Crypto Correlation Matrix: How to Read Asset Co-Movement

Why correlation is the foundation of every portfolio

Most retail traders pick coins by narrative or thesis. That's how exposure gets stacked: long BTC, long ETH, long SOL, long a couple of DeFi tokens, long a memecoin. On paper this looks like five different bets. In practice it's one bet — they're all 0.7+ correlated to BTC, which means the portfolio is essentially levered Bitcoin with extra volatility.

The correlation matrix is how you see this clearly. It's a square grid showing every pairwise correlation across the assets you're considering. A glance tells you which coins are independent risks and which are duplicates of an exposure you already own.

If you're constructing a portfolio, sizing positions, building a hedge, or running a pair trade — you start here.

What's correlated right now

The matrix below pulls live data from the Sharpe correlation engine — 30-day Pearson correlations of daily log returns across BTC, ETH, SOL, BNB, XRP, and DOGE. Green cells are positively correlated, red cells are inversely correlated, and shades indicate magnitude.

A few things to notice:

• The diagonal is always 1.0 — every asset is perfectly correlated with itself.
• The matrix is symmetric — corr(BTC, ETH) = corr(ETH, BTC).
• In a typical risk-on regime, the off-diagonals run 0.7–0.95 across the top 6 — meaning if you're long all six, you're effectively long one risk factor with leverage.
• During narrative regimes (AI agents rallying, memecoin season, ETH ETF flows), specific pairs decorrelate temporarily — that's the signal for sector rotation trades.

How the math works

Pearson correlation measures the linear relationship between two return streams. The formula:

r = Σ((x − x̄)(y − ȳ)) / √(Σ(x − x̄)² × Σ(y − ȳ)²)

Where x and y are the daily log return series, x̄ and ȳ are their means, and r is bounded in [−1, +1].

Why log returns and not simple percentage returns? Three reasons:

1. Additive over time. A 10% gain followed by a 10% loss is not zero in simple returns (1.10 × 0.90 = 0.99, a 1% loss). In log returns, ln(1.10) + ln(0.90) ≈ 0 — properly additive.
2. Symmetric. A 50% gain and a 50% loss have equal-magnitude log returns. In simple returns, the loss is more impactful (you need 100% to recover from −50%).
3. Statistically well-behaved. Log returns are closer to normally distributed than simple returns, which makes Pearson correlation a meaningful estimator.

Sharpe requires at least 70% data coverage in the lookback window for a pair to be included — meaning if a coin only has price data for 70% of the trailing 30 days, it makes the cut; below that, it's excluded because the correlation estimate would be noise.

Reading the cells

In risk-on crypto markets, almost everything sits in the 0.6–0.9 range. That's why crypto-only portfolios are so directionally biased — the structure of the asset class limits diversification within it.

Three practical uses

1. Position sizing. If you're running a 5-coin portfolio and the average pairwise correlation is 0.85, you have what looks like 5 names but the risk profile of ~1.4 names. A correlation-aware position size reduces individual coin weights to compensate. Standard rule: scale each weight by 1 / sqrt(1 + (n − 1) × ρ̄), where n is number of names and ρ̄ is average pairwise correlation. For 5 coins at 0.85 average, that's a 70% scaling — meaning your "5x diversified" portfolio should be sized like a 1.5x portfolio.

2. Pair trades. Two highly correlated assets that have dislocated in price are pair-trade candidates. ETH and SOL run a structural 0.85 correlation; if ETH is up 25% over 30 days and SOL is up 8%, the 17-point gap is the dislocation. Long the laggard, short the leader, target convergence. Use the matrix to screen for high-correlation pairs, then check price dispersion separately to find the dislocation.

3. Hedge construction. Want to hedge BTC exposure cheaply? Find something with high correlation but lower implied volatility, and short that. ETH and BTC run 0.85+ correlation but ETH options have higher IV, making short-ETH a more capital-efficient hedge than short-BTC for the same delta exposure. The matrix is the first-pass filter.

Multiple lookback periods tell different stories

The same pair can have wildly different correlations on different lookbacks. ETH and BTC running 0.95 on 30d but 0.78 on 1y means the two are in a tighter regime now than usual — useful for short-term trading, but don't assume the regime persists.

I check three lookbacks before any structural decision:

• 30-day — current regime. Useful for tactical decisions: which coins are decorrelating right now, where the rotation is.
• 90-day — trailing quarter. Smooths out single-news shocks; the picture you'd use for portfolio construction.
• 1-year — structural baseline. The "fundamental" correlation picture. Use this for strategic allocation decisions.

The Sharpe matrix supports all four (including 3-year for very long-term views) — switch the period in the full matrix to compare side by side.

Crypto vs traditional finance correlations

Crypto's relationship with traditional finance is the macro question that matters most. The headline correlations:

• BTC vs S&P 500 — 0.4 long-run, but spikes to 0.7+ in risk-off regimes. The "BTC is a hedge" thesis only holds outside crisis.
• BTC vs gold (PAXG) — 0.2 long-run. The "digital gold" thesis is partially supported but the correlation is weak.
• ETH vs Nasdaq (NDX) — 0.55 long-run. ETH trades like high-beta tech.
• DeFi tokens vs financial-sector ETFs — surprisingly low (≤ 0.3). Different risk premium structure.

The TradFi compare tool lets you pair any crypto asset with stocks, gold, bonds, or FX — useful for asset-allocation decisions and macro hedging.

Common mistakes

Trusting a 30-day correlation as structural. A coin that's 0.95 correlated to BTC over 30 days might be 0.5 over 1 year — the recent regime is unusual. Always check the 1-year view before betting that correlation will persist.

Ignoring volatility. Two assets at 0.9 correlation but with very different volatilities are not interchangeable. ETH might be 0.9 correlated to BTC but with 1.5x the realized volatility — long ETH short BTC isn't market-neutral, it's net-long beta. Correlation tells you direction; volatility tells you magnitude. Both matter.

Building a pair trade without checking the dislocation history. Two coins might run 0.9 correlation and currently be 20 points dislocated — but if 20 points is normal noise, there's no edge. Check the historical price dispersion alongside the correlation before you size the trade.

Forgetting that correlations regime-shift. Crypto correlation is not stable. A bear market spike, an ETF approval, a major hack — any of these can flip a 0.5 correlation to 0.9 within days. Don't bet that today's matrix is tomorrow's matrix.

Where to go from here

If you're building a portfolio, start with the correlation matrix and add coins one at a time, checking how each addition changes the average pairwise correlation. Anything that pushes the average above 0.7 is duplicate exposure.

If you're hunting pair trades, run the matrix on 30d, sort by highest positive correlation, then drill into coin-compare for each top pair to find the dislocations.

If you're worried about macro exposure, use the TradFi compare view to see how each crypto asset relates to the broader financial system. The correlations are counterintuitive in places — RWA tokens run lower correlation to TradFi assets than you'd expect.

The matrix updates daily. The methodology is open. The point is to give traders the structural map they need before they put capital to work — that's the entire reason the correlation engine exists.