One database for every market question.

The problem

A simple market question still takes a data team. Prices sit in one system and fundamentals in another, indicators are computed in overnight batch jobs, research lives in notebooks, and alerts in yet another tool.

• Five or six systems glued together, each with its own symbols, calendars and gaps.
• Answers computed overnight, so screens are stale by the open.
• Research that quietly uses hindsight, and backtests that look better than reality.
• Coverage that stops at US large caps the moment you ask about Tokyo, Mumbai or a small cap.
• AI assistants that sound confident and make numbers up.

What EquationDB is

One purpose-built database in place of that stack. Four ideas make it different.

From what happened, to why, to what if

Most market data stops at what happened. Here, why and what if are questions the database answers too: with evidence from the past and its sample size, never a forecast.

WHY NVDA LAST 1mo

SIMULATE NVDA -5% ON top500 FWD 1d, 5d, 20d

SERIES AAPL.close, AAPL.insiders.sell_value LAST 1y ALIGN monthly

Ask the market anything

One language goes from a quick screen to an event study, an explanation or a simulation. Pick a question to see the query that answers it.

GET top500
WHERE rsi(14) < 30 AND close > sma(200)
SORT rsi(14)

EVENTS * IN top500 LAST 1d
WHERE rarity >= 95
SORT rarity DESC

OUTCOMES AFTER ret < pctl(5, 2y) AND volume > pctl(95, 2y)
           AND $TH.AI.INFRA@drawdown < -10
  IN (GRAPH MATCH company -supplies-> company -has_theme-> TH.AI.INFRA WHERE mcap > 10b)
  SINCE 2018-01-01 FWD 5d, 20d, 60d GROUP BY sector

LET movers = GET us.movers SELECT symbol, ret SORT abs(ret) DESC LIMIT 5;
FOR EACH s IN movers WHY $s LIMIT 4

SIMULATE SPY -2% OVER 3d AND $VIX > 18
  ON (GRAPH NEIGHBORS OF NVDA ACROSS supplies)
  FWD 5d, 20d

SERIES CORR NVDA.insiders.net_buy_value, NVDA.close LAST 5y ALIGN monthly LAG 12

GET all WHERE has_theme('TH.ENERGY.NUCLEAR', 0.5) AND has_theme('TH.AI.INFRA.POWER', 0.5)
SELECT symbol, name, exchange, node_name,
       theme_score('TH.ENERGY.NUCLEAR') AS nuclear, theme_score('TH.AI.INFRA.POWER') AS ai_power,
       roc(126) AS ret_6m, roc(126) - $TH.ENERGY.NUCLEAR@roc(126) AS vs_theme_index, pe, ev_ebitda
SORT ret_6m DESC

BACKTEST LONG IN (GRAPH MATCH company -supplies-> company -has_theme-> TH.AI.INFRA WHERE mcap > 10b)
  SINCE 2019-01-01
  ENTER WHEN close CROSS ABOVE highest(high, 55)[-1] AND rvol(20) > 1.2
         AND $TH.AI.INFRA@close > $TH.AI.INFRA@sma(200)
  EXIT WHEN close < sma(50) OR $TH.AI.INFRA@close < $TH.AI.INFRA@sma(200)
  STOP 12% COST 10 bps MAX 8

LET setups = GET top500 WHERE rsi(14) < 40 AND close > sma(200) AND $BRD.A200@close > 50
             SELECT symbol, rsi(14) AS r SORT r;
IF count(setups) = 0 THEN STOP;
FOR EACH s IN setups LIMIT 5
  OUTCOMES AFTER rsi(14) CROSS BELOW 40 AND close > sma(200) IN $s SINCE 2010-01-01 FWD 5d, 20d

Or type the question in plain English and see the query it becomes. Turn any condition into an alert or a paper-trading flow, and view any result as a table, a chart or a pivot.

One language, or plain SQL

Full SQL over the same data: WITH, JOIN, window functions, QUALIFY, SAMPLE BY, user-defined functions, procedures and in-database models. Indicators such as rsi(14) or pe are just columns, and any statement can be a table in a query.

