Research
How historical data becomes falsifiable evidence — datasets, backtests, walk-forward, OOS, replay, counterfactuals.
Methodology
The end-to-end research methodology — how evidence is produced, graded, and allowed to influence the platform.
Datasets
Canonical dataset contracts — fingerprints, quality gates, provenance, and the tick-dataset cache.
Backtesting
Deterministic, friction-aware backtests — what is simulated, what is honest about it.
Walk-Forward
Purged and embargoed walk-forward validation — why the purge exists and what happens when it was missing.
Out-of-Sample Gate
The hard OOS gate — floors, rejection semantics, and the evidence that it actually rejects.
Replay
Replay and streaming replay — bit-exact parity with the dataset, the anti-leakage guarantee, and how it differs from backtesting.
Counterfactuals
The NO_TRADE counterfactual engine — walking the decisions the engine didn't take, with stratified evidence.
Validation
The validation philosophy — what "validated" means, the 12 gates, and why rejections are published.
Reproducibility
How NSE makes research runs reproducible — fingerprints, frozen captures, deterministic seeds, provenance snapshots.