RiskModels API

Your portfolio already has a benchmark. You just haven’t seen it.

Decompose every position into market, sector, subsector, and stock-specific layers — with ETF hedge ratios you can use from one API call.

Numbers-first answers from live ERM3 data · hedge ratios in dollars of ETF per $1 of stock

Get API Key View Docs See the Worked Example

Read-only · Keys server-sideERM3 (Equity Risk Model v3)Sub-120 msAgent-callable

Worked example — MSFT

Per $1 long MSFT, ERM3 returns three hedge ratios you can trade.

L1 · market

short $0.03 SPY

+20% explained risk

L2 · tech sector

short $0.05 XLK

+4% explained risk

L3 · systems software

short $0.62 IGV

+18% explained risk · dominant

What’s left — the bet

MSFT’s stock-specific return. Not hedgeable with ETFs.

57% residual

Live decomposition from the ERM3 API · single call, no extrapolation · snapshot 2026-05-26

Run on your portfolio See the methodology

Built for three workflows

Pick the path that matches your job.

For developers & agents

Call risk decomposition from your app, notebook, or agent workflow. JSON in, hedge ratios out.

View API Docs

For portfolio managers

See what each position is really betting on — and what ETF trades hedge it. NVDA single-stock and AGTHX fund-fit snapshots, rendered from one API call.

See sample snapshots

For allocators & diligence

Separate benchmark exposure from stock selection in disclosed 13F books.

Read 13F Research

Live Demo

See the API in Action

Stylized demos (not live recordings) — REST and SDK shapes match the API; CLI 'agent' and some MCP tool names illustrate product workflows. Python SDK: python3 -m pip install "riskmodels-py>=0.3.4" on PyPI ([xarray] optional for dataset helpers; [viz] for Kaleido static PNG). CLI: npm install -g riskmodels@latest on npm. Use riskmodels --help and MCP tools/list for what your install exposes. sdk/ · cli/.

riskmodels agent decompose — L3 portfolio attributionportfolio: positions.json

[INFO] Loading portfolio: 8 holdings detected...

[INFO] Calling POST /batch/analyze (ERM3-L3-v30)...

[INFO] L3 attribution complete. Building response...

{

"avg_l3_mkt_hr": 1.14,

"dominant_factor": "market",

"holdings": {

"NVDA": { "l3_mkt_hr": 0.97, "l3_sec_hr": 0.14, "sector_etf": "SOXX" },

"MSFT": { "l3_mkt_hr": 0.88, "l3_sec_hr": 0.11, "sector_etf": "XLK" },

"AAPL": { "l3_mkt_hr": 0.85, "l3_sec_hr": 0.12, "sector_etf": "XLK" }

"_metadata": { "model_version": "ERM3-L3-v30", "universe_size": 2987 }

}

The research behind the model

“Much of what gets labeled as the ‘factor zoo’ is already sitting in plain sight — embedded in sector and subsector exposures as tradable risk. The proliferation of factors is often a labeling exercise.”

— Part 1 · One Position, Four Bets

Published methodology. Open research examples. Reproducible decomposition logic. Real citations (Frisch-Waugh-Lovell, Cremers-Petajisto, Harvey-Liu-Zhu, Grinold-Kahn).

Part 1 · One Position, Four Bets

AAPL vs NVDA, XOM vs KMI, MAG7 DNA.

Read on riskmodels.org

Part 2 · Risk Structure in 13F Filings

Buffett, Ackman, Lone Pine, Tiger Global, Baupost.

Read on riskmodels.org

Methodology · ERM3 Engine Design

Hierarchical orthogonalization, L-star, hedge ratios, capacity.

Read on riskmodels.org

See full institutional snapshots — what ERM3 renders end-to-end