docs/research/2026-07-25-market-data-source-picks.md

Market-data source picks — historical, live, and options (cheap-first)

Date: 2026-07-25 · Status: adopted (operator-requested /bandits pull) · Total cost: $0/mo

The trader's data needs, the source picked for each, and why — verified empirically this session, not from vendor marketing. Upgrade paths listed for when a need outgrows free.

Need Pick Cost Verified how Limits
Underlying: historical + live bars (stocks/ETFs) Yahoo (keyless)lib/market-data-yahoo.js, already integrated everywhere $0 powers every backtest in this arc (10y daily, ~1–2mo 15m) intraday capped ~30–60d; occasional bad wicks (ADR-0029)
Options: live chain + quotes (shadow trader, UI) Alpaca indicative feed (free tier of the existing broker keys) — chain snapshots + latest quotes $0 live probe: full SPY 5-leg ladder + penny strike with real bid/asks indicative (estimated NBBO),calls/min, no OPRA depth
Options: HISTORICAL (backtests) Alpaca historical option bars (same free keys) — full OHLCV for any listed/expired contract since 2024-02 $0 pulled realSPY contract bars (o/h/l/c/volume,days) on free keys; scripts/options-hist-backtest.js runs the overnight ladder/penny strategy on ~2.4y of real prices bars only (no historical quotes/greeks on free); floor 2024-02; entry proxied by last trade of day

Rejected / deferred

  • DoltHub community options DB (post-no-preference/options, free SQL API, 2019+):

verified live — schema is good (bid/ask + greeks EOD) but SPY carries only 2–6-week expirations; no next-day contracts, so it cannot price the overnight 0–1DTE strategies. Kept in mind for longer-dated strategy research.

  • optionsDX (free EOD chains, 2013+): real, but signup + manual quarterly ZIP

downloads — not automatable by the trader. Backup if we ever need pre-2024 depth.

  • MarketData.app (free tier: 1y history, greeks,credits/day): credible, but

requires an account/token and the credit budget is tiny for chain work. Candidate first paid step at $30/mo if greeks history becomes necessary.

  • **ThetaData $80/mo · Alpaca Algo Trader Plus $99/mo (OPRA) · Polygon/Massive

real-time $2.5k/mo · Intrinio $1k+/mo**: the upgrade ladder, in order, once a measured edge justifies paying — per the loop's rule, spend follows evidence.

The rule this encodes

Every data need is served free until a measured edge demands better: quotes-level historical fidelity (ThetaData) only if the bars-level backtest shows a real edge worth refining; real-time OPRA ($99) only when a proven strategy actually trades options live.