BL-SD-003: Directional Alpha, Taker Entry with 8/10 bps Brackets

Status: RELEASED

Classification: PUBLIC

Revision: 1

Date: 2026-09-07

Venue: Bybit perpetuals Programme: I Verdict: FALSIFIED Axis of death: Cost Transferability: ENVELOPE-BOUND

1. Identification

Named Variant B at design time, distinguishing it from a rejected pure-maker pitch (Variant C) and a rejected pure-taker scalp (Variant A) considered in the same review. It is the design the programme actually shadow-tested for a full 24-hour window.

2. Economic thesis

Order-flow imbalance over the top five book levels, taken above a 2.0 sigma threshold, predicts short-horizon directional continuation with an empirically high hit rate at a 100 millisecond measurement horizon. The design widens the exit brackets to 8 basis points take-profit and 10 basis points stop-loss and extends the hold to a 30-second timeout, on the reasoning that a Taker entry capturing the signal immediately, combined with a Maker exit at the profit target, halves the round-trip cost relative to crossing on both legs.

Who is forced to lose. The passive counterparty resting on the side the order-flow imbalance is about to sweep, the same structural loser named across this family of hypotheses. The mechanism was not in question here; the magnitude available to capture it at this horizon was.

3. Falsifiable claim

Pre-registered before the shadow window began:

Directional alpha entries at Taker, exiting via Maker take-profit or Taker stop-loss/timeout, produce a shadow win rate at or above 60 percent, with a median time-to-take-profit under 15 seconds and no per-symbol stop-loss rate above 30 percent, sufficient to clear the round-trip cost.

4. Mathematical specification

Notation per BL-D-011.
 
Signal:
 
  s_t  =  ofi_l5_norm(t)      z-score of level-5 order-flow imbalance,
                               warmup-gated
 
Entry:
 
  fire at t when  |s_t| >= 2.0  AND  is_warmed_up(t) = true
  cooldown: 10,000 ms per symbol since last exit
  direction  d_t = sign(s_t)
  order type: Market IOC (taker)
  entry_mid = (bid + ask) / 2  at fill
 
Exit, path-dependent, first touch wins:
 
  tp_price  =  entry_mid + d_t * entry_mid * 8  / 10000     (PostOnly, maker)
  sl_price  =  entry_mid - d_t * entry_mid * 10 / 10000     (Market IOC, taker)
  timeout: 30,000 ms, Market IOC reduce-only if neither touched
 
Sizing:
 
  notional_usd = 5.0     per entry, max_concurrent = 2 cross-symbol
 
Cost, branch by outcome, per source assumption:
 
  c_rt(tp_hit)      =  6.0 (taker entry) + 1.0 (maker exit) = 7.0 bps
  c_rt(sl_hit)      =  6.0 (taker entry) + 6.0 (taker exit) = 12.0 bps
  c_rt(timeout)     =  6.0 (taker entry) + 6.0 (taker exit) = 12.0 bps
 
Net return per event:
 
  r_net(t) = d_t * (exit_mid - entry_mid) / entry_mid * 1e4  -  c_rt(outcome)

Causality holds by construction: s_t and the warmup flag are read from state at or before t, the cooldown check reads only the timestamp of the prior exit, and the exit rule evaluates price against fixed brackets set at entry rather than against any information from after t.

5. Operating envelope

6. Instrument

The directional-alpha strategy module, with pure, independently testable functions for entry decision, position construction, and outcome checking, run in shadow mode against live market data with no order submission. A dedicated offline aggregation tool performed the shadow analysis. This is a properly instrumented measurement rather than a hand-reconciled one.

7. Results

Direction was confirmed positive and consistent, at a magnitude the exit brackets and cost structure could not monetise.

8. Verdict, axis of death, and transferability

Verdict: FALSIFIED. Axis: cost. Transferability: ENVELOPE-BOUND.

The pre-registered kill criterion, p_tp at or above 60 percent, was missed by 25 points on a sample of 7,916, which is a falsification with power rather than an underpowered null. The mean realised move of +0.78 basis points confirms the underlying order-flow direction is real; the 8/10 basis point brackets simply require a magnitude the signal does not carry at this horizon.

Applying the test in BL-D-002 section 4: at a sufficiently low round-trip cost, a mean edge of +0.78 basis points produces a positive expectation, so the classification is envelope-bound rather than universal. The magnitude itself, not the direction, is what a cheaper channel would need to rescue.

9. Cost and capacity analysis

Mean realised move (gross)       +0.78 bps
EV per trade (fee-adjusted)      -6.22 bps
Implied round-trip cost used     approximately 7.0 bps blended
Projected daily return           -24.6 pct on $100 notional basis

Capacity was not the binding question; frequency at 7,916 trades in 24 hours was more than sufficient for a shadow sample, and the strategy was closed on unit economics rather than on throughput.

10. Envelope sensitivity and reopening conditions

Binding parameter: round-trip cost against a sub-basis-point gross edge.

Gross edge, mean realised          +0.78 bps
Reversal condition                 c_rt < 0.78 bps

No channel this venue publishes reaches that level. The declared maker-to-maker round trip of 4.0 basis points (BL-FM-002 section 3.1) leaves a net of -3.22 basis points; taker-to-taker at 11.0 and maker-to-taker at 7.5 fail by a wider margin. The mechanism is envelope-bound by the strict test in section 8, since an arbitrarily favourable round trip would clear it, but no reachable Bybit tier gets within an order of magnitude of doing so.

This finding is grouped in the Kill Ledger's envelope-bound register together with BL-SD-004, whose measured gross edge of +0.64 basis points sits in the same sub-basis-point range against the same 11.0 basis point taker-to-taker floor. Both are the same underlying finding at two different bracket and threshold configurations: the order-flow signal is real, and its magnitude at short horizons does not clear Bybit's non-VIP fee schedule at any declared channel.

11. Reference implementation

Published as 05_INSTRUMENTS/strategies/rust/src/directional_alpha.rs, with the 20 test cases that pinned it in the source. No defects were found in the entry, position-construction, or outcome-check functions during this measurement; the shadow result reflects the mechanism and cost structure as designed. The async run loop and its cache transport are not reproduced.

12. Evidence register

This measurement is recorded in narrative decision documents; registration as a standalone document-evidence artifact is scheduled per BL-SD-000 section 8.

13. Lessons extracted

This is the measurement that established the magnitude-versus-fee-floor pattern that recurs across BL-SD-004 through BL-SD-008: order-flow direction at Bybit Tier 1.5 altcoin is real and consistent, and its typical magnitude at horizons under one minute sits below one basis point, which is structurally smaller than the venue's cheapest reachable round trip. No single dossier in this family should be read as an isolated failure; they are five measurements of the same ceiling from different angles.

14. Cross-references

BL-SD-001 for the predecessor Taker configuration. BL-SD-004 for the recalibrated variant tested one cycle later on the same fee-floor problem. BL-D-002 section 4 for the envelope classification test. Ledger row DA-B.

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