BL-IS-011: Spread Census

Status: RELEASED

Classification: PUBLIC

Revision: 1

Date: 2026-09-09

1. What it measures

Quoted spread, in basis points and in exchange ticks, across every tradeable perpetual on the venue, from three time-separated live order-book reads aggregated by median.

spread_bps(bid, ask)              = (ask - bid) / mid * 1e4
spread_ticks(bid, ask, tick_size) = (ask - bid) / tick_size

Reporting both units in the same row matters. A wide spread in basis points can mean a book with genuine room to quote inside it, or it can mean a symbol whose minimum price increment is itself worth several basis points, in which case no quote can legally sit inside the touch. The two situations have opposite consequences and are indistinguishable from the bps figure alone.

2. Why it exists

Market making on this venue was closed once, on the strength of seven symbols. The recorder that produced that sample captured order-book snapshots only around liquidation events, which means only symbols active enough to trigger a liquidation ever entered it. That selects toward the most liquid names in the universe, which are also the narrowest-spread names. The sample was not merely small; it was drawn in the direction of the conclusion it produced. Its median was 1.31 basis points. Two of the seven had in fact cleared the maker round-trip threshold, DEXE at 4.80 and LAB at 6.53, and were discarded as illiquid, which is precisely the condition under which passive market making is expected to work at all.

This instrument answers the question the seven-symbol sample could not: what does the full universe actually quote. Measured across all 676 instruments trading at the time, the median corrected to 5.26 basis points, roughly four times the earlier figure [E-0051]. The correction did not change which door it closed on. It changed why the door was worth reopening in the first place.

3. Three disciplines the instrument enforces

The maker fee is reported as a grid, never assumed. The constant this door's verdict depends on entirely was marked unverified at the time of measurement: absent from every configuration file in the repository, and unreachable because the venue's fee-rate endpoint requires authentication the programme does not hold. Every result is therefore computed across a fee grid rather than at one asserted number.

Capacity is measured beside spread, not separately. A thirty basis point spread on a book that holds two hundred dollars of depth is not an opportunity; it is a number. Best-quote notional is recorded for every symbol precisely so the two questions cannot be answered independently and then silently recombined.

Three passes, median aggregation. A single snapshot is one instant. Spread on this venue widens and narrows inside seconds, and a median across three passes, forty-five seconds apart, discards a single-snapshot outlier without assuming a distribution to correct it against.

4. Interface

spread_census(universe, snapshots, tick_sizes, maker_fee_grid) -> SpreadCensusReport
 
  universe        every Trading, linear, USDT perpetual, non pre-listing
  snapshots       three independent L1 order-book reads, 45 s apart
  tick_sizes      venue-declared minimum price increment, per symbol
  maker_fee_grid  candidate maker fee per fill; the true value is unverified
 
Returns per symbol: median spread in bps across passes, spread in ticks,
best-quote depth in USD, 24-hour turnover, open interest.
 
Returns per fee in the grid: count and percentage of symbols net-positive,
count with two or more ticks of quoting room, count with depth above the
minimum gate, median net spread of the positive set, an upper-bound revenue
pool at assumed shares of 24-hour volume captured.

5. Specification

median_spread(sym)  = median_{k=1..3}( spread_bps_k(sym) )
room(sym)           = spread_ticks(sym) >= 2.0
tick_bound(sym)      = spread_ticks(sym) <= 1.0
tradeable(sym)       = depth_usd(sym) >= 1,000
net_bps(sym, fee)    = median_spread(sym) - 2 * fee

Cross-checked against a second, independent venue endpoint on every symbol; the two disagree by a median relative error of 0.23 percent, which is the basis for trusting the order-book read rather than a coincidence.

6. Results

Full universe, three passes, 676 of 676 symbols measured [E-0051]:

Mean 8.077 bps. Spread in ticks: median 3.0, with 37.9 percent of the universe at one tick or less (mechanically unquotable inside the touch) and 62.1 percent at two ticks or more (genuine room to quote). At a maker fee of 2.0 bps per fill, round trip 4.0, 412 of 676 symbols, 60.95 percent, quote a spread above that cost by median alone, before any adverse selection is considered.

7. Two caveats the artifact carries that the original narrative did not

The maker fee is a grid, not a fact. The 2.0 bps figure used above is one point on that grid, chosen because it matched a prior internal assumption, and that assumption itself carries no confirmation anywhere in the repository. Every percentage in section 6 that depends on a fee should be read as conditional on that fee, not as a statement about what any specific operator would actually net.

Depth clears the minimum gate on 33 of 676 symbols. The 62.1 percent with two ticks of room and the 60.95 percent net-positive at fee 2.0 both describe what the universe quotes. Only 33 instruments cleared even a thousand-dollar depth gate at the best quote. For the remaining 643, quoted width and available liquidity are separate, unanswered questions: a wide, net-positive quote on a book with negligible depth behind it is not evidence of an opportunity, and this census cannot by itself tell the two apart.

8. What it does not do

It is a snapshot taken on a single day, not a measurement over the 2022 to 2025 research era, and it cannot be used to backtest anything against that era. It answers one question only: whether any part of today's universe quotes wider than today's maker round trip.

It says nothing about adverse selection. A wide, liquid, net-positive quote is a candidate for the next gate, not a verdict; whether the price moves against a maker after the fill is a separate measurement, on a separate instrument, over a separate horizon.

It does not resolve which of the 643 depth-ungated symbols are actually tradeable. That requires a liquidity study built for the purpose, which this instrument is not.

9. Cross-references

BL-P2-001 section 4, door 3. BL-SD-046 for the two-gate closure this census fed as the first gate. BL-SD-000 section 7 for the corrected figures against the original narrative. BL-EC-005 for the universe-size context, 676 instruments as of 2026-08-01 against 570 nine weeks earlier. Evidence E-0051.

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