Task 12
Seasonality / Calendar Effects → Strategy → BacktestEnter a ticker. We compute historical calendar statistics (month-of-year returns, sell-in-May split, turn-of-month) and an LLM picks a calendar rule. The rule itself is lookahead-free (the calendar is known in advance) — but the pattern is estimated in-sample, the weakest form of edge, so a weak signal correctly defaults to buy-and-hold.