SSStrategicSignal

Exceptional Capital methodology v0.3.1

Delayed filings can still contain signal — if the clock and the sample are honest.

The method compares consecutive public Form 13F holdings, identifies material position changes, clusters event outcomes by filing quarter, and qualifies managers only when held-out and recent evidence both clear published gates.

Core principles

01Use public SEC Form 13F filings as the source of record for institutional holdings changes.
02Measure outcomes from the first trading close strictly after the public filing date, never from quarter-end or the filing-day close.
03Classify comparable quarter-over-quarter disclosed holdings changes as NEW, ADD, REDUCE, or EXIT.
04Describe EXIT as a disclosed-position exit; do not infer bearish intent, complete portfolio exposure, or replacement exposure outside Form 13F.
05Treat each manager filing quarter, not each individual security event, as the independent evidence unit.
06Equal-weight qualifying event outcomes within each manager-quarter-horizon before calculating evidence statistics.
07Require both time-ordered held-out evidence and a separate recent-six-quarter stability test before a manager can qualify.
08Treat manager reputation as candidate selection only; historical evidence determines eligibility for live signal scoring.
09Publish security-level convergence only from independently disclosed events belonging to qualified managers.
10Keep U.S. Senate disclosures methodologically separate; Senators are never assigned manager-quality or investment-skill scores.
11A qualified manager does not make every action predictive: an event receives a public score only when the relevant event-type evidence is at least 50, or an allowed broader fallback clears its published threshold.

Timing and benchmark

The return clock begins at the first regular trading close strictly after the SEC filing date. NEW and ADD use security return minus SPY; REDUCE and EXIT use SPY minus security return. Quarter-end performance is never treated as if the filing were already public.

Independent evidence unit

Each manager filing quarter is one independent observation per horizon. Qualifying event returns within that manager-quarter-horizon are equal-weighted before hit rate and mean directional excess return are calculated.

Manager Evidence Score

For each horizon, raw evidence is: 50 + 30*tanh(mean directional excess / 10%) + 20*tanh((hit rate - 50%) / 15%), bounded to 0–100. The score is reliability-shrunk toward 50 using completed 12-month filing-quarter count, distinct issuer breadth, and price coverage. Overall evidence weights 3 months at 25%, 6 months at 35%, and 12 months at 40%.

Event Strength calculation

Event Strength is not company quality or a return forecast. It measures one disclosed position-change event. The exact calculation is round(35% × materiality + 25% × portfolio weight + 20% × manager/action evidence + 20% × convergence).

35% · EVENT MATERIALITYCanonical raw input = 1 for NEW/EXIT and abs(share_change_pct)/100 for ADD/REDUCE. Score = min(raw, 1) × 100.
25% · PORTFOLIO WEIGHTNEW/ADD use current disclosed portfolio weight; REDUCE/EXIT use prior weight. Score = min(weight % / 5, 1) × 100.
20% · MANAGER/ACTION EVIDENCEUses the qualifying historical action-type score. When action history is insufficient, direction evidence ≥50 may be used, then held-out overall evidence ≥55. A below-50 action score is not rescued.
20% · CONVERGENCE1 qualified manager = 0; 2 = 60; 3+ = 100.

The four unrounded weighted point contributions are summed and JavaScript Math.round produces the published integer Event Strength. Qualitative event labels are descriptive only: strong 70–100, moderate 50–69, limited 0–49.

Worked example · MU / Coatue ADD

Coatue increased its disclosed MU position from 165,931 shares to 3,142,269 shares, a 1793.72% increase (canonical materiality raw input 17.9372, capped to a 100 materiality score). MU represented 7.46% of Coatue’s disclosed 13F equity value. Materiality scores 100.0 → 35.00 points; portfolio weight scores 100.0 → 25.00 points; Coatue ADD evidence is 78.3 → 15.66 points; 1 qualified manager means convergence scores 0.0 → 0.00 points. The unrounded total is 75.66; Math.round(75.66) = 76. This is strong accumulation-event evidence, not a probability of appreciation or a Micron quality score.

Qualification gates

A manager needs at least 5 years of history, 12 completed notable events, 6 distinct issuers, 80% 12-month price coverage, 6 held-out completed filing quarters, and 6 recent completed filing quarters. Held-out overall evidence must be at least 55, with 6-month and 12-month raw scores at least 50. The recent-six-quarter overall score must also be at least 55 and recent 12-month raw evidence at least 50.

Statuses. Candidate = insufficient data gate. Watch = sufficient data but failed evidence/stability gate. Qualified = every data, held-out, and recent-stability gate passes. Only qualified managers are eligible for live signal scoring.

Action-specific publication gate

Qualified manager status is necessary but not sufficient. If an event-type evidence score exists, it must be at least 50. A below-50 subtype is not rescued by a stronger broader manager score. When subtype history is insufficient, direction evidence may be used at 50 or above, followed only then by held-out overall evidence at 55 or above.

Notable exits

EXIT remains first-class. It qualifies when the prior disclosed position was at least 0.5% of the disclosed equity book or ranked in the top 20. The wording remains deliberately narrow: “disclosed position exited.”

Publication boundary

Version 0.3.1 publishes validated manager evidence and the first scored event register for Q2 2026. The register is a delayed-disclosure evidence layer, not a real-time trade feed or investment recommendation.