A 252-session drawdown gave no clear gain in weekly Nasdaq direction forecasts in a 2018–2025 archive test

We tested whether the log gap from the past 252-session closing high improves weekly Nasdaq Composite direction probabilities after past returns and volatility are included. Dated ALFRED records supplied the inputs. Each forecast used a Saturday archive and prices through Thursday. Its target ran from the next Monday session close to the following Monday session close. The main test has 401 fresh-input weeks in 2018–2025. The drawdown model increased mean Brier loss by 0.27% against the matched comparison model. The 95% interval for loss reduction was −0.68% to +0.15%, from paired 13-week blocks. The past positive-return share had lower loss than either fitted model. All fitted probabilities were above 0.5, so their 61.60% accuracy matched an always-positive rule. Sixteen stale-input weeks remain in a separate check. All 834 training states passed timing checks. A fresh-process repeat matched 14 result files exactly. The package has code, forecasts, fit states, source records, uncertainty draws, figures and repeat instructions. This is a fixed historical test for one index and one input. It does not establish reliable return skill or trading profit.

stock return prediction · direction probabilities · drawdown · 52-week high · point-in-time data · chronological test · Brier loss · negative result