SELECT symbol, sector, roc(63) AS momentum_3m,
       rank() OVER (PARTITION BY sector ORDER BY roc(63) DESC) AS rk
FROM top500
QUALIFY rk = 1
ORDER BY momentum_3m DESC

WITH q AS (SELECT symbol, period_end, filed,
                  sum(eps_diluted) OVER (PARTITION BY symbol ORDER BY period_end ROWS BETWEEN 3 PRECEDING AND CURRENT ROW) AS ttm_eps,
                  count(*) OVER (PARTITION BY symbol ORDER BY period_end ROWS BETWEEN 3 PRECEDING AND CURRENT ROW) AS quarters
           FROM financials
           WHERE symbol IN ('AAPL', 'MSFT', 'NVDA', 'GOOGL', 'META') AND period <> 'FY' AND period_end >= '2019-06-01'),
     pe AS (SELECT symbol, filed, ttm_eps, price_asof(symbol, filed) / ttm_eps AS pe_at_filing
            FROM q WHERE quarters = 4 AND ttm_eps > 0 AND filed IS NOT NULL),
     h AS (SELECT symbol, count(*) AS filings, min(pe_at_filing) AS pe_low, median(pe_at_filing) AS pe_median,
                  max(pe_at_filing) AS pe_high, avg(pe_at_filing) AS pe_avg, stddev(pe_at_filing) AS pe_sd, last(ttm_eps) AS latest_ttm_eps
           FROM pe GROUP BY symbol)
SELECT h.symbol, h.filings, h.pe_low, h.pe_median, h.pe_high,
       s.close / h.latest_ttm_eps AS pe_now, (s.close / h.latest_ttm_eps - h.pe_avg) / h.pe_sd AS z_vs_own_history
FROM h JOIN (AAPL, MSFT, NVDA, GOOGL, META) s ON h.symbol = s.symbol
ORDER BY z_vs_own_history

SELECT p.symbol, p.name, p.sector, p.expected_move_pct AS shock_move_pct, t.exp_move_pct AS implied_daily_move_pct,
       abs(p.expected_move_pct) / t.exp_move_pct AS shock_vs_implied, t.iv30, p.corr_60d, p.path
FROM (GRAPH PROPAGATE NVDA -8% VIA beta DEPTH 2 LIMIT 40) p
JOIN top500 t ON p.symbol = t.symbol
WHERE t.exp_move_pct > 0
ORDER BY shock_vs_implied DESC
LIMIT 15

SELECT o.group AS sector, o.horizon, o.n, o.mean_pct, o.edge_pct, o.t_stat,
       rank() OVER (PARTITION BY o.horizon ORDER BY o.edge_pct DESC) AS rk
FROM (OUTCOMES AFTER close CROSS ABOVE highest(high, 252)[-1] AND rvol(20) > 1.5
      IN top500 SINCE 2010-01-01 FWD 5d, 20d, 60d GROUP BY sector) o
WHERE o.n >= 20
QUALIFY rk <= 3
ORDER BY o.horizon, rk

WITH j AS (SELECT s.timestamp, s.close AS spy, t.close AS tlt,
                  corr(s.ret, t.ret) OVER (ORDER BY s.timestamp ROWS BETWEEN 59 PRECEDING AND CURRENT ROW) AS corr_60d,
                  0.6 * (lead(s.close, 63) OVER (ORDER BY s.timestamp) / s.close - 1) * 100
                  + 0.4 * (lead(t.close, 63) OVER (ORDER BY s.timestamp) / t.close - 1) * 100 AS fwd_6040
           FROM bars s JOIN bars t ON s.timestamp = t.timestamp
           WHERE s.symbol = 'SPY' AND t.symbol = 'TLT' AND s.timestamp >= '2008-01-01')
SELECT CASE WHEN corr_60d < -0.2 THEN '1 bonds hedge (below -0.2)' WHEN corr_60d < 0.2 THEN '2 unrelated' ELSE '3 bonds move with stocks (above 0.2)' END AS regime,
       count(*) AS days, avg(corr_60d) AS avg_corr, avg(fwd_6040) AS fwd_63d_avg, min(fwd_6040) AS fwd_63d_worst,
       sum(CASE WHEN fwd_6040 > 0 THEN 1 ELSE 0 END) * 100.0 / count(*) AS pct_positive
FROM j
WHERE corr_60d IS NOT NULL AND fwd_6040 IS NOT NULL
GROUP BY regime ORDER BY regime

