Session-Boundary Queue-Carryover Assumption Slippage Playbook

When Queue Age Is Reset but the Router Still Prices It as Intact

Why this note: Many execution stacks implicitly assume passive queue age survives across boundaries (midday breaks, auction→continuous handoffs, venue micro-reopens, broker session resets). If priority is actually reset or partially degraded, fill probability is overstated and catch-up aggressiveness arrives late—creating avoidable slippage.

1) Failure Mode in One Sentence

If your model prices post-boundary passive orders with pre-boundary queue priority assumptions, you understate implementation shortfall and delay necessary re-entry actions.

2) Extend the Objective with Boundary Priority Risk

For action (a) under context (x):

[ J(a|x)=\mathbb{E}[IS_{gross}|x,a] + \lambda,\mathrm{CVaR}_{q}(IS|x,a) + \rho,\mathrm{BoundaryQueueRisk}(x,a) ]

Where:

• (IS_{gross}): spread + impact + timing/opportunity cost,
• (\mathrm{BoundaryQueueRisk}): expected extra cost from queue-priority non-carryover,
• (\rho): penalty weight increasing near known boundary windows.

Without (\mathrm{BoundaryQueueRisk}), the router over-allocates passive flow immediately after boundaries and pays the correction cost later via urgent aggressing.

3) Minimal Boundary-Reset Model

Let:

• (p^{assumed}_{fill}): modeled post-boundary fill probability assuming carryover,
• (p^{real}_{fill}): realized fill probability,
• (\Delta p = p^{assumed}{fill} - p^{real}{fill}),
• (E_{passive}): expected passive edge (spread capture minus adverse selection),
• (C_{catchup}): later impact paid when residual inventory is forced through.

Define boundary mispricing cost:

[ C^{boundary} = \Delta p \cdot E_{passive} + C_{catchup} ]

And branch-conditional risk:

[ \mathrm{BoundaryQueueRisk}=\sum_{b\in\mathcal{B}} P(B=b|x,a)\cdot \mathbb{E}[C^{boundary}|B=b,x,a] ]

Typical branches (\mathcal{B}):

1. Full Carryover (assumption valid),
2. Priority Degrade (partial queue-age loss),
3. Hard Reset (priority fully reset),
4. Eligibility Drift (order flags/session mode changed, reducing fillability),
5. Reopen Shock (queue survives structurally but is economically irrelevant due to regime jump).

4) Telemetry Contract (Required)

A) Boundary Metadata

• boundary_type (lunch-break reopen, auction->continuous, volatility pause reopen, maintenance micro-reopen)
• boundary_start_ts, boundary_end_ts
• venue_session_state, broker_session_state
• priority_carryover_rule_version (venue/broker interpretation used by router)

B) Order-Lineage Facts

• parent_id, child_id, replace_chain_id
• pre_boundary_order_id, post_boundary_order_id
• amend_vs_cancel_replace
• queue_proxy_pre / queue_proxy_post (estimated ahead volume or rank proxy)

C) Post-Boundary Outcomes

• post_boundary_fill_prob_5s/30s/120s
• first_fill_latency_ms
• residual_after_window
• catchup_action_type and catchup_cost_bps

D) Reconciliation Fields

• assumed_carryover (bool)
• realized_carryover_class (full/degraded/reset)
• queue_carryover_error_bps
• boundary_branch_label

If boundary-state assumptions are not stored per child order, this risk is un-auditable.

5) Label Design

Create labels for training/monitoring:

1. CarryoverMismatchEvent
   • modeled carryover != realized carryover class
2. PostBoundaryFillDeficitEvent
   • realized fill probability in first N seconds is materially below model expectation
3. CatchupImpactEvent
   • residual inventory forces aggressive catch-up with elevated impact
4. BoundaryRuleDriftEvent
   • venue/broker behavior changed relative to configured carryover rule

6) Modeling Stack (Practical)

Layer A — Carryover Classifier

Estimate:

[ P(B\in{full,degraded,reset}|x,a) ]

Features: venue, symbol, boundary type, order type/flags, amend vs cancel-replace path, reopen imbalance, spread regime, queue density, volatility jump.

