BL-EC-001: Cost Structure

Status: RELEASED

Classification: PUBLIC

Revision: 1

Date: 2026-09-07

1. Why this volume comes before capacity

In most research programmes cost is an implementation detail addressed after an edge is found. In this one it is the binding constraint on seven of thirty-seven settled verdicts, and it determined which classes of strategy were worth proposing at all.

The numbers below are therefore not background. They are the parameter that every cost-bound verdict in the record is conditional on, and the one a reader substitutes first.

2. Quoted schedules

Tier progression on both venues was measured and treated as fixed:

Door 4 in BL-P2-001 closed on that arithmetic and took minutes to settle.

3. Quoted is not realised

The schedule is a floor. Measured end to end, on the final cross-sectional candidate [E-0014]:

The gap between 11.0 quoted and 15.02 measured is spread paid at the moment of crossing, slippage against real depth at the intended size, and the difference between the price a model assumed and the price that was available.

A layer-zero check run against the quoted schedule is optimistic by roughly a third. That correction should be applied by anyone reusing these numbers.

4. Patience is not a cost lever

The obvious response to a taker cost of 15.02 basis points is to rest the order and wait. The programme measured what waiting costs.

The orders that fail to fill are not a random subset of the intended trades. They are the names moving away, which is precisely why they did not fill, and their drift compounds for as long as the order rests. The names that do fill are disproportionately the ones going nowhere.

Total execution cost therefore rises with the wait budget rather than falling. Extending the budget from 60 seconds to one hour makes the strategy worse.

This closed a lever the programme had expected to use, and it is the least intuitive cost finding in the record.

5. The cost identity, term by term

E[r_net]  =  E[r_gross] - c_rt
 
c_rt      =  2 * fee + spread_cost + slippage + funding_accrual

The funding term is the one most often omitted, and omitting it is not conservative. It is signed. A short position through a positive-funding settlement receives, and a long pays, so leaving the term out biases some strategies favourably and others unfavourably with no consistent direction.

One strategy in the archive had this term stubbed to zero throughout its evaluation, which overstated its return by an amount that was never fully recovered because the strategy was retired for other reasons.

6. What cost actually decided

Seven of the thirty-seven settled verdicts died on cost. Their gross effects, against the round trip they faced:

The pattern is consistent. The effects are real, between roughly half a basis point and three, and the round trip is between nine and fifteen. The programme was not failing to find signal. It was finding signal an order of magnitude below what it cost to act on.

That is a statement about the operator, not about the market, and it is why every one of these rows is classified ENVELOPE-BOUND.

7. The reversal points, collected

For a reader with a different schedule, the cost-bound rows and the round trip each would need:

Only one of the four is within reach of a schedule that exists, and even that one requires a passive entry leg on a burst-triggered signal, which is discussed in BL-SD-029 section 10.

Publishing this table matters more than publishing the verdicts. A reader can see immediately that three of these are dead at any plausible schedule, and that one is a genuine question for an operator with a rebate tier.

8. What was not measured

Market impact. No strategy in this programme ran at a size where its own transacting moved the book it was measuring against, so impact does not appear in the cost identity and is not estimated anywhere.

An operator running at size should treat every cost figure in this record as a floor that omits their own footprint.

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