title: "why no priors" slug: "why-no-priors" date: "2026-04-12" updated: null type: "general" excerpt: "A uniform prior is flat. A posterior, after evidence, is sharpened. This is a public notebook about doing that work out loud." tags: ["launch", "meta"] readTime: 4 draft: false
A uniform prior is flat. You assign equal probability to every outcome because you don't yet know any better. As evidence accumulates, the distribution sharpens — some regions tighten, others collapse, and eventually what started as an even field of possibilities turns into a pointed, defensible posterior. That movement, from flat to sharpened, is the only thing this site is about.
— —
no priors is a personal research and trading blog. There is a bot behind it — a systematic options engine that scans a universe of liquid U.S. equities every morning, scores each ticker across eight dimensions, and writes trade cards for the signals it believes in. It has been running long enough to have opinions, short enough that the opinions are still being tested against reality. The site is the place where I write down what it does, how it does it, and when it fails — not as a sales channel, not as a newsletter funnel, just as the running log of a project I find worth taking seriously.
The name is the thesis. I don't want to start a trade, a post, or a backtest by importing a belief about what should work. I want to start with a flat prior and let the evidence do the shaping. Sometimes that produces uncomfortable results — a factor I expected to matter doesn't, a rule I thought was clever turns out to be a wash once you isolate it. Those are the interesting outcomes. They're also the ones that don't survive in environments where the author has already committed to a narrative.
what lives here
Four kinds of posts, roughly. Scan recaps cover what the bot saw this week — the top five scores, one ticker that surprised me, a sentence about what I'm watching next. Methodology deep dives are the long ones: how a scoring dimension was built, why a weight is what it is, what changed after a backtest. Postmortems are for when something breaks — a trade the system was confident in that went sideways, a bug in the scoring code, a mismatch between predicted win rate and realized win rate. And research notes are short observations I want to log publicly without committing to a full writeup.
None of these are a recommendation to trade anything. The ledger page shows what the system did, not what you should do. If that distinction doesn't feel load-bearing to you, this probably isn't the site you're looking for.
what doesn't
No hot takes on macro. No "AI is going to change everything" posts. No screenshots of TradingView where I drew the trendline after the move happened. If the finding is strong, the numbers make the case on their own, and the only adjective I'll reach for is the one that describes the data.
I'll be wrong on this site. The goal is to be wrong in specific, recoverable ways — the kind of wrong you can look back at a year later and see exactly what you got tangled up on. Vague wrongness doesn't teach you anything. Quantitative wrongness is the whole point.
— —
Three posts go up today. The first is this one. The second is a tour of the scoring engine's eight dimensions — what each one measures, roughly how it's computed, and why it earned its weight. The third is the story of replacing the original regime score, a satisfying little case study in what happens when you actually look at your factors under a backtest rather than trusting the intuition that built them. After that, cadence is one post a week, alternating between scan recaps and longer pieces, with research notes dropped in when a single chart is enough.
If you got here from somewhere and the project sounds interesting, the best way to follow along is the RSS feed. There is no email list yet. There may be, eventually. What there will not be, ever, is a Discord with trade alerts and a monthly subscription fee. The bar for any extension to the site is that it has to serve the reader who's here for the work, not the reader who's here to be sold to. If I can't pass that test, the extension doesn't ship.
Thanks for coming.