BL-SD-016: Cash and Carry, Delta Neutral

Status: RELEASED

Classification: PUBLIC

Revision: 1

Date: 2026-09-07

Venue: Bybit Programme: I Verdict: CLOSED Axis of death: Venue access Transferability: ENVELOPE-BOUND

1. Identification

2. Economic thesis

Long a coin's spot and short its perpetual in equal notional cancels price exposure. What remains is the funding cashflow: when the perpetual trades at a premium and funding is positive, the perpetual's long side pays the short side at each settlement, and the carry trader, sitting short the perpetual and long the cancelling spot, collects that payment while bearing approximately zero net directional risk.

Who is forced to lose. The payer side of a persistently positive funding rate: the crowded leveraged-long side of a perpetual trading at a premium to its own spot. That side is structurally compelled to pay for as long as it wants to hold leveraged long exposure across a settlement, and it cannot decline the payment while the position is open. This is a real, if mild, structural transfer, distinct from a forced-liquidation mechanism but genuine in the sense that it does not depend on the payer being wrong about price, only on the payer wanting the leverage.

3. Falsifiable claim

The evaluation was framed as a build decision rather than a classical kill-criterion pre-registration, and is reproduced here in that form:

The monitor's positive-funding fire, net of a realistic round-trip cost and amortized over a 30-day hold, identifies carry that is both net positive and realizable, meaning executable within Bybit's own spot and perpetual markets on a single venue, at a rate that clears an approximately 10 percent annualized bar that would justify building spot execution infrastructure.

4. Mathematical specification

Notation per BL-D-011. Bybit funding settles every 8 hours, 1095 periods per year.
 
basis_bps(perp, spot)             = (perp - spot) / spot * 1e4
 
annualized_funding_bps(mean_f_8h) = mean_f_8h * 1095 * 1e4
 
amortized_roundtrip_annual(c_rt, hold_days) = c_rt * (365.0 / hold_days)
 
net_carry_bps_annual  = annualized_funding_bps(mean_f_8h)
                         - amortized_roundtrip_annual(c_rt, hold_days)
 
entry: fire when the ROLLING MEAN of realised funding over a trailing warmup
       of at least 48 samples is positive and debounced (a sustained level,
       not a single-sample spike)
 
position: long the spot leg, short the perpetual leg, equal notional, delta neutral
 
hold: amortize the round-trip cost over an assumed 30-day hold
 
cost: c_rt = 35 bps stated round trip

Causality holds for the entry rule: the rolling-mean funding signal at any decision time is computed from settlements observed at or before that time, with a warmup requirement that excludes an undersampled estimate from firing, and a debounce that requires the sustained level rather than the latest sample alone, so the signal is measurable with respect to F_t rather than reacting to a single noisy print.

The source states the round-trip cost as 35 basis points, composed of a spot leg at 10 basis points per side, taken twice, and a perpetual leg at 5.5 basis points per side, taken twice. Those two components as stated sum to 31 basis points. The source does not reconcile the 4 basis point difference against the stated 35, and this record reports the figure as given rather than silently correcting it to the component sum.

5. Operating envelope

6. Instrument

The monitor logic: rolling-mean sustained funding rather than a snapshot rate, a warmup floor of 48 samples, debounce, and a positive-funding-only fire condition. A companion module held the pure yield mathematics, unit-pinned to 1095 funding periods per year, the single figure its own documentation names as the primary unit-risk in this calculation.

The evaluation itself was a read-only decomposition, not a new statistical test: every symbol the monitor had fired on was checked against a stock-and-commodity exclusion filter, which the funding-rate universe did not apply at the source, and against whether a Bybit USDT spot market existed for that symbol, the precondition for executing the long leg at all on this venue.

7. Results

8. Verdict, axis of death, and transferability

Verdict: CLOSED. Axis: venue access. Transferability: ENVELOPE-BOUND.

The method carries no defect. The rolling-mean construction, the warmup floor, the debounce, and the realistic 35 basis point round trip amortized over a 30-day hold are all sound, and the monitor's fire is not a false-positive of its own logic. What defeats it is that the fire is concentrated in names this venue cannot execute on a single-venue, spot-plus-perpetual basis: thirteen of fourteen net-positive symbols have no Bybit spot market, so the long leg the strategy requires does not exist here.

