Slippage Jump-Risk in News Windows: A Practical Playbook

Why this matters

Average slippage models fail exactly when it hurts most: around scheduled/unscheduled news. In these windows, spread, depth, and impact parameters move discontinuously (jump risk), so static execution settings become expensive fast.

Objective

Build a news-window aware execution policy that:

1. Detects event-risk regime in real time,
2. Shrinks participation and aggressiveness before liquidity cliffs,
3. Restores speed only after microstructure stabilizes,
4. Measures whether policy reduced tail slippage, not just average bps.

1) Regime definition (Normal / Event-Watch / Event-Shock)

Use a compact state machine:

• Normal: baseline spread/depth/imbalance behavior.
• Event-Watch: known macro/earnings window approaching or elevated pre-event signals.
• Event-Shock: post-release dislocation detected (spread jump + depth collapse + high short-horizon volatility).

Transition triggers (example):

• t in [event_time - 10m, event_time + 20m] => at least Event-Watch.
• spread_z > 3 AND top_depth_z < -2 AND rv_1m_z > 2 => Event-Shock.
• Require persistence (e.g., 20-30s) before downshifting from Shock to Watch.

2) Slippage model extension with jump component

Baseline: slippage = spread_cost + temp_impact + delay_cost + opportunity_cost

Event extension: slippage_event = baseline + jump_term

Where jump_term is conditional on regime and captures abrupt repricing risk:

• Event-Watch: small positive premium
• Event-Shock: large premium tied to realized micro-vol + depth scarcity

Practical parameterization:

• jump_term = a_r + b_r * rv_1m + c_r * (1/depth_top) where coefficients depend on regime r.
• Fit with robust regression / quantile regression (p70, p90) to focus on tail control.

3) Execution controls by regime

Normal

• Standard POV and urgency.
• Passive-first if toxicity is low.

Event-Watch

• Reduce POV cap (e.g., 30-50% cut).
• Tighten max child size.
• Expand minimum refill interval.
• Cancel stale quotes faster.

Event-Shock

• Hard cap on aggression and marketable size.
• Prefer slicing with stricter pause/re-evaluate gates.
• If liquidity is one-sided and markout deteriorates, trigger temporary participation freeze.
• Allow kill-switch when projected residual cost exceeds trade thesis edge.

4) Pre-event checklist (operational)

Before each high-impact event:

• Confirm event schedule + symbol exposure map.
• Load event-specific caps (POV, child size, max sweep).
• Arm automated Shock detector.
• Predefine “resume-normal” conditions.
• Assign explicit fallback (skip/halve/continue) if detector or feed fails.

5) Monitoring metrics that actually matter

Track separately for Normal/Watch/Shock:

• Slippage median, p90, p95
• Adverse markout (30s, 5m)
• Fill rate and underfill ratio
• Budget-breach frequency
• Time spent in each regime

Key KPI:

• Tail Slippage Reduction: p95 in Watch/Shock vs prior policy
• Cost of Defense: extra underfill/opportunity cost from defensive throttling

Goal is better tail survival at acceptable opportunity trade-off.

6) Calibration loop (weekly)

1. Re-label event windows from realized data.
2. Re-fit jump coefficients by regime.
3. Re-test thresholds for false-positive/false-negative balance.
4. Evaluate policy frontier:
   • x-axis: underfill/opportunity cost
   • y-axis: p95 slippage
5. Select parameters on efficient frontier, not single best average.

7) Common implementation mistakes

1. Using one global slippage coefficient for all times.
2. Optimizing mean slippage only while tails explode.
3. No hysteresis in state transitions (policy flaps).
4. Missing fail-safe path when event feed is delayed.
5. No post-event re-acceleration logic, causing unnecessary underfills.

Minimal rollout plan

• Week 1: Regime labeling + monitoring only (no control changes).
• Week 2: Enable Event-Watch throttles.
• Week 3: Enable Event-Shock hard guards with manual override.
• Week 4: Tail-focused review and threshold tuning.

This keeps rollout reversible while quickly reducing “one bad print ruined the day” outcomes.

Bottom line

News-window execution is not a faster/slower toggle. It is a regime-switching risk problem. If you price jump risk explicitly and bind execution controls to regime states, you usually sacrifice a bit of average speed to gain a lot of tail robustness—which is the trade that keeps compounding alive.