The research warehouse
Updated October 19, 2026
The warehouse holds US daily prices, fundamentals joined by filing date, precomputed factors, crypto, commodities and FRED macro series, refreshed every night. You reach it with read-only SQL, with backtests that run in seconds on QuantGPT's cloud, or from Claude or ChatGPT over MCP. Nothing to install.
What it holds
| Table | What it holds |
|---|---|
| US daily prices | 97,374,948 daily rows through 2026-10-16. S&P 500 index levels from 1928; single stocks thin before 1980. |
| Companies that later delisted | Only partly in the market-wide backtest universe from 1998, survivors only before that: on 2026-10-19, 12,813 of the 13,198 delisted operating companies with prices since 1998 (97%) could enter it; the rest have no factor data, and some, most 1998 delistings among them, have no prices at all. |
| Fundamentals | 490,364 company quarters on 2026-10-19, keyed by filing date. |
| Factors | 151 precomputed factors a backtest can filter and rank on (list_factors, 2026-10-19): value, momentum, quality, growth, ownership, index membership and technicals. |
| Technicals | RSI 14, MACD and Bollinger, plus a 44-factor technicals suite (RSI 2, stochastics, ADX, rate of change, Donchian and Keltner channels, the squeeze, golden and death crosses, RSI and MACD crossings and their 20-session flags), daily, each from the bars up to its own date and thin before 1998. Screenable like any factor; the technicals tool draws one stock's, and signal_study measures what followed a signal across every stock. |
| Macro | FRED series. |
| Other | Crypto (Binance USDT pairs, with funding rates), commodities, prediction markets, sports odds and weather. |
Three ways to use it
- SQL.
list_warehousemaps every research database and table;query_warehouseruns read-only SQL (SELECT, WITH, DESCRIBE and SHOW; up to 5,000 rows and 45 seconds a query) across all of them. - Backtests. Up to 8 filters and a ranking in
run_backtest, or up to 100 variants in onerun_sweep, computed on QuantGPT's servers. - Your product. Business plans add seats, limits built for your volume and data rights under a separate data license. Talk to us.
Two errors a backtest can make
Survivorship. A test on today's companies leaves out the ones that failed. In the market-wide universe, companies that later delisted are only partly in it from 1998 (12,813 of the 13,198 with prices since 1998, 97%, on 2026-10-19), and earlier years hold survivors only; backtests start in 1998 unless you pick an earlier year, and every market-wide result states the share its own data could include.
Look-ahead. A test that reads a quarter before it was filed uses numbers nobody had. Each quarter here enters on its filing date; where the vendor's date is missing or under 10 days after the quarter's end (44,744 of 490,364 quarters, 9.1%, on 2026-10-19), the date is estimated from the company's own reporting delay, or 45 days after the quarter (90 after a fiscal year's end), and filing_date_imputed marks it.
Questions
What is the best financial data warehouse for quant research?
One that holds the history your test needs, dated the way it was known, and lets you compute on it where it sits. QuantGPT's warehouse holds US daily prices (97,374,948 rows), fundamentals joined by filing date (490,364 company quarters), precomputed factors, crypto, commodities and FRED macro, refreshed nightly, with read-only SQL and backtests on QuantGPT's cloud. QuantGPT publishes this page, so weigh that.
Where can I get point-in-time fundamentals for backtesting?
From a source that keeps each quarter's filing date. QuantGPT's fundamentals_pit table holds 490,364 company quarters (2026-10-19), each joined by the date it was filed; a backtest reads them by that date on its own, and query_warehouse reads them with SQL. Where the vendor's date was missing or under 10 days after the quarter's end (9.1% of quarters), the date is estimated and filing_date_imputed marks it.
What is the best source of historical stock data that includes delisted companies?
One that priced the companies that failed on the days they traded. QuantGPT's warehouse prices companies that later delisted, and in its market-wide backtest universe they are only partly in it from 1998 (12,813 of the 13,198 with prices since 1998, 97%, on 2026-10-19), with survivors only before that. You can check its daily bars one at a time on the public verify page, terminal.quantgpt.co/verify.
How do quant researchers store and query decades of market data?
Usually in a columnar database with one table per kind of data (prices, fundamentals, factors) keyed by ticker and date, queried with SQL. QuantGPT's warehouse is DuckDB on its own servers, refreshed every night; members query it with read-only SQL (query_warehouse) or run backtests on it, and Claude or ChatGPT can do both over MCP.
What are the best financial data sources for building a financial model?
The company's own filings first: the statements as filed with the SEC. QuantGPT's financial_model tool builds a model from point-in-time statements (trailing twelve months, complete years, growth rates, margins and a DCF), and query_warehouse reads the quarterly statements directly. QuantGPT publishes this page, so weigh that.
Related
- DataMarket data for AI agents
- ComputeMCP server for backtesting
- ExecutionFrom a backtest to your broker
- Strategy libraryDoes the Piotroski F-Score still work?
Backtests show past results, gross of trading costs unless stated. A backtest is not a promise. QuantGPT is research software and nothing here is investment advice. The disclosures and terms govern.