risk-threshold-engine 0.2.0

Factor-score-driven market risk evaluation: composite score, regime classification, and deployment decisions.

risk-threshold-engine

A lightweight, zero-dependency Python library that converts factor scores into a market risk regime, deployment decision, and recommended mitigation actions.

Overview

Feed in five 0–100 risk scores (volatility, momentum, breadth, macro, drawdown), get back a structured result telling you whether to deploy capital at full pace, reduce lot sizes, pause, or rebalance.

from risk_threshold_engine import RiskThresholdEngine, FactorScores

engine = RiskThresholdEngine()

scores = FactorScores(
    volatility_score=72.0,
    momentum_score=60.0,
    breadth_score=55.0,
    macro_score=65.0,
    drawdown_score=50.0,
)

result = engine.evaluate(scores)
print(result.summary())
# [2026-05-01T...] Risk Engine Result
#   Composite Score : 61.25 / 100
#   Regime          : ELEVATED
#   Decision        : REDUCED_DCA
#   ...

Installation

pip install risk-threshold-engine

Or install from source:

git clone https://github.com/BrainScanStudios/risk-threshold-engine.git
cd risk-threshold-engine
pip install -e .

Requirements

Python 3.10+. No external dependencies — only the standard library.

Risk Regimes

Composite Score	Regime	Default Decision
0 – 30	LOW	FULL_DCA
31 – 55	MODERATE	REDUCED_DCA
56 – 75	ELEVATED	REDUCED_DCA / PAUSE*
76 – 100	SEVERE	REBALANCE

*ELEVATED switches to PAUSE when the portfolio's cash/bond buffer exceeds 30%.

Default Factor Weights

Factor	Weight	What it captures
volatility_score	25%	Realised vol / VIX signal
momentum_score	20%	Price momentum across assets
breadth_score	15%	Market breadth (advance/decline)
macro_score	20%	Rates, credit spreads, inflation
drawdown_score	20%	Drawdown vs high-water mark

All weights are configurable. Scores are 0–100 where higher = more risk.

API

RiskThresholdEngine

RiskThresholdEngine(
    factor_weights: dict[str, float] | None = None,
    ticker_beta:    dict[str, float] | None = None,
    db_path:        str | None = None,
)

• factor_weights — Override the default weight map. Must sum to 1.0.
• ticker_beta — Map of ticker → beta coefficient. When provided, the engine generates TRIM and OPTIONS_SIGNAL actions for high-beta tickers held in the portfolio during elevated/severe regimes.
• db_path — Optional path to a SQLite file. When set, every evaluate() call persists results to a risk_engine_results table.

evaluate(scores, portfolio=None) -> EngineResult

Runs the full evaluation pipeline.

FactorScores

FactorScores(
    volatility_score: float,   # 0–100
    momentum_score:   float,   # 0–100
    breadth_score:    float,   # 0–100
    macro_score:      float,   # 0–100
    drawdown_score:   float,   # 0–100
)

All five scores are required and must be in [0, 100]. How you produce them is up to you — any factor scoring module works.

PortfolioState (optional)

PortfolioState(
    buffer_pct:   float,              # cash/short-term bond % of portfolio
    cash_usd:   float,
    positions:  dict[str, float],   # ticker → market value in USD
)

Passing a PortfolioState enables portfolio-aware decisions (buffer checks, beta-weighted trim candidates, options signals).

EngineResult

Field	Type	Description
composite_score	float	Weighted composite (0–100)
regime	RiskRegime	LOW / MODERATE / ELEVATED / SEVERE
decision	DeploymentDecision	FULL_DCA / REDUCED_DCA / PAUSE / REBALANCE
actions	list[MitigationAction]	Ordered list of recommended actions
weighted_breakdown	dict[str, float]	Per-factor contribution to composite
notes	str	Contextual caveats
evaluated_at	str	ISO 8601 UTC timestamp

result.to_dict()    # fully serialisable dict
result.summary()    # human-readable string

Custom Configuration

engine = RiskThresholdEngine(
    factor_weights={
        "volatility_score": 0.40,
        "momentum_score":   0.20,
        "breadth_score":    0.10,
        "macro_score":      0.20,
        "drawdown_score":   0.10,
    },
    ticker_beta={
        "QQQ":  1.20,
        "SPY":  1.00,
        "TLT":  0.30,
    },
    db_path="risk_results.db",
)

Development

Clone the repo and set up an isolated environment:

git clone https://github.com/BrainScanStudios/risk-threshold-engine.git
cd risk-threshold-engine

python3 -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate

pip install -e ".[dev]"

The [dev] extra installs pytest. The library itself has no runtime dependencies beyond the standard library.

pytest
pytest -v          # verbose
pytest -k regime   # run tests matching a keyword

Project layout

src/
    risk_threshold_engine/
        __init__.py    public API surface
        engine.py      all core logic lives here
tests/
    test_engine.py     full test suite

The entire implementation is in engine.py. If you're adding a new action type or regime, the three methods to touch are _build_actions, _decide, and _contextual_notes. If you're changing the scoring model, adjust DEFAULT_FACTOR_WEIGHTS or pass custom weights at construction time.

Submitting changes

1. Fork the repository and create a branch from main.
2. Add or update tests for any behaviour you change — pytest must pass before opening a PR.
3. Keep the zero-dependency constraint: no new imports outside the standard library.

License

MIT — see LICENSE.

Copyright (c) 2026 Brain Scan Studios