Health Score

Methodology · Portfolio structure, 0 to 100

The Health Score is one number, 0 to 100, that measures the structure of a crypto portfolio: how its pieces relate to each other. It is a weighted average of five checks, with same-bet correlation carrying the largest weight at 35%. It does not predict returns, and it never says what to do.

What it measures, and what it does not

The score reports what a portfolio is. It does not measure alignment with the current market (that is the Regime Fit Score), it does not forecast price, and it produces no instruction.

A letter grade from A to F and a one-word band accompany the number for readability. The verdict sentence is generated from whichever check scored weakest, so the headline always points at the real problem rather than an average of everything.

The five checks and their weights

1. Same-bet · 35%

The largest weight, because hidden sameness is the most common structural failure in crypto portfolios. Pairwise correlations between holdings are used to compute the effective number of independent positions, one divided by the sum of weight-weighted pairwise correlations. One effective bet scores 0. Four or more independent bets scores 100.

Holdings are also grouped into named clusters, where a coin joins a cluster once its correlation with the cluster seed reaches 0.8 or higher. That is what lets the read say “8 coins, 2 bets” and name the two bets rather than just asserting a number.

2. Concentration · 20%

The Herfindahl-Hirschman index over position weights. An HHI of 1,200 or lower scores 100, an HHI of 5,500 or higher scores 0, linear in between. For scale, a portfolio that is 67% one coin has an HHI above 4,500.

Concentration sits below same-bet on purpose. Nominal weight spread across many tickers can look diversified while every holding moves as one, and correlation-based same-bet is the truer measure of that.

3. Sector spread · 15%

How many of the ten tracked crypto sectors the portfolio holds at 5% weight or more. Four or more scores 100. When a strong sector is absent, the check names it.

This check is calibrated to be fair within crypto. Majors (BTC, ETH) and stablecoins are the core of a book rather than a missed narrative, so spread is judged on the slice beyond that core. A book that is 75% or more core floors this sub-score at 60, and 50% or more core floors it at 40. A majors-heavy portfolio gets criticised by the same-bet and concentration checks, where the criticism is true, not here, where it is not.

4. Stability · 15%

Starts at 50. Rises with the share held in majors and stablecoins. Falls with the share held in assets under $100M market capitalisation, and with untracked assets. Small caps move hardest in both directions, and this check measures how much of the portfolio rides that.

5. Volatility · 15%

The weighted average absolute 30-day move of the holdings. 20% or less scores 100, 60% or more scores 0, linear in between.

The single-factor cap

The composite carries a cap. When the effective number of independent bets is below 2, the overall score is capped: 42 at one effective bet, rising to roughly 65 as it approaches two. A portfolio that is structurally one bet cannot grade as healthy however many tickers spread its nominal weight across. Portfolios with two or more effective bets are never capped.

Grades and bands

  • 85 or higher · A · Sound
  • 70 to 84 · B · Solid
  • 55 to 69 · C · Uneven
  • 40 to 54 · D · Strained
  • Below 40 · F · Fragile

Where the numbers come from

Correlations and beta to Bitcoin are computed from real daily returns over the trailing year, taken from a daily price-history capture of the top coins by market capitalisation. A coin needs at least 60 overlapping daily returns before its measured number is trusted.

Below that threshold, and for any coin outside the tracked universe, the engine falls back to an asset-class correlation model and a per-class beta lookup. So an established portfolio is graded on how its holdings have actually behaved, while a brand-new listing degrades to a defensible estimate instead of producing noise.

Known limitations, stated plainly

  • The asset-class fallback is calibrated on cycle behaviour, not a per-coin regression, and applies only to coins with too little price history to measure. Coins that cannot be classified are treated as high-correlation alts, which errs toward overstating risk rather than hiding it.
  • The score knows nothing about cost basis, time horizon, income or anything else about the holder. It measures the portfolio, not the investor. That is deliberate: the analysis is mechanical and impersonal.
  • Market data covers the top coins by market capitalisation. Holdings outside that universe reduce the stability check and are labelled untracked.
  • Score history accumulates from the day a portfolio is first analysed. There is no backfilled history for a portfolio entered today.

Legal posture

Every sentence the Health Score produces is descriptive. It reports weights, correlations and market data. It never produces buy, sell, trim, rotate, rebalance or any other directive aimed at the reader. CryptoArrow publishes impersonal analytics and is not an investment adviser.

Run it on your own holdings with the free Portfolio Risk X-Ray. No signup, no wallet connect.

Mechanical analysis. Not investment advice. See the Terms of Service.