Per-Asset-Class Risk

Different rules for equities, options, perps, and prediction markets

Different asset classes have different risk profiles. A 10% stop makes sense for an equity position but is way too wide for an options position and misleading for a prediction market where the native concept is “days to resolution” not “unrealized P&L.” per_asset_class lets you configure each class with its own AssetRisk bundle.

The dataclass

python
@dataclass
class AssetRisk:
    # Stops for this asset class
    stop_loss: Any = None
    time_stop: Any = None

    # Generic sizing caps
    max_gross_notional_usd: float | None = None
    max_position_pct_of_adv: float | None = None
    max_position_notional_usd: float | None = None

    # Equity-specific
    max_sector_exposure: float | None = None
    exclude_earnings_within_days: int | None = None

    # Option-specific
    max_portfolio_delta: float | None = None
    max_portfolio_gamma: float | None = None
    max_portfolio_vega: float | None = None
    max_theta_burn_per_day_usd: float | None = None

    # Perp-specific
    max_leverage: float | None = None
    liquidation_buffer_pct: float | None = None
    max_funding_exposure_per_day_usd: float | None = None

    # Prediction-market-specific
    max_per_event_usd: float | None = None
    min_days_to_resolution: int | None = None

Usage

python
from horizon.asset_classes import Equity, Option, Perp, Prediction
from horizon.risk import (
    AssetRisk,
    GreekStop,
    RiskConfig,
    StopLoss,
)

risk = RiskConfig(
    per_asset_class={
        Equity: AssetRisk(
            stop_loss=StopLoss(per_position_pct=0.10, trailing_pct=0.05),
            max_sector_exposure=0.40,
            exclude_earnings_within_days=3,
        ),
        Option: AssetRisk(
            stop_loss=GreekStop(
                close_if_delta_over=0.70,
                close_if_gamma_over=0.08,
                close_if_dte_under=2,
            ),
            max_portfolio_delta=1000,
            max_portfolio_gamma=50,
            max_portfolio_vega=500,
            max_theta_burn_per_day_usd=200,
        ),
        Perp: AssetRisk(
            stop_loss=StopLoss(per_position_pct=0.08),
            max_leverage=3.0,
            liquidation_buffer_pct=0.30,
            max_funding_exposure_per_day_usd=50,
        ),
        Prediction: AssetRisk(
            stop_loss=StopLoss(per_position_usd=500),
            max_per_event_usd=2_000,
            min_days_to_resolution=3,
        ),
    },
)

Equity

Option

Perp

Prediction

Combining with global risk

Per-asset-class rules live alongside the global risk layers:

python
risk = RiskConfig(
    # Global. applies across all asset classes
    per_order=OrderRisk(max_order_notional_usd=25_000),
    drawdown=[DrawdownGuard(daily_pct=0.05, action=HaltNew)],
    kill_switch=KillSwitch(auto_trigger_on=[PortfolioDrawdownEvent(0.25)]),

    # Per asset class. applies only to that class
    per_asset_class={
        Equity: AssetRisk(stop_loss=StopLoss(per_position_pct=0.10)),
        Option: AssetRisk(stop_loss=GreekStop(close_if_delta_over=0.70)),
    },
)

Both run. Per-order checks apply to every order regardless of class; drawdown guards apply to total equity regardless of composition; kill switch fires on total drawdown. The per-class stops fire only on positions in the matching class.

When the strategy is multi-asset

If your strategy trades multiple asset classes and you want different rules for each, per_asset_class is the cleanest way to express it. Without it, you’d need separate risk configs per strategy, which wouldn’t work because risk is evaluated at the portfolio level.

Next

Risk profiles Conservative / moderate / aggressive presets. Back to overview The 7 layers in one page.