Turn-of-month means gave no clear forecast gain for daily Nasdaq Composite returns in a 2015–2025 archive test

We tested a fixed turn-of-month calendar forecast for daily Nasdaq Composite price returns. The calendar group contains the last market session and first three sessions of each month. Each monthly fit used twenty years of returns from an ALFRED version available before the first entry trade. The 2015–2025 test has 2,766 daily returns and 132 fits. Calendar means had 0.0366% more mean squared error than the overall past mean. The paired three-month-block 95% interval was −0.0325% to +0.1259%. A fixed 25% mix and weekday controls also gave no clear gain. Calendar direction probabilities had slightly more Brier loss. All direction rules predicted a rise on every date and had 55.71% sign accuracy. At a charge of 5 basis points per one-way trade, a fixed calendar paper allocation had a 1.51% annual mean price return, compared with 3.35% for a constant 20% index allocation. Their difference had an interval that included zero, and the calendar rule had more risk. The simulation omits cash interest, dividends and tracking costs. Independent forecast and cash-balance checks passed. A full replay matched all 13 original numeric result files. We preserve code, source provenance, forecasts, fits, 60,000 uncertainty draws, figures and negative results. Raw provider prices remain separately retained under provider terms. The result applies to one index, calendar rule, horizon and update policy. It does not establish that all calendar effects are absent or that a real fund could earn the paper returns. The outcomes overlap related SCI work and are not an independent market sample.

return prediction · turn of the month · Nasdaq Composite · point-in-time data · chronological tests · negative result · paper trading · transaction costs