position-sizerlisted

# position-sizer You hand over a basket of tickers and (optionally) per-name edges. The skill returns position sizes under four canonical methods, side-by-side, plus the portfolio-level vol, exposure, and binding constraints for each. This is NOT alpha. The script doesn't pick names — you bring the names. It doesn't predict returns — you bring the edges (for Kelly). What it does is the descriptive math that says "given these names and these vols, here's how each sizing method allocates." ## When to invoke - PM has 3-15 names they want in the book; deciding how to weight - Researcher wants to compare Kelly vs vol-target on the same basket - Risk team checking that a proposed book sits inside a target vol - A trader sanity-checking that the discretionary book they built by feel doesn't have one position dominating the variance budget ## What you need - 3-15 tickers (more is fine; risk-parity slows quadratically in N) - `MASSIVE_API_KEY` exported Optional: - Edges per ticker (only needed for Kelly; e.g., "I think NVDA returns 15% annualized, AMZN 10%") - Target vol (default 12%) - Leverage cap (default 1.0x; Σ|w| ≤ cap across every method) - Max single-position weight (default no cap) - Vol estimator (default `realized`; `ewma` for RiskMetrics EWMA with configurable λ, default 0.94, that responds faster to recent regime shifts, so sizing cuts exposure into rising vol rather than lagging it) ## What you get back Two output layers from one analysis. **Layer 1: c