Research note 016: the inverse fails in its best month, before borrow
Date: 2026-10-16 · Status: one month, hourly bars, stated borrow range. Fills the empty cell in the carry matrix: every carry note so far ran long-spot/short-perp. Short-spot/long-perp collects funding when rates go negative — November 2022, 32 of 90 payments negative, is the kindest possible window. Same legs, fees, and fills; borrow added as an explicit sensitivity (0/2/5/10% APR over the 30-day hold) because short spot pays borrow and the rate is an assumption, not data.
Question
Does the inverse survive where the standard direction cannot — a negative-funding month — once borrow is priced?
Dataset and method
- November 2022 funding (90 payments) plus hourly closes, spot
$20,456.57 → $17,163.64 through the FTX collapse aftermath. Three
slices pinned by extracted-CSV SHA-256 in
research/scripts/inverse_016.py(stdlib only, one command). - Fees: 4 fills at 5 bps taker = $20.00. Basis drift signed for the inverse direction (gains when the discount narrows). Borrow pro-rated over exact days held at each stated APR.
Results
| Measure | Value |
|---|---|
| Funding received (inverse) | +$17.37 |
| Fees | $20.00 |
| Basis drift PnL | +$0.41 (discount −$11.17 → −$10.34) |
| Net at 0% borrow | −$2.23 |
| Net at 2% / 5% / 10% borrow | −$18.67 / −$43.32 / −$84.42 |
Reading
Two facts, and the second is redundant by design. First, the inverse fails before borrow exists: $17.37 of the kindest funding on record against $20 of unavoidable fees nets −$2.23 with borrow at zero. No rate assumption is load-bearing because the position is already underwater without one — the sensitivity table exists to show that borrow only deepens a hole fees already dug. Second, the matrix is now complete and asymmetric: long/short survives hot funding at 28-to-1, the inverse dies in cold funding at any borrow rate including zero. Carry is a one- directional trade at retail size, and the direction is whichever side funding pays.
What would strengthen or overturn this: a deeper negative-funding regime (the sample is one month — though it is the most negative month in the two years surveyed); maker fills halving the $20 floor (the fee boundary again, as everywhere in this series); and the live-captured version, where fills stop being assumed.
Reproduce
python3 research/scripts/inverse_016.py --data-dir /tmp/inv016
Three CSV content hashes pinned in the script, refused on any mismatch — verified by a from-scratch rerun with identical figures.