Correlation and diversification for traders
Holding five positions is not diversification if all five move together. Correlation measures how closely markets move with each other. This guide shows how to calculate it, how to read it and how it should change your risk.
Key takeaways
- Correlation runs from −1 to +1 and must be measured on returns, not prices.
- Large crypto coins tend to move with Bitcoin; treat them as one exposure.
- Forex pairs that share a currency often duplicate the same bet.
- Combined risk depends on correlation, not on the number of positions.
- Correlations change and tend to rise in sell-offs.
What correlation measures
The correlation coefficient runs from −1 to +1. A value of +1 means two markets’ returns move perfectly together; −1 means they move perfectly opposite; 0 means there is no linear relationship. Most real pairs sit somewhere in between, and the value changes over time.
Correlation = Covariance(RA, RB) ÷ (σA × σB)
Use returns, not prices
Two markets that both rose over a year will show a high correlation of prices even if their daily moves were unrelated, simply because both lines go up. This is called spurious correlation. The correct method measures the correlation of returns: daily percentage changes, or log returns as above. The compare tool on TradeProfy computes correlations from daily log returns on dates that both markets share.
Align the dates
Crypto trades every day; forex reference rates are published on working days only. To compare Bitcoin with EUR/USD you must use only the days where both have a price, otherwise weekend crypto moves are matched against nothing. The compare tool aligns the series automatically before calculating.
Reading the numbers
| Correlation | Plain reading | Diversification effect |
|---|---|---|
| 0.8 to 1.0 | Move almost together | Very little; treat as one position |
| 0.5 to 0.8 | Strong link | Some, but losses often arrive together |
| 0.2 to 0.5 | Moderate link | Meaningful |
| −0.2 to 0.2 | Little linear link | Strong |
| Below −0.2 | Tend to move opposite | Acts partly as a hedge |
Crypto: one big factor
Most large crypto coins have historically shown high correlations with Bitcoin, often in the 0.6–0.9 range over daily returns. When Bitcoin falls sharply, most coins fall with it. Holding Bitcoin, Ether, Solana and three other large coins is therefore much closer to one large crypto position than to six independent bets. Run the compare tool with BTC, ETH, SOL, XRP and DOGE over one year to see the current values.
Forex: shared currencies
Every currency pair contains two currencies, and the US dollar is on one side of most trades. According to the Bank for International Settlements’ 2025 triennial survey, the dollar was on one side of 89% of all foreign exchange trades. Pairs that share a currency are often linked:
- EUR/USD and GBP/USD have historically been positively correlated, because both are priced against the dollar and the euro and pound often move together against it.
- EUR/USD and USD/CHF have often been strongly negatively correlated: the dollar is the base in one and the quote in the other, and the euro and franc tend to move similarly.
- Being long EUR/USD and short USD/CHF at the same time is largely the same bet twice: short the dollar.
Why correlation matters for risk
Suppose you risk 1% on each of three trades. If they were independent, the chance of all three losing together would be moderate. If they are highly correlated, they behave like one trade risking 3%. A losing day then costs three times what you planned for “one idea”.
Worked example
Two positions each have a daily volatility of $200. The volatility of the combined position is √(200² + 200² + 2 × ρ × 200 × 200). With ρ = 0 it is $283; with ρ = 0.5 it is $346; with ρ = 0.9 it is $390, almost the same as one doubled position ($400). With ρ = −0.5 it falls to $200. Correlation, not the number of positions, decides how much risk you hold.
Practical rules
- Group correlated positions. Treat positions with correlations above about 0.7 as one exposure and give the group one risk budget.
- Look for different drivers. Diversification comes from markets with different causes of movement: a commodity currency, a safe-haven currency, an equity index, a crypto asset.
- Check net currency exposure. List every forex position as long or short of each currency and add them up. Hidden doubling is common.
- Expect correlations to rise in a crisis. In sell-offs, many assets fall together as traders cut risk everywhere at once. Diversification that looked good in calm markets often weakens exactly when needed.
Correlation trading in practice
“Correlation trading” is used for two different things. One is a pairs trade: two instruments that normally move together drift apart, and the trader buys one and sells the other, expecting the gap to close. The other, more common in a retail account, is risk control: checking that three open positions are not really one bet. Both start from the same number, measured on returns over the same dates, and both fail the same way when the correlation changes. A pairs trade also pays two sets of charges, so the expected convergence has to be larger than the round-trip cost on both legs.
