What Is Bill Ackman Buying Next? How Prior Moves Predicts 13F Moves Before They File
By the time a 13F lands in your inbox, the trade inside it is old. The filing tells you what Bill Ackman’s Pershing Square owned at the end of last quarter, and it can arrive up to 45 days after that quarter closed. So the “news” you are reading describes a position that has been on the books for weeks, sometimes months.
Prior Moves takes a different starting point. Instead of reporting the filing after it drops, it estimates the next likely buy before the filing is public, using a model trained on this one investor’s own history. Below is how that works, what the number means, and the parts we get wrong and publish anyway.
Why a 13F is already old news when you read it
A 13F is a quarterly snapshot. Large institutional managers report their U.S. equity holdings to the SEC, and the deadline is 45 days after the quarter ends. A position opened in early January can sit hidden until mid-May.
That lag is the whole problem with trading off filings. The crowd reads the same 13F on the same day, the obvious names have already moved, and you are acting on a portfolio photo taken a quarter ago. Following filings is following footprints in the snow after the person has walked on.
The window we care about is the gap between when an investor likely acts and when the rest of the market gets to see it on paper. That gap is where a forecast can be useful, if it is honest about its limits.
How a per-investor model learns one investor’s style
Most trackers pool every famous investor into one blended signal. Prior Moves does the opposite. There is a separate model for each of the 44 investors tracked, and each one is trained only on that investor’s own 13F filings going back five or more years, plus roughly 250 public signals (sector tilts, position sizing patterns, turnover, how a manager tends to scale into a name).
For Ackman, that means the model studies how Pershing Square actually behaves: a concentrated book, a handful of high-conviction positions, slow turnover, a preference for large quality businesses. It does not average Ackman together with a quant shop that holds 150 names or a macro fund that rotates every quarter. The point of a per-investor model is to capture the fingerprint of one decision-maker rather than the blur of forty.
From all of that, each investor’s model produces a single conviction number for candidate names, the PriorScore. We will get to how to read that honestly.
What “predicted next buy” means, roughly six weeks before the filing
The headline output is a predicted next buy: the name the model thinks this investor is most likely to have added, surfaced about six weeks before the 13F that would confirm it becomes public. Pershing Square’s current top predicted buy and its PriorScore are shown live on the Bill Ackman mirror, and they update each filing cycle.
Here is what that does not mean. It is not a tip. It is not a claim that Ackman has bought the name, or will, or that you should. It is a model’s estimate of where this investor’s pattern points, stated as a probability, ahead of the paperwork that would settle the question. When the next 13F files, you can check the prediction against the real filing yourself. That is the entire test, run in the open.
Prior Moves never touches your money, never holds custody, and never places a trade. It describes what a model predicts. Any decision and any trade is yours, at your own broker.
What the model got wrong last quarter
The uncomfortable part, kept in plain sight: the model misses, and we publish the misses. For the Ackman mirror, last quarter’s predicted buys and what actually showed up in the filing are on the live mirror and the running track record. Some predictions matched. Some did not. We do not quietly drop the bad calls or rescore them after the fact.
Across the full consensus mirror (the model that combines the high-conviction long-only investors, with the quant and multi-strat shops deliberately left out), the backtest shows an edge of about +1.8 points per quarter versus the S&P 500 over 47 complete quarters from 2014 to 2026, net of costs. That edge is directional, and it is not statistically significant: the t-statistic is about 1.45 and the 95% confidence interval includes zero. In plain terms, the result leans positive, and the data is not yet strong enough to rule out that the true edge is nothing. We lead with that sentence because hiding it would make every other claim here worth less.
How to read the PriorScore conviction number honestly
The PriorScore is a probability, not a price target and not a rating. A name with a 0.92 score is one the model considers a strong fit for how this investor tends to act. A name at 0.55 is a coin-flippier guess.
- A high PriorScore says the prediction fits the investor’s historical pattern well. It does not say the stock will go up.
- Per-investor samples are small. A model trained on one manager’s filings has fewer data points than the combined mirror, so individual scores are noisier and individual-investor edges are not statistically significant either.
- The score answers one narrow question: how likely is it that this investor adds this name, given how they have behaved before. Treat it as a directional read on a manager’s style, weighted by your own judgment, never as instruction.
See the live mirror
The live page carries the current predicted buys, PriorScores, and the running record of hits and misses.
- See the live Bill Ackman mirror.
- Compare Prior Moves against 18 other 13F trackers on price, custody, regulatory posture, whether they predict next buys, and whether they publish their misses: the comparison page.
The core is free. You read what the model predicts; you decide what, if anything, to do with it. This is research, not investment advice, see the full disclaimer.