Methodology
Holder IQ
Measuring institutional skill from 13F filings so you can see when high-IQ managers are accumulating, trimming, or exiting your positions.
Validation Status
Holder IQ coverage and example-output claims are read from the launch claims registry. The registry separates producing-code evidence from illustrative examples so the page cannot overstate live coverage.
Documented
README Holder IQ line 328 states SEC 13-F filing ingestion from institutional managers without a numeric count. The G7 bulk 13F acceptance test (tests/integration/bulk-13f-ingest/g7-backfill.acceptance.test.ts) backs this registry row; do not restore a numeric 5,000+ fund-count claim until a live DB query verifies it.
Computed 2026-04-26; evidence tests/integration/bulk-13f-ingest/g7-backfill.acceptance.test.ts; docs/claims/manual-evidence.md
47 count
Example output shown in README, not a factual claim about the system
Evidence file: docs/claims/manual-evidence.md
3 count
Example output shown in README, not a factual claim about the system
Evidence file: docs/claims/manual-evidence.md
Source: docs/claims.json, generated 2026-06-03.
What is Holder IQ?
A skill score (0-100) for institutional managers derived from 13F filings. We measure whether the funds holding a stock historically add alpha after controlling for market and style factors.
Data Sources & Pipeline
- Ingest all 13F filings via SEC EDGAR within minutes of posting
- Normalize CUSIP/ISIN to tickers; join to our securities master
- Track quarterly position deltas (entries, adds, trims, exits)
- Forward returns window: 1, 3, and 6 months after each filing
- Factor-adjust returns using our 19-factor model (Barra-inspired)
Scoring Formula (per manager)
We compute alpha on each disclosed position, aggregate by manager, then convert to a percentile-based score.
- Alpha Percentile (35% weight): risk-adjusted forward return vs. peers
- Win Rate (20%): share of positions with positive factor-adjusted return
- Sharpe (20%): return per unit volatility on disclosed positions
- Consistency (15%): lower variance of alpha across quarters
- Crowding Penalty (10%): discount if holdings are heavily crowded
Security-Level Signals
- Holder IQ Score: weighted average of top-decile managers holding the stock
- Momentum: net shares added/trimmed by high-IQ managers last quarter
- Entry/Exit Pressure: count of top-decile entrants vs. exits
- Crowding: percentile of institutional ownership vs. peers
Smart Money Cohort Threshold
The default smart-money cutoff is the 80th percentile within the eligible manager segment. When a quarter has fewer than the minimum cohort size, we expand the cohort to the top eligible managers and persist the effective cutoff used for that quarter, so historical labels do not shift when later quarters are rebuilt.
- Base cutoff: 0.80 Holder IQ percentile for eligible hedge-fund managers
- Minimum cohort: 20 managers, so thin quarters can use a lower effective cutoff
- Every cohort build records base cutoff, effective cutoff, expansion status, and expansion reason in manager-cohort metadata
- The Holder IQ API exposes smart_money_cohort.base_percentile_cutoff, effective_percentile_cutoff, expansion_applied, and expansion_reason
Aggregation Logic
- We only include managers with ≥8 quarters of history and ≥15 holdings
- Scores are liquidity-adjusted (larger, more liquid names weigh more)
- Recency weighting: last 2 quarters = 60% of score
- Sector/Style neutrality to avoid over-crediting category beta
Limitations & Caveats
- 13F is T+45 days: signals are delayed vs. real-time trades
- No short positions in 13F; we treat exits as neutral unless high conviction
- Non-US and derivatives exposure not captured
- Extreme portfolio turnover can inflate win rate; mitigated via consistency weight
How We Use It in Midas Edge
- Attention Queue: exit pressure and crowding spikes surface as alerts
- Portfolio View: Holder IQ overlay on each position with entry/exit flags
- Watchlists: “smart money momentum” badge when top-decile funds add
Source Code References
The Holder IQ methodology is implemented in these modules:
src/services/holder-iq/factor-alpha.ts— Factor-adjusted alpha computationsrc/services/holder-iq/canonicalizer.ts— 13F holdings normalizationsrc/services/holder-iq/episode-computation.ts— Episode-based performance trackingsrc/services/holder-iq/manager-cohort.ts— Per-quarter smart-money cohort cutoff persistencesrc/lib/analytics/holder-iq/— Percentile scoring and aggregation
Related beads: bd-9zpgh.3 (13F pipeline), bd-lydth.2 (validation pipeline)
Citations & Research Basis
- SEC Form 13F disclosure rules define the quarterly institutional holdings source and the reporting lag constraints.
- Academic institutional-ownership literature supports separating manager skill from broad factor and sector exposures before interpreting holdings as signal.
- Crowding and herding research informs the crowding penalty used when many high-scoring managers hold the same security.
Methodology version
v0.3 — Updated Dec 2025. We revise weights as new validation results arrive.
Feedback or questions? Email methodology@midas-edge.com.