drawdown figure as the price of that lag, not as a flaw in the test.

Two things deserve scrutiny. The trade count is small — a crossover this slow fires only a few times per symbol — so the profit factor and win rate describe particular holds rather than a repeatable process, and a single long hold can dominate the profit-factor line. And the five tickers are not five independent bets: SPY and QQQ overlap heavily with AAPL, MSFT and NVDA, so the 20%-per-position sizing that looks diversified behaves closer to one concentrated position that enters and exits together.