Stealth Execution

Impact-aware execution algorithms: adaptive TWAP, randomized iceberg, sniper mode, and smart cross-exchange routing.

Ultra Feature. Requires an Ultra subscription. Get started at api.mathematicalcompany.com

Stealth Execution

Institutional-grade execution for prediction markets. Estimate market impact before trading, route across venues, and execute with algorithms that minimise information leakage.

Impact Estimation

Kyle’s lambda + spread + volatility to predict slippage before you trade.

Smart Routing

Greedy cost-minimising allocation across Paper, Polymarket, and Kalshi.

Adaptive TWAP

Fill-rate adaptive time-slicing with spread-widening circuit breaker.

Sniper Mode

Waits for tight spread + favourable imbalance, then fires instantly.

---

Impact Estimation

import horizon as hz

engine = hz.Engine()
impact = hz.estimate_impact(engine, "0xcondition...", size=500)
print(impact.estimated_slippage_bps)   # e.g. 12.5
print(impact.recommended_strategy)      # "twap"

Field	Description
estimated_slippage_bps	Predicted slippage in basis points
kyle_lambda	Kyle’s lambda (price impact coefficient)
effective_spread_bps	Current effective spread
recommended_strategy	"market", "twap", "iceberg", or "sniper"
estimated_duration_secs	Recommended execution window

---

One-Call Execution

plan = hz.stealth_execute(
    engine, "0xcondition...",
    side=hz.Side.Yes,
    order_side=hz.OrderSide.Buy,
    size=500, price=0.55,
    strategy="auto",  # auto-selects based on impact
)
print(plan.strategy, plan.estimated_cost_bps)

---

Pipeline Mode

hz.run(
    name="stealth-mm",
    markets=["0xcondition..."],
    pipeline=[
        hz.stealth_executor(config=hz.StealthConfig(
            patience_secs=60, randomize=True,
        )),
    ],
    interval=1.0,
)

The pipeline reads ctx.params["stealth_requests"] (a list of request dicts) and drives active algorithms each tick.

---

Algorithms

AdaptiveTWAP

Splits a parent order into time-slices with fill-rate feedback:

• Fill rate > 80% → slow down (multiply interval by 1.5)
• Fill rate < 20% → speed up (multiply interval by 0.75)
• Spread widens > 2× → pause until it narrows
• Optional ±20% slice size randomisation

IcebergPlus

Shows a random fraction of the total order:

• Visible size randomised in [0.5×, 1.5×] of base
• Capped at max_visible_pct of remaining quantity
• Random delay between refills

SniperAlgo

Waits for optimal entry conditions:

• Monitors effective_spread and lob_imbalance
• Fires when spread < target AND imbalance favours our side
• Patience mechanism progressively relaxes the threshold

---

StealthConfig

Parameter	Default	Description
randomize	True	Randomise slice sizes and delays
min_delay_secs	0.5	Minimum delay between iceberg refills
max_delay_secs	5.0	Maximum delay between iceberg refills
max_visible_pct	0.10	Maximum visible fraction (iceberg)
impact_limit_bps	50.0	Abort if estimated impact exceeds this
patience_secs	120.0	Sniper patience window
adaptive_aggression	True	Enable fill-rate adaptive behaviour