---
title: Why the Market Regime Scores Have Decimals
canonical: "https://themacrodashboard.com/blog/why-market-regime-scores-have-decimals/"
pubDate: "2026-06-01T00:00:00.000Z"
updatedDate: "2026-06-01T00:00:00.000Z"
author: The Macro Dashboard
description: How four equal-weight evidence domains produce fractional scores and prevent a crowded group of equity indicators from dominating the regime.
categories: [Field Notes]
---

A score of 37.5 looks suspiciously precise until you know where it came from. It is not a 37.5% probability, and it does not mean 37.5 indicators voted for anything.

The model balances evidence inside four domains and gives each domain 25% of the final score. Dividing those domain weights among the available signals produces the decimals.

## The problem with raw vote counting

Imagine a model with eight equity indicators, two credit indicators, two rates indicators, and two funding indicators. If every indicator receives one equal vote, equities control more than half the result before the data arrive.

That setup is not diversified. It quietly gives equities most of the vote.

The problem becomes worse when the equity indicators are closely related. Small caps versus large caps, cyclicals versus defensives, high beta versus low volatility, and equal weight versus cap weight all tell useful stories. They are not four fully independent facts.

Goldman Sachs' discussion of [S&P 500 concentration](https://www.goldmansachs.com/intelligence/pages/is-the-sp-too-concentrated.html) helps explain why participation beneath the headline index matters. It does not mean every participation ratio deserves a separate full vote.

## Four domains, equal influence

The dashboard assigns each allocation signal to one primary domain:

1. participation and risk appetite;
2. credit and refinancing conditions;
3. rates, inflation pricing, and duration;
4. funding, the broad dollar, and financial conditions.

Signals are normalized inside their domain. Each domain then contributes 25% of the combined score, even when one domain has more available series than another.

Here is the arithmetic. If one of three available signals in a 25-point domain confirms a regime, its contribution can be about 8.3 points. Half of a 25-point domain is 12.5. Combine those pieces across four domains and decimals are unavoidable.

## A score is not a probability

A Goldilocks score of 37.5 does not mean there is a 37.5% chance that Goldilocks is true. It means Goldilocks received 37.5 points from the model's weighted confirmation structure.

The dashboard also publishes the winner margin, data coverage, domain agreement, and confidence. A high score with a tied rival still produces Mixed. A narrow lead under two points also produces Mixed.

The Federal Reserve's [Financial Stability Report](https://www.federalreserve.gov/publications/financial-stability-report.htm) uses the same broad habit of separating evidence into categories. Concentration in one category does not settle the whole question.

## Why not add every available indicator?

More data can make a model worse when the extra series repeat the same message. The BIS describes global liquidity as something that cannot be captured by [one indicator](https://www.bis.org/statistics/dataportal/gli.htm). The reverse is also true: fifty overlapping indicators do not create fifty independent observations.

The dashboard therefore favors explicit domains, fixed weights, and visible sensitivity checks. Removing one signal or one domain should produce an explainable change. Adding another equity ETF should not automatically make the model more risk-on.

AQR's [trend-following research](https://www.aqr.com/Insights/Research/Journal-Article/A-Century-of-Evidence-on-Trend-Following-Investing) and CFA Institute's [asset-allocation framework](https://www.cfainstitute.org/insights/professional-learning/refresher-readings/2026/principles-asset-allocation) both support the broader habit of defining a process before judging one result.

## What the number tells you

Read the regime scores as balanced evidence points. Then check the winner margin and confidence.

The decimal is simply the residue of honest weighting. It shows that the model divided domain weight among available evidence instead of pretending every ticker was independent.