I reconstructed Bill Ackman's convictions from 2 years of public data. Now I run the news through them.

I’ve been working for a while on an idea: can you reconstruct how a famous investor builds his convictions, and then use that to understand which news actually matters to his portfolio?

Starting with Bill Ackman, I scraped roughly the last 2 years of public material I could get: filings, interviews, letters, public statements, portfolio information. Then manually cleaned and labelled the corpus.

I wanted an engine that understands the drivers behind his convictions. How strong are they? What would make him change his mind? Which stocks does each conviction apply to? What would reinforce or weaken it?

And then when news comes in: **which part of that investment thesis actually moved?**

The hard part was stability. If I run the same evidence 3 times against the same investor, I need roughly the same conclusion, for the same reasons. Otherwise the investor apparently changes philosophy overnight.

So a lot of the work went into keeping the underlying representation stable. I’m using some of my own equity-research methodology in the middle.

**Link**: [conviction.ekheon.com](https://conviction.ekheon.com/)

Giving it away for free in case people find it useful. The demo currently has two scores:

* **Source trust:** 0–5, based on the type of source.
* **Public evidence:** −5 to +5, depending on whether new information challenges or reinforces a conviction.

The more interesting experiment is to see how different investors owning the same stock decode new information in their own distinct way. And also to run the engine using my own portfolio and investment convictions.

**EDIT**:
Chill down, not claiming Ackman is the best investor. I started with him for purely technical reasons: he has the most comprehensive corpus of publicly available data, as opposed to Buffett or Lynch. I'll probably pick Dan Loeb, Jeff Smith or David Einhorn as better examples.