QR · independent research · final stage

Can the Treasury curve forecast a five-day 10Y quote change?

Work from original U.S. Treasury 2010–2025 daily nominal par-yield CSVs. Define a falsifiable claim, implement the study, compare zero-change and training-mean baselines, and defend the result. The target is a quoted yield change in percentage points, not a tradable bond return.

Remove labels crossing splits and embargo at least five trading observations. The primary measure is MAE; report direction, year-specific errors and block-bootstrap uncertainty. Retain every failed trial. Export the original files, hashes, fixed dependencies, code, log and report in one local package.

Treasury source and methodology · Full QR path