Role
Normalized seniority
InsidersTradesSigma
See what we ingest, what is complete enough to quantify, what reaches a T+90 observation, and what can actually influence a recommendation. Coverage is not proof.
A connector is a feed. A market groups one or more feeds. Only per-trade events can enter the current recommendation research corpus.
Scoring coverage and price-outcome coverage are related, but neither is a subset of the other across historical versions. They are shown as separate branches.
Choose a field, then a market. Identity means a linked, non-placeholder person or entity. The database does not split first and last names reliably.
No market currently has sufficient local evidence for a return allocation. Market returns stay observations, never promises.
The strict statistical corpus uses reported open-market purchases without quarantine, then requires a publication-time price and a T+90 outcome. Portfolio simulations add capital, overlapping positions, costs and execution constraints.
The live multi-market path blends market-local and global publication-time percentiles. Sparse markets borrow more from the global distribution. Missing enriched fields are neutral, not rewarded.
Normalized seniority
Distinct same-direction insiders over the past 30 days
Reported EUR notional
Closed prior T+90 windows with a 91-day embargo
local confidence = n / (n + 300)A score of 80 is a rank, not an 80% return forecast. Display score and ordering score are still separate technical values.
Improve execution and live/backtest parity before tuning country coefficients. Local models activate progressively only when their out-of-sample evidence matures.
hybrid rank = global rank × (1 - local confidence) + local rank × local confidenceReplay exact V15 snapshots, eligibility and publication-time ordering.
Continue down the ranked list when a symbol is already held or has no entry price.
Fixed partial pooling, market and region caps, market-specific costs.