How CORVIX actually works

CORVIX scores eight secular themes against seven weighted subsignals, reads the current macro regime, and shows the reasoning behind every number instead of a black-box output. This page is the long version of that, including what the system does not do.

Seven subsignals, IC-weighted

Every theme and every regional read is built from the same seven subsignals. Each one is weighted by its information coefficient, a measure of how reliably it has actually predicted outcomes historically, so a subsignal that has been noisy gets shrunk toward zero influence rather than treated the same as a subsignal with a real track record. No single input can dominate the composite score on its own.

These combine into one composite score from 0 to 100 per theme, and the same framework is used to read the overall regime, so a theme score and a regime score are directly comparable rather than built on different logic.

Conviction bands

The composite score maps to a conviction band, a plain-English label for how strongly the evidence currently supports a theme:

A band is a summary, not a black box: every band change is traceable back to which subsignal moved and by how much.

Falsify conditions

Every theme ships with a plain-English condition that would prove the thesis wrong, written down in advance and reviewed on a quarterly cycle. When a theme hits its falsify condition, it moves down a band automatically. This exists specifically to stop a stale conviction from just sitting there unquestioned because nobody wanted to revisit it. Themes are graded against the condition, not against how they're doing overall, which is what keeps the check honest.

Point-in-time backtesting

A backtest is only useful if it reflects what an investor could have actually known and done at the time. CORVIX's backtester uses only data through the prior period at every step, no look-ahead, and applies transaction costs on turnover so a strategy that trades often doesn't look artificially cheap. The signal overlay can be toggled on or off to isolate exactly what a given signal contributed to risk-adjusted return, separate from the underlying benchmark.

Update cadence

Core signals refresh hourly. Market prices refresh daily after major exchanges close. Structural and policy subsignals, which by nature don't change hour to hour, are reviewed on a fixed quarterly schedule rather than being left stale indefinitely or refreshed so often that noise gets mistaken for signal.

What CORVIX does not do

This section exists because a methodology page that only lists strengths isn't a complete one.