BL-D-002 section 4 asks whether the result would be positive under arbitrarily favourable envelope parameters. Here it would be: funding dispersion is a measured, real phenomenon elsewhere in this record, at roughly 15 basis points per 8 hours across a 14-symbol cross-section, and an operator with the venue access to execute the currently unreachable names would face a materially different answer. That is precisely the definition of envelope-bound rather than universal.

A second, independent finding belongs alongside the verdict without changing it. Even where cash and carry is executable, it is structurally a rate collected rather than an alpha generated: a delta-neutral coupon that compounds at single-digit annual percentages under normal conditions, roughly a nine-year doubling period at 8 percent per year, and one whose available rate falls as capital grows, because scale forces migration into thinner, lower-yield, spot-unavailable names, and a large enough short itself compresses the funding it is trying to collect. This is not the axis that closed the strategy, but it is the reason the record does not recommend reviving it even where venue access improves, except as a treasury overlay rather than a growth engine.

9. Cost and capacity analysis

Round-trip cost, stated                     35 bps
Amortized annually over a 30-day hold       35 * (365 / 30) is approximately 426 bps per year
Best realizable candidate, net annualized   +2.7 pct

The one executable, clean, and liquid candidate cleared its cost after amortization and still landed well short of the roughly 10 percent annual bar the evaluation set for justifying execution infrastructure. Capacity was addressed qualitatively rather than measured numerically: as working capital grows, the strategy is forced into progressively thinner names where spot liquidity is scarcer and where the carry trader's own short position compresses the funding rate being collected, so the achievable rate falls as assets under management rise. No specific capacity ceiling in USD is given in the source.

10. Envelope sensitivity and reopening conditions

Binding parameter: venue access, specifically Bybit spot-market availability for the funding-rich names identified by the monitor.

Reversal condition:  a venue, or a cross-venue execution capability, where the
                      13 currently unexecutable symbols carry a spot market,
                      which changes the realizable set directly.

A second, regime-dependent reopening condition is stated honestly in the source rather than omitted because it complicates a clean verdict. In euphoric bull regimes, funding on the liquid majors, which do have Bybit spot markets, has been observed to spike to 20 to 50 percent annualized for periods of months. That window is real, executable today on this venue, and transient: it compresses as carry traders arbitrage it back toward cost, so it does not change the structural verdict in section 8, but it does mean the reversal point for this specific hypothesis is regime-conditional rather than fixed.

What would reopen it: cross-venue spot access to the 13 currently unreachable names, or a sustained bull-regime funding spike on the majors, evaluated through the same monitor with an added executable-carry gate: crypto only, a confirmed Bybit USDT spot market, and a liquidity floor of approximately 20,000,000 USD, none of which was built here because building it was judged effort spent on a non-priority sleeve.

What would not reopen it: a lower assumed round-trip cost alone. The binding constraint is which symbols can be executed at all, not the cost of executing the ones that can.

11. Reference implementation

Published as 05_INSTRUMENTS/strategies/python/src/cash_and_carry.py, with 20 tests. The yield mathematics, the fire condition, and the spot-availability screen that carried this verdict are all in it; the runner and its scheduling are not, being transport. The monitor was stopped rather than deleted; its container and its historical archive were both retained.

12. Evidence register

13. Lessons extracted

The contamination finding in section 7, tokenized stocks and commodities present in an ostensibly crypto funding-rate universe, is the same defect class recorded in BL-IN-004: universe hygiene at the source, not at the detector, and it produced the recommendation for an executable-carry gate analogous to the liquidity-floor and denylist discipline R-19 already established.

The structural finding in section 8, that a successfully executed carry strategy is a linear yield sleeve rather than a compounding growth engine, was the basis for redirecting the programme's growth-oriented research toward scalable directional and structural-flow candidates instead, a redirection that led directly to the design of BL-SD-017.

14. Cross-references

BL-SD-017 for the growth-oriented redirection this evaluation motivated. BL-IN-004 and rule R-19 for the universe-hygiene defect class shared with this hypothesis. BL-D-002 section 4 for the envelope-substitution test applied in section 8. BL-SD-000 ledger row CARRY.

Copyright (c) 2026 Bijective Labs. All rights reserved.