Lower-correlation strategies: how much risk they remove
Hold two positions of equal size, each with 20% yearly volatility. The risk of the pair depends only on how they move together: combined volatility = 20% × √((1 + correlation) ÷ 2).
| Correlation | Volatility of the pair | Risk removed |
|---|---|---|
| +0.9 | 19.5% | 3% |
| +0.7 | 18.4% | 8% |
| +0.5 | 17.3% | 13% |
| +0.2 | 15.5% | 23% |
| +0.0 | 14.1% | 29% |
| -0.3 | 11.8% | 41% |
At 0.9 the second position removes almost nothing; it is the same bet twice. The gain comes below about 0.5, and it is there that a lower-correlation strategy earns its place: a trend system next to a mean-reversion system, an index position next to a commodity, or the same idea on a different timeframe. Measure the correlation of the strategies’ own daily results, not of the instruments they trade, because two systems on one market can still win and lose on different days.
Correlation is not stable
Correlations change with the period measured. A one-month window reacts to recent events but is noisy; a one-year window is steadier but slow. Comparing several windows is the honest approach. In the compare tool, switch between 1M, 3M, 6M and 1Y: if the correlation between two markets jumps around, do not rely on any single value.
Correlation and causation
A high correlation does not mean one market drives the other. Two coins may both follow Bitcoin; two currency pairs may both follow the dollar; a stock index and a crypto asset may both follow general risk appetite. For risk management the cause does not matter much, but for trading ideas it does: a correlation with no reason behind it is likely to break.
Using correlation with the planner and risk lab
When you plan several trades in the Trade Planner, add their risks together for any group of correlated markets and keep the total within your daily loss limit. To see how a run of simultaneous losses affects an account, model the group as one larger risk per trade in the Risk Lab.
Calculating correlation by hand
A small example shows the mechanics. Over five days, market A returns +1%, −2%, +3%, 0% and +1%; market B returns +2%, −1%, +2%, −1% and +1%. The averages are 0.6% and 0.6%. The deviations from the average are A: 0.4, −2.6, 2.4, −0.6, 0.4 and B: 1.4, −1.6, 1.4, −1.6, 0.4. The products are 0.56, 4.16, 3.36, 0.96 and 0.16, which sum to 9.2. The squared deviations sum to 13.2 for A and 9.2 for B. Correlation = 9.2 ÷ √(13.2 × 9.2) = 9.2 ÷ 11.02 = 0.83. Five days are far too few to trust, but the arithmetic is the same for 250 days.
Building a net exposure table
For forex traders, a currency exposure table reveals hidden concentration. List each position and split it into its two currencies:
| Position | EUR | GBP | USD | CHF |
|---|---|---|---|---|
| Long 1 lot EUR/USD | +1 | −1 | ||
| Long 1 lot GBP/USD | +1 | −1 | ||
| Short 1 lot USD/CHF | −1 | +1 | ||
| Net | +1 | +1 | −3 | +1 |
Three separate-looking trades add up to a short dollar position three times the size of each one. If the dollar rallies on one piece of US news, all three lose together.
Diversification across asset classes
Real diversification comes from markets driven by different things. A rough map: equity indices follow growth and earnings; government bonds follow inflation and central-bank expectations; safe-haven currencies such as the yen and Swiss franc often strengthen in market stress; commodity currencies such as the Australian and Canadian dollars follow commodity prices and global growth; crypto assets follow their own adoption and liquidity cycles but have at times moved with risk appetite. None of these relationships is fixed, which is why measuring them regularly matters more than memorising them.
Key terms
- Correlation coefficient
- A number from −1 to +1 describing how two return series move together.
- Log return
- The natural logarithm of today’s price divided by yesterday’s.
- Spurious correlation
- An apparent link created by two trends rather than related moves.
- Net exposure
- The sum of long and short positions in each currency or asset.
- Rolling window
- The period over which a statistic is measured, moved forward one day at a time.
Related guides
- Average True Range: volatility-based stops and position sizingATR does not tell you direction. It tells you how much a market typically moves in one…
- Options Greeks explained with worked examplesThe Greeks measure how an option’s price reacts to changes in the underlying price, time,…
- How to backtest a trading strategy without fooling yourselfA backtest runs a set of rules over past prices to see what would have happened. Done…
Frequently asked questions
What is a good correlation for diversification?
Lower is better. Positions with correlations below about 0.3 add real diversification; above about 0.7 they behave like one position.
Why calculate correlation from returns instead of prices?
Prices that both trend upward appear correlated even when their daily moves are unrelated. Returns remove the trend and measure how the moves line up.
Is Bitcoin correlated with stocks?
The relationship has changed over time. In some periods Bitcoin’s daily returns have been noticeably correlated with US technology shares; in others the link has been weak. Measure it over the period that matters to you.
Can correlation change quickly?
Yes. Correlations often rise sharply in market stress, when many assets fall together.
Which forex pairs are most correlated?
Pairs that share a currency and have similar partners, such as EUR/USD and GBP/USD, often move together. EUR/USD and USD/CHF have often moved opposite to each other.
