New positions
Growing sampleA manager disclosed a stock that was absent from the prior filing.
- Complete
- 139
- Supported
- 43.9%
- Median vs SPY
- -2.3 pts
+22 still tracking
Signal Outcomes Lab
Which disclosed decisions held up after they became public?
We pool material 13F actions across managers, start the clock after disclosure, and observe each stock against SPY for 63 market sessions. This evaluates patterns—not personalities.
- Complete outcomes
- 1,250
- Direction supported
- 51.2%
- Median vs SPY
- +0.8 pts
- Outcome window
- 63 sessions
Separate prospective register
Track evidence from the day we register it.
14 curated decisions registered from 2026-09-07. 0 complete windows, 13 tracking, 1 waiting for eligible closes. The historical results on this page do not count as prospective performance.
For each version, the clock starts strictly after both its source filing date and its recorded registration date. A correction preserves the previous source and observations, then restarts the revised claim's clock. Editorial selection and overlapping stock returns still prevent a claim of predictive validation.
Inspect a registered decision and its source versions →Two products. Two different questions.
This is not another investor ranking.
Here · Signal Outcomes Lab
The unit is a decision.
NEW, ADD, TRIM, EXIT, reversal, agreement, or disagreement. The clock starts once the filing is public. Results are pooled across managers and direction-adjusted against SPY.
Elsewhere · Investor track records
The unit is a clone portfolio.
Each manager's disclosed portfolio is reconstructed, rebalanced quarterly, and compared with SPY over years. That asks whether a naive follower kept pace—not whether one filing action worked.
Compare investor track records →Read the score correctly
One yardstick, normalized for direction.
NEW / ADD: positive means the stock beat SPY after disclosure.
TRIM / EXIT: positive means the stock lagged SPY after disclosure.
Complete: a full 63-market-session window, so rows are comparable.
Completed 63-session windows only
What happened after each action?
These cohorts separate the four observable filing actions. They do not infer the manager's trade price, intent, or undisclosed exposure.
Additions
Broader sampleA manager increased an existing reported position materially.
- Complete
- 410
- Supported
- 43.9%
- Median vs SPY
- -1.3 pts
+84 still tracking
Trims
Broader sampleA manager reduced, but did not eliminate, a reported position.
- Complete
- 542
- Supported
- 53.7%
- Median vs SPY
- +1.8 pts
+97 still tracking
Exits
Broader sampleA reported position disappeared from the manager's next filing.
- Complete
- 159
- Supported
- 67.9%
- Median vs SPY
- +6.2 pts
+39 still tracking
Completed 63-session windows only
Did context change the signal?
These retrospective groups compare peer alignment and a manager's reported sequence. Same-quarter context can include filings published after the first manager disclosed, so it is not an available-at-first-disclosure trading signal. Sequence groups include every qualifying follow-on event in the eight-quarter window.
Same-side clusters
Growing sampleAt least three material decisions on the same stock pointed in one direction that quarter.
- Complete
- 103
- Supported
- 59.2%
- Median vs SPY
- +6.8 pts
+14 still tracking
Split-quarter decisions
Broader sampleTracked managers took material actions on both sides of the same stock in the same quarter.
- Complete
- 1,061
- Supported
- 50.6%
- Median vs SPY
- +0.3 pts
+216 still tracking
Reversals
Broader sampleA manager's next material action flipped from buying to selling, or vice versa.
- Complete
- 329
- Supported
- 55.9%
- Median vs SPY
- +2.8 pts
+80 still tracking
Continued builds
Growing sampleA buy-side material action followed an earlier buy-side action in the same stock.
- Complete
- 147
- Supported
- 39.5%
- Median vs SPY
- -1.8 pts
+37 still tracking
Continued reductions
Broader sampleA sell-side material action followed an earlier sell-side action in the same stock.
- Complete
- 204
- Supported
- 56.4%
- Median vs SPY
- +2.8 pts
+49 still tracking
Methodology and limits
Public-information clock. The measurement begins at the first market close after the filing became available—not at quarter end and not at an unknowable trade date.
Comparable primary sample. The table includes only completed 63-session observations. 242 newer decisions are still tracking; 0 do not yet have an observation window.
Material decisions only. NEW and EXIT qualify; ADD and TRIM require a 20% share change plus either 0.5% portfolio weight or $100M reported value.
Observation, not causation. The sample covers 8 reported quarters and a curated stock universe. Overlapping decisions can share the same stock return, so this is not a statistical claim of independence.
Selection rule. The 40 most widely owned mapped stocks in each archived prior quarter, plus the editorial Decision Ledger roster. Retrospective coverage: current manager membership and ticker mappings remain in use; archived holdings can include later amendments. This is not a fully point-in-time investable universe or a prospective test.
13F blind spots remain. Filings omit shorts, most derivatives, foreign listings, bonds, cash, exact trade dates, and the manager's thesis.
Fresh through Oct 2, 2026. Positive historical observations are research prompts, not forecasts or recommendations.