Adding VXN gave no clear forecast gain for monthly Nasdaq-100 losses of at least 5% with fixed logistic models, 2017–2026

We test whether Cboe VXN improves next-month Nasdaq-100 event probabilities after past variation and return trend are known. Each forecast uses an ALFRED historical input vintage. The test has 116 months from January 2017 to August 2026, including 16 price losses of at least 5%. Adding VXN to a fixed logistic model reduces Brier error from 0.119223 to 0.118775. The relative gain is 0.38%, with a paired 95% block-bootstrap interval from -1.15% to +2.29%. Fixed penalty, block-length and source-freshness checks also give intervals that include zero. Direction forecasts show no clear added gain. In a secondary test, the combined model reduces error for absolute moves of at least 5% by 11.54% against an expanding event rate, but past-return features already supply most of that gain. We preserve 4,176 forecasts, 3,132 fit records, all fixed settings, source hashes, code, figures, uncertainty draws and independent checks. This is a retrospective test on the same source sample as earlier SCI risk studies. It does not establish a new model, independent market validation, or a net trading advantage.

Nasdaq-100 · VXN · monthly event probability · historical data vintage · chronological forecast test · Brier score · negative result