WITH pf(symbol, weight) AS (VALUES ('AAPL', 0.20), ('MSFT', 0.20), ('AMZN', 0.15), ('JPM', 0.15), ('XOM', 0.10), ('LLY', 0.10), ('CAT', 0.10)),
     hits AS (SELECT 'NVIDIA falls 10%' AS scenario, pf.symbol, pf.weight, p.expected_move_pct AS move
              FROM pf JOIN (GRAPH PROPAGATE NVDA -10% VIA beta DEPTH 2 LIMIT 500) p ON pf.symbol = p.symbol
              UNION ALL
              SELECT 'Exxon falls 8%', pf.symbol, pf.weight, p.expected_move_pct
              FROM pf JOIN (GRAPH PROPAGATE XOM -8% VIA beta DEPTH 2 LIMIT 500) p ON pf.symbol = p.symbol
              UNION ALL
              SELECT 'JPMorgan falls 7%', pf.symbol, pf.weight, p.expected_move_pct
              FROM pf JOIN (GRAPH PROPAGATE JPM -7% VIA beta DEPTH 2 LIMIT 500) p ON pf.symbol = p.symbol),
     r AS (SELECT scenario, symbol, weight, move, row_number() OVER (PARTITION BY scenario ORDER BY move) AS rk FROM hits)
SELECT scenario, sum(weight * move) AS portfolio_move_pct, count(*) AS holdings_hit,
       max(CASE WHEN rk = 1 THEN symbol END) AS worst_holding, min(move) AS worst_move_pct
FROM r
GROUP BY scenario
ORDER BY portfolio_move_pct

These are taken from the runnable Demo in the app: a quick tour, a demo sheet of 700+ runnable commands and 260+ Power queries, Advanced and Expert, each with the question it answers, and each runs against the live database with one click. Open the Demo (sign-in).

What's inside

More than a query box. Everything below runs on the same data and the same language.

Links to the documentation and the Demo open in the app and ask you to sign in first.

Built for AI agents

Language models reason well and remember numbers badly. EquationDB gives them a database to query instead of a guess, through the same door people use.

Open the Agent view (sign-in), or talk to us about agent integrations.

One engine, many databases

EquationDB is an engine, not a market product. It works on a few principles that hold wherever things are measured over time. Each database applies the same principles to one domain. Financial markets was the first; Weather is the second.

Databases

1. 01
   Active · the first database

   Financial markets

   Stocks, ETFs, funds, indexes, FX and crypto on 70 exchanges, with fundamentals, market events, themes and supply chains.

   70 exchanges ~70,000 stocks, ETFs and funds; companies from 108 countriesSince 1962 312M price bars, split & dividend adjustedFX, crypto, indexes 1,400 currency pairs · 2,100 tokens · 10,000+ theme and industry indexesFundamentals statements for ~51,000 companies; earnings, dividends, estimates

2. 02
   Active

   Weather

   Daily weather for 399 US stations in all 50 states and 55 world cities, back to 1869: temperature, rain, snow, wind and degree days as series; records, heat waves, cold snaps and storms as events; what markets did after them.

   399 US stations first-order stations in all 50 states, updated daily55 world cities history; international data arrives about a year lateSince 1869 daily records, most stations from the 1940s10 event types records, heat waves, cold snaps, heavy rain and snow, freezes, gusts

3. 03
   Exploring

   Energy & power

   Load, generation and prices per plant, grid node or region as series; outages and price spikes as events, rare for that node at that hour; what followed, without hindsight; a graph of plants, nodes, interconnectors and fuels.

4. 04
   Exploring

   Supply chains & logistics

   Transit times, port calls and freight rates per port, lane or carrier as series; congestion and diversions as events, rare for that lane; what followed downstream, without hindsight; a graph of ports, lanes, carriers and shippers.

5. 05
   Exploring

   Public health

   Aggregate case and admission counts per region as series; outbreak onsets and early seasonal peaks as events, rare for that region's own seasons; what followed, without hindsight; a graph of regions, facilities and conditions.

Financial markets and Weather are the databases you can use today. The others show where the same engine goes next; they carry no dates.

Data & coverage

About 70,000 instruments on 70 exchanges, not just the index names: ~52,600 stocks, ~12,300 ETFs and ~5,400 mutual and closed-end funds, from New York, Toronto and São Paulo to London, Frankfurt, Mumbai, Shanghai, Hong Kong, Tokyo, Sydney and Dubai. Plus currencies, crypto and 10,000+ indexes the database builds itself.

Run SHOW DATASETS in the app for live counts, date ranges and update times.

See it on your own questions

Tell us what your desk or your agents need to answer, and we will set up a trial around it.

For hedge funds and prop desks, quant and research teams, advisors and platforms. Encrypted in transit, hashed credentials, two-factor authentication, tenant isolation and a full audit trail. Dedicated and private deployments available.