Dual-Horizon Ensemble
Fast 5-day forecasts blended with slower structural views to reduce whipsaw. Signals are evaluated separately by horizon, side, and asset bucket.
Research & Development
ProjectV Lab
00Capital’s internal quant lab, focused on building a regime-aware, ML-driven, cross-asset engine that can survive real markets — not just look good in a backtest.
What ProjectV Actually Is
Not a single “model” — a full engine around data, signals, and risk.
Universe & Data
Liquid ETFs, large-cap equities, and macro proxies cleaned into a research-ready dataset.
- Equities, sector & factor ETFs, REITs, and commodity exposure.
- Price-based features, volatility & trend measures, breadth indicators.
Signals & Alpha
Ensemble of models predicting multi-day returns and relative strength, blended across horizons.
- Random Forest, Gradient Boosting, and Ridge Regression ensemble.
- Outputs directional views (long / short / flat) plus confidence estimates.
Portfolio & Risk Layer
Position sizing, exposure bands, and cost modeling wrapped around the raw predictions.
- Position caps, turnover controls, and regime-dependent exposure.
- Focus on drawdown behavior and survivability, not just peak CAGR.
00Capital Quant Stack
ProjectV sits inside a stack of alpha, risk, and execution modules that all have a defined role.
Core Alpha & Regime
- ProjectV ensemble — RF / GB / Ridge predicting multi-day returns and directional signals (buy / sell / long / short) across the universe.
- Regime classifier — GMM-style labels for bull, bear, sideways, and high-volatility states that gate how aggressive the engine is allowed to be.
Risk, Vol & Sizing
- Volatility model (GARCH-style) to forecast near-term volatility and scale positions up or down.
- Monte Carlo risk engine simulating thousands of paths to estimate drawdowns, tail risk, and ruin probabilities.
- Kelly / fractional Kelly sizing converts edge + variance into conservative position sizes as a % of portfolio.
- Drawdown monitor & kill-switches that can throttle or pause the system when behavior moves outside expected ranges.
Portfolio, Execution & Analytics
- Allocation module that spreads capital across names with caps by asset, sector, and net exposure bands.
- Execution bot (broker API) that turns final signals + sizes into orders, rebalances, and trade logs.
- Analytics layer for equity curves, Sharpe / Sortino, and per-asset / per-regime attribution.
In short: ProjectV generates signals, the regime & risk stack decides how much risk is acceptable, and the allocation & execution layer turns that into an actual portfolio with measurable behavior.
Signal → Risk → Execution
How an idea moves through the stack before it ever becomes a trade.
1. Alpha
ProjectV scores the universe and produces long / short / flat views.
2. Regime & Volatility
Regime classifier and volatility forecasts adjust how much we trust those signals.
3. Sizing & Limits
Kelly-inspired sizing, exposure bands, and drawdown rules shape position sizes.
4. Portfolio & Execution
Allocation module and the bot translate the final book into actual orders.
Current Generation: ProjectV 6.x
Dual-horizon, cost-aware, regime-aware.
Regime Labeling & Stress Tests
Bull, sideways, and stress regimes tagged using volatility and breadth. Any change to the engine is graded by how it behaves in each environment.
Cost-Aware Portfolio Construction
Turnover limits, spread / slippage assumptions, and per-asset cost estimates baked into allocations instead of ignored in hindsight.
Performance analytics for backtests and paper portfolios live on the Performance page. The lab is focused on robustness, not marketing-grade equity curves.
Version History at a Glance
Each version fixed a specific pain point in the previous one.
v1 — Equity-Only Sandbox
Tiny long-only US equity universe with basic ML on 5-day returns. Backtests looked great, robustness did not. First lesson in overfitting.
v2 — Cross-Asset Expansion
Added ETFs, REITs, and simple macro proxies to see which signals survived outside a hand-picked equity sandbox.
v3 — First Risk Discipline
Introduced position caps, sector limits, and drawdown awareness, trading a bit of headline CAGR for a portfolio that didn’t implode in bad weeks.
v5 — Regime-Aware Analytics
Split performance by regime instead of averaging it away. A lot of “good” ideas quietly died when they failed stress regimes.
v6 — Dual-Horizon, Cost-Aware Engine
Current architecture: blended horizons, regime-specific reporting, and transaction-cost-aware allocations. Everything from here is iteration.
In the Lab Right Now
Active experiments that feed the next versions of ProjectV.
Signal Stability Under Regime Flips
Testing which features decay the fastest when volatility spikes or correlations break, and which stay usable across environments.
Portfolio Throttles & Kill-Switches
Designing circuit-breaker rules based on drawdown, regime, and cross-asset behavior so the system can cut risk before it becomes fatal.
Multi-Asset Roadmap
Laying groundwork to extend beyond listed ETFs and equities into additional asset classes while keeping execution realistic.
Deeper Technical Access
Select investors and collaborators can receive a more detailed ProjectV technical brief.
If you’re interested in the engine under the hood — feature sets, validation framework, or portfolio construction details — reach out via the Access form with “ProjectV Lab” in the subject line. All materials are provided for research discussion only and do not represent an offer to manage capital.