Why we built a 7-layer signal engine

Most retail trading platforms ship a handful of indicators stitched together with a moving-average cross, and call that "intelligence." We took a different route.

CentoFlow models conviction as a probabilistic belief that updates as new evidence arrives. We split the world into seven independent signal layers — macro regime, technicals, options flow, news sentiment, positioning (COT), geopolitical events, and SEC filings — then fuse them via Bayesian updating into a single 0–100 conviction score per asset.

Why seven layers, not three?

Because no single layer is right all the time. Macro is wrong at turning points. Technicals lag in regime shifts. News sentiment reacts. Options flow front-runs but is noisy. By weighting each layer dynamically — based on its historical accuracy in the current regime — we get a score that is more honest about the things it doesn't know.

Why Bayesian fusion instead of an average?

A simple average treats all signals as equally important and equally independent. They aren't. A bullish technical signal during a hawkish Fed cycle is meaningfully less informative than the same signal during a dovish pivot. Bayesian updating captures that. It also gives us a proper uncertainty estimate, which is what makes the conviction score actually useful for position sizing — not just for picking direction.

Why this matters for retail

Institutional desks have analysts dedicated to each of those seven layers. Most retail traders have a single chart and a Twitter feed. We built CentoFlow to close that gap.

You don't need a Bloomberg terminal or a research budget to think probabilistically about the market. You just need the right tools.