StockDriftsStockDriftsA filing from a 5-star activist is a fundamentally different signal than the same filing from an unknown passive fund. We built a 1-5 star rating system to quantify exactly that — based on every EDINET filer's verified track record.
Japan's large-shareholding system (大量保有報告書) requires any investor crossing 5% ownership to disclose their purpose and transaction type. We use those fields to assign signal tiers (S/A/B/C) based on historical returns by Purpose × Transaction type.
But two filings with identical purpose and type can come from an investor with a proven 10-year track record or one who has never generated a positive return. The investor identity is a primary signal carrier — not a secondary filter. Our ratings make this distinction explicit.
We tracked every investor who has ever filed on EDINET — across 50,000+ large-shareholding reports. For each investor, we computed the return on every BUY and INITIAL filing (filing price → current price) and aggregated across two time windows: 5-year and 3-month.
(current_price − filing_price) / filing_price for buys. Averaged across all qualifying filings in each window.We combine two return windows into a single composite score that drives the star rating:
| Return window | Weight | Rationale |
|---|---|---|
| 5-year | 70% | Long-term foundation — captures a full market cycle of proven alpha |
| 3-month | 30% | Strongest statistical predictor in our data — especially for activists |
The formula: composite = 0.70 × 5y + 0.30 × 3m. When a window has fewer than 2 qualifying buys, its weight is redistributed proportionally to the remaining window.
Five years captures a full market cycle — bull runs, corrections, and recoveries. An investor who has generated alpha across these regimes has demonstrated genuine skill, not just momentum. We give 5-year returns the dominant 70% weight because long-term consistency is the strongest indicator of an investor's true ability.
Our dataset spans from March 2021 to present — almost exactly five years of EDINET filings. This means the 5-year window captures virtually all available data while still being bounded enough to reflect the investor's current-era performance rather than outdated positions.
We ran a Spearman rank-correlation analysis across all investors, comparing their return rank over different lookback windows against their 90-day forward return. The 3-month window was the clear winner — 3× more predictive than all-time scores:
| Window | Spearman r | p-value | Top vs Bottom Spread |
|---|---|---|---|
| 3 month(Strongest) | +0.153 | <0.0001 | +6.9% |
| 6 month | +0.089 | 0.003 | +4.9% |
| 12 month | +0.049 | 0.094 | +1.8% |
| All-time | +0.057 | 0.051 | +1.8% |
The 3-month signal is especially strong for activist investors (r=+0.19, p=0.005) — the "hot hand" effect is real and concentrated in this category. Passive and strategic investors show little recency signal, so the 3-month weight is conservative enough not to overfit to this subset.
Investors are ranked by composite score, then assigned stars by percentile:
Exceptional long-term track record with recent confirmation
Consistently strong returns across windows
Above-median performer
Below-median but not a clear red flag
Weak historical performance — treat with caution
On the dashboard, the investor rating is combined with the filing's signal tier (S/A/B/C) into a single conviction score. The signal tier (based on Purpose × Transaction Type) carries 60% of the weight, and the investor's star rating carries 40%.
See the full conviction score breakdown and tier tables on the Signal Tiers page →
Investor ratings are based on historical return data from EDINET filings and are updated periodically. Past performance does not guarantee future results. Ratings should not be used as the sole basis for investment decisions. Returns are calculated from filing date to current market price and will change over time. Always conduct your own research before investing.
See investor ratings live alongside every filing on the Japan Insider Dashboard.