quantOS· Engine

Open-Core Simulator & TUI · v1.0.0

AlphaEngine 1.0

Local Python execution and terminal interface for institutional strategy validation, parameter sweeps, and out-of-sample stress tournaments.

Read CLI DocsGitHub
AlphaEngine 1.0.01.0

AlphaEngine 1.0.0. The trace is the stack. Figures illustrate one run.

Out-of-Sample Stress Tournament

Stress candidate rules before selecting a winner.

Test competing portfolio rules against outlier-day removal, high-volatility windows, and holdout splits—scoring the survivor on its worst-case regime performance.

Closed-Room Adversarial Bakeoff · 6 Proposals vs. Referee Attack
CANDIDATE EQUITY CURVES UNDER DROP_BEST ATTACKDETERMINISTIC REFEREE STANDING (IN-SAMPLE → ATTACKED)1. Risk Parity + Walk-Forward ★1.68 →1.42SURVIVOR2. Volatility Target (60d)1.44 →1.08KEPT3. Inverse Volatility1.31 →0.92KEPT4. Equal-Weight (1/N)1.18 →0.74THINNED5. Unhedged Momentum (20d)1.92 →−0.18KILLED6. Short-Horizon Reversion (5d)1.75 →−0.34KILLED
Referee Survivor

Risk Parity + Walk-Forward

Worst-Attack Sharpe

1.42

Deflated Sharpe (DSR)

0.84 (6 Calls)

Hard Wall-Clock Cap

45s Default (60s Max)

  1. 01

    Candidate Sleeves

    Generate competing allocation rules (risk parity, inverse-vol, equal weight, momentum, mean reversion) across 5d to 60d lookbacks.

  2. 02

    Out-of-Sample Stress

    Subject each candidate to outlier-day removal (drop_best), 95th-percentile volatility windows (high_vol), and strict holdout splits.

  3. 03

    Selection-Adjusted Scoring

    Deterministic code ranks candidates by their worst stressed Sharpe and adjusts the survivor for multiple-testing breadth.

Parameter Stability & Selection Audit

Sweep the surface. Discount for search breadth.

Evaluate lookbacks from 5d to 120d and holding horizons from 5d to 63d as one surface—separating durable parameter plateaus from isolated backtest spikes.

validate_study · Parameter Surface Heatmap & Deflated Winner Path
LOOKBACK × HOLDING HORIZON SURFACE (ONE CELL DID NOT RUN)5d10d21d42d63d5d10d20d40d60d80d100d120dWINNER EQUITY CURVE & UNDERWATER DRAWDOWN (LOOKBACK 60D · HORIZON 21D)
Selected Winner Cell

Lookback 60d · Horizon 21d

Raw In-Sample Sharpe

1.64

Deflated Sharpe (DSR)

0.86

CSCV Overfitting (PBO)

14.2%

  1. 01

    Natural-language or CLI execution

    Type a plain-English research question or run a scripted CLI command. AlphaEngine sequences up to eight deterministic tool calls.

  2. 02

    Live execution trace

    Every tool returns verified portfolio metrics and renders the live parameter stability surface inside the terminal session.

  3. 03

    Local JSON or Portal seal

    Save the versioned study artifact locally or seal the findings directly onto your active thesis in /portal.

Deterministic Primitives

Fifteen tools. Zero hallucinated math.

The model selects the tool order; deterministic Python computes every figure. Any number not returned by a tool is refused.

01 · Market Data & Regimes5 tools

Hygiene & Regimes

  • profileAudit missing closes, split spikes, and stale series.
  • alignPoint-in-time calendar alignment; gaps left open.
  • regimeTwo-state volatility and macro regime filter.
  • processDrift, volatility, jump, and mean-reversion fit.
  • workflowRun any canonical platform workflow by name.
02 · Derivatives & Relative Value5 tools

Surface & Pairs

  • priceOptions chain valuation and full-surface Greeks.
  • repriceScenario revaluation under spot and vol shocks.
  • cointegrationRelative-value pair spreads and stationarity.
  • correlationCross-sectional and rolling correlation matrix.
  • covarianceShrinkage covariance for portfolio construction.
03 · Portfolio & Risk Audit5 tools

Sizing & Stress

  • portfolioRisk-parity (HRP) and volatility-targeted weights.
  • overlapMarginal beta and drawdown overlap vs. the book.
  • performanceSharpe, drawdown duration, and Treasury excess.
  • varHistorical 5% VaR and 95% tail CVaR.
  • bakeoffMulti-strategy stress tournament and scoring.

Shared Workflows

Eight canonical checks.

Run interactively inside the TUI or as one-shot CLI commands for scripts and CI pipelines.

Local Custody

Your strategy module never leaves your machine.

Execute custom Python modules locally and pin the sealed statistical artifact to the hypothesis Ideate filed.

  1. 01

    Zero vendor lock-in

    Run offline on your own CSV panels or registered desk universes using your local environment key.

  2. 02

    Proprietary strategy code stays on disk

    validate_study executes your custom Python module locally and uploads only the sealed summary metrics.

  3. 03

    Pinned numerical parity

    Exact floating-point parity with the cloud workspace so local CLI runs and portal studies match byte-for-byte.

$ pip install alphaengine
$ alphaengine run validate_study --project ./my_strategy.py
$ alphaengine run validate_study --thesis hyp_semis_rv_04

A sentence builds a stack. A command runs one workflow.

stress this book, then bake off a survivor

align · regime · bakeoff·the referee names the survivor

$ pip install alphaengine

Get startedRead CLI Docs