This is the direction of travel for the QuantJourney Backtester, not a schedule. Items are grouped by theme and listed without dates or ordering commitments; priorities shift with user feedback. Nothing here is a promise of delivery.
The guiding goal is unchanged: make research assumptions explicit, keep runs reproducible, and close the gap between a good-looking backtest and one that survives due diligence.
- Point-in-time (PIT) fundamentals. Company financials and derived ratios stamped with the date each figure actually became public, so strategies never see a number before it was knowable. Sourced from processed regulatory filings with restatement tracking.
- Survivorship-free universes. Historical index membership as it was on each date, including names that later delisted, so index backtests stop implicitly betting on today's winners.
- Point-in-time reference data. Temporal ticker mapping and sector/industry classifications, so long histories do not break on symbol changes.
- Corporate-actions ledger. Dividends credited as real cash, delisting returns realized, and splits handled on position quantities - separating price return from total return in an auditable way.
- Financing, borrow, and margin. Short borrow fees, funding on leverage, interest on cash, and margin modeling - so long/short and futures research reflects the cost of carry instead of omitting it. This closes the borrow and financing assumptions noted in the README limitations.
- Multi-currency accounting. Converting non-base-currency positions and PnL to the portfolio's base currency for global and cross-asset books.
- Faster reproducible starts. Predefined data bundles for the example universes, so common strategies run in seconds without repeated data fetches.
- Cross-sectional and factor tooling. First-class helpers for ranking, neutralizing, and combining signals across a universe, building on the weight-mode long/short examples.
- Deeper overfit detection. Combinatorial purged cross-validation and canonical CSCV probability of backtest overfitting on top of the existing walk-forward, pre-OOS purge, DSR and rolling rank-stability diagnostics.
- Multi-timeframe research. Combining signals observed on one cadence with execution on another.
- Richer reporting. More report formats and interactive views alongside the current static charts and PDF tear sheets.
Have a request or a use case that is not covered? Feedback shapes what moves up this list.