Layer B — Post-Boundary Fill Hazard

Model early-window fill hazard conditional on carryover class.

Layer C — Catch-Up Cost Model

Estimate impact penalty when residual inventory exceeds schedule budget after boundary.

Layer D — Boundary-Aware Policy Scorer

Rank tactics by net cost:

• passive-first if full-carryover probability is high,
• mixed tactic if degraded risk is rising,
• controlled aggressive re-entry if hard-reset probability dominates.

7) KPIs That Expose Hidden Boundary Slippage

1. Boundary Priority Survival Rate (BPSR) [ BPSR = \frac{N_{full\ carryover}}{N_{boundary\ transitions}+\epsilon} ]

2. Queue Carryover Error (QCE, bps) Difference between expected and realized post-boundary passive edge.

3. Post-Boundary Fill Deficit (PBFD) [ PBFD = p^{assumed}{fill,Ns} - p^{real}{fill,Ns} ]

4. Boundary Re-entry Impact Tax (BRIT) Incremental bps paid by catch-up actions attributable to carryover mismatch.

5. Rule-Drift Detection Lag (RDDL) Time from behavior change onset to model/config update.

6. Top-Bucket Concentration (TBC) Share of boundary slippage explained by top-k venue×boundary buckets.

8) Control Policy (NORMAL -> BOUNDARY_GUARD -> RESET_SAFE)

• NORMAL

  • standard routing weights.

• BOUNDARY_GUARD

  • activated near known boundaries,
  • raises carryover uncertainty penalties,
  • limits oversized passive exposure right after reopen.

• DEGRADED_CONFIDENCE

  • when PBFD and QCE rise,
  • use shorter passive patience windows,
  • pre-authorize selective aggressive clips.

• RESET_SAFE

  • assume no carryover until evidence proves otherwise,
  • tighten residual budget thresholds,
  • prioritize schedule recovery over nominal passive edge.

Use hysteresis + minimum dwell time to avoid oscillation around boundary transitions.

9) Rollout Blueprint

1. Backfill boundary episodes across symbols/venues with lineage joins.
2. Shadow dashboard for BPSR/QCE/PBFD/BRIT before policy changes.
3. Counterfactual replay: compare carryover-assuming vs carryover-aware policies.
4. Canary deployment on bounded notional + high-observability symbols.
5. Promotion gate: improve net slippage and schedule adherence without completion-rate degradation.

10) Common Mistakes

• Treating all session boundaries as equivalent.
• Assuming amend preserves economics when broker internally performs cancel-replace.
• Measuring only eventual completion, ignoring first-window fill deficit.
• Updating boundary rule tables manually without drift monitors.
• Reacting to residual too late (impact convexity already kicked in).

11) Fast Implementation Checklist

[ ] Persist boundary type/state and carryover rule version at decision time
[ ] Link pre/post-boundary order lineage (amend vs cancel-replace)
[ ] Label full/degraded/reset carryover outcomes
[ ] Model early-window fill hazard and catch-up impact jointly
[ ] Deploy NORMAL->BOUNDARY_GUARD->RESET_SAFE controller with hysteresis
[ ] Gate rollout on QCE/PBFD/BRIT reduction, not just gross fill rate

References

• Kissell, R. (2014), The Science of Algorithmic Trading and Portfolio Management.
• Cartea, Á., Jaimungal, S., Penalva, J. (2015), Algorithmic and High-Frequency Trading.
• Venue market-structure notices (auction/reopen/session handling) and broker session-behavior technical specs.
• SEC Rule 605 / Rule 606 datasets for execution-quality and routing diagnostics context.

TL;DR

Queue priority around session boundaries is a model assumption, not a given. Explicitly model carryover/degrade/reset branches, track post-boundary fill deficits, and switch to boundary-aware controls before residual inventory forces expensive catch-up slippage.