An educational dashboard for historical event studies. Import daily prices, prediction-market probabilities, catalyst dates and political or insider disclosures, then measure what happened around those events, honestly: every event is used only from the first trading session after it was published, the baseline period and sample size are always visible, and small samples are labelled as small.
Status: v0.1, published. Try the demo or run it locally (below). It is a descriptive, historical tool. It does not forecast, recommend or trade, and it makes no claim of guaranteed or multiplied returns. Not investment advice.
- Data with provenance. CSV import of prices (with optional volume), event probabilities and events. Each series and event keeps its source, access level (public, licensed, restricted, synthetic) and retrieval date; events keep both their event or transaction date and their publication date. Import is all-or-nothing per file with line-numbered errors; missing values stay missing, never zero.
- Timelines. Price chart with each event's transaction date, the data-arrival delay and day 0 (first session after publication); an events table with the delay in days. Dates can be edited, and a publication date before the event is refused.
- Event study. Pre and post cumulative abnormal returns (CAR) against a constant-mean baseline, with the baseline window and n shown for every event, exclusion reasons (short baseline, out of range, missing data, not yet tradable), overlap flags, a mean CAR path chart, a normal-approximation 95 % interval and sample-size badges.
- Indicators and relationships. Rolling volatility, Wilder RSI, volume z-score shocks and lagged return correlations between two series, labelled as co-movement only, never causation.
- Exploratory rules. Combine two to six conditions that must cross at least two evidence types (event, volume, momentum, price, prediction market). The app reports how often the rule fired and what followed.
- Backtest. Train/test split by date, transaction costs per side, one position at a time, buy-and-hold comparison, max drawdown and uncertainty badges. Every signal and entry passes a no-look-ahead assertion; if one fails, no results are shown.
- Saved work and exports. Saved configurations, workspace JSON export and import, and CSV exports
(event study, indicators, rule hits, backtest trades), each headed by provenance and parameter
comments and protected against spreadsheet formula injection. An export that contains data
labelled
restrictedstays disabled until you tick "Include restricted data". - Sample market. A seeded, synthetic market with invented companies, officials and a prediction market, marked by a banner and Sample badges wherever it appears.
Needs Node.js 20.19 or newer.
npm ci
npm run dev # http://localhost:5176/signal-scope/
npm run verify # format, lint, types, unit tests, build, secret scan, e2escripts\verify.ps1 and scripts/verify.sh run the same steps and stop on the first failure.
To try the import flow, use the invented example files in examples/: import
demo-prices.csv as a price series with symbol DEMO (a source is required; choose access
"Synthetic" since the file is invented), then demo-catalysts.csv as events (labelled synthetic).
Probabilities from another market (prediction markets) are dated but not timed, so a rule deciding at a session's close only uses probabilities dated before that session.
Day 0 for an event is the first session strictly after its publication date (the data carry dates, not times, so a same-day release is treated as after the close). Rule conditions are evaluated at a session's close and trades enter on the next session. The full rules are in docs/spec/00-data-and-time.md; calculations are in docs/spec/01-calculations.md and docs/spec/02-rules-backtest-io.md.
No backend, accounts, live data or network calls. Daily data only: no intraday timestamps, exchange holiday calendars or corporate-action adjustment. The baseline is a constant mean (no market model) and the interval is a normal approximation. See docs/known-limitations.md.
docs/architecture.md, docs/demo.md, docs/test-report.md, docs/handoff.md, docs/backlog.md, SECURITY.md, CONTRIBUTING.md, AI_ASSISTANCE.md.
MIT, see LICENSE. All bundled data (the sample market and the example CSVs) is synthetic.