Autopilot lets you copy an investor after their filing is public. Prior Moves ranks the same move about six weeks sooner, for 58 investors across five markets. Names in our top-ranked decile get bought 38.8% of the time against a 12.5% base rate. You place the trades at your own broker; we publish the playbook, disclosed record and predicted next move.
Model quality, for the technically minded: the median investor model scores 0.659 AUC on held-out quarters, across 49 scored investors. It splits: 0.649 on names an investor has never held, 0.539 on buy-backs of names they used to own, measured on the 30 investors carrying at least 20 of each. How it is measured
Every 13F from 58 tracked investors, plus ~250 public signals: insider buys, congressional trades, activist stakes, options flow, news tone.
Each investor gets their own model, trained on their own history, predicting their next new position, scored as a probability, published with its as-of date.
Export a copy sheet (target weights and a dollar split) and place it at your own broker. We never execute and never hold your money.
A 13F shows what a manager held 45+ days ago. Copy-trading products mirror that stale snapshot. We model what shows up in the next filing, about six weeks before it is public.
The Top-15 basket's simulated spread against the S&P 500 was 2.2 points per quarter net of costs across 35 backtested quarters, and the 95% confidence interval includes zero. With t ≈ 1.31, this is a small directional research signal, not proof. Most of that average is a single quarter (2026Q1, about half of it two AI-hardware names marked at a local peak); leave it out and the edge is 0.9 points. We publish that sentence on purpose.
If your question is “will this make me rich?” then no one can promise that, and we won’t.
No. Prior Moves is a research publication. We publish model portfolios and probabilities; we never execute trades, hold money, or know your situation. What you do at your broker is your decision.
A 13F is filed up to 45 days after quarter end. Our predictions publish about six weeks before the filing is public, each with an as-of date and the filing due date.
It often is: a P 0.68 pick should fail about a third of the time, and we grade ourselves on exactly that calibration, published per conviction tier on the track record page.
Not yet. +2.2 pts/quarter over 35 backtested quarters with t ≈ 1.31. The 95% confidence interval includes zero, and most of the average is one quarter (2026Q1); excluding it the edge is 0.9 pts. It is a directional signal, and we would rather under-claim.
The research is free. A paid weekly digest and partner licensing are the business; your portfolio never is. We hold no license to your money because we never touch it.
The predicted next buys, fresh congressional trades, and the week’s catalysts. One email a week. Honest numbers, never advice.