Are Two Highly Correlated Coins Suitable for Pairs Trading?

Trading Strategies

01 November 2026

Two coins being highly correlated is not a sufficient condition for pairs trading. Correlation describes the degree of short-term co-movement. Pairs trading requires long-term predictable reversion of the spread. These two things are answered by different statistical tests: correlation looks at whether returns rise and fall together, while pairs trading looks at whether the spread is stationary and tends to return to its mean. A trading pair with a correlation of 0.95 can still have a spread that keeps widening without reverting. In that case, doing pairs trading will cause losses on both sides.

OKX Exchange
A leading global cryptocurrency platform,suitable for both beginners and experienced traders.
New user benefit: 20% off trading fees upon registration!！

What Correlation Can and Cannot Tell You

High correlation means that two coins often rose and fell together during a certain past period. This is very common in the crypto market. The correlation between BTC and ETH has long fluctuated between 0.6 and 0.9, and they move almost in sync when the broader market rises or falls. If you only look at this number, you may feel that "their relationship is very stable."

But the profit logic of pairs trading is not "they will rise together." It is "when one rises too much and the other rises too little, the spread will shrink back." High correlation does not guarantee that the spread will shrink back. An analysis by Deribit gave a clear example: BTC and LTC are correlated, but they are not cointegrated. This means that if you go long LTC and short BTC, betting that the spread between the two will revert, this bet will not succeed in the long run.

Correlation measures consistency of direction. Cointegration measures the stationarity of a linear combination. The price of each asset can be non-stationary on its own, with trends or random walks. But some linear combination of them may be stationary: the spread fluctuates around a mean and tends to revert after deviating. This is the property that pairs trading needs.

The Real Test You Should Run: Cointegration, Not Correlation

A practical judgment process looks like this:

Step one, take price series from the same time window, not returns. The spread in pairs trading is built on price levels. Using returns to calculate correlation will miss the long-term relationship between price levels.

Step two, run a linear regression on the two price series to get a spread series, the residuals. Specifically, regress the price of A on the price of B, or the other way around, depending on which direction you want to go long or short. The residuals from the regression are the spread you would trade.

Step three, run a stationarity test on the residuals. The most common is the Augmented Dickey-Fuller (ADF) test. If the residuals are stationary, meaning the unit root hypothesis is rejected, it shows that the two prices have a cointegration relationship and the spread tends to revert. If they are not stationary, then even if the original price series are highly correlated, they are not suitable for pairs trading.

VectorBT's tutorial clearly states this judgment: when selecting pairs, cointegration is more important than correlation. Correlated assets can move in the same direction, but the price ratio may deviate greatly over time. Cointegrated assets have spreads that "pull back together." Academic research also uses the same framework, using cointegration tests and error correction models to screen crypto asset combinations suitable for pairs trading.

Step four, if the pair passes the cointegration test, then look back at correlation as a supplementary confirmation. High correlation but no cointegration: give up. Cointegration but very low correlation: also be cautious. The spread reversion may come from an accidental match of long-term trends rather than a stable economic relationship. Only when both are satisfied should you enter the strategy design stage.

Passing Cointegration Does Not Mean You Can Trade

There is a trap here that is easy to overlook. Cointegration tests are done on historical samples. You find a pair of coins that were cointegrated historically. You test data from 2021 to 2024 and get a p-value below 0.05. Then you run the strategy on 2025 data. The problem is that the cointegration relationship itself may have changed.

The fundamentals of the crypto market change much faster than traditional financial markets. Project upgrades, tokenomics adjustments, exchange listings or delistings, and regulatory events can all break the original price relationship within a few months. A Chinese analysis on Gate Square stated this risk very directly: changes in project fundamentals, industry structure adjustments, shifts in capital flows, or extreme market conditions can all break the original relationship. If the spread keeps widening instead of reverting, traders face losses on both sides and cannot wait indefinitely for the market to recover.

This means cointegration tests cannot be run only once. A reasonable approach is to rerun the test regularly using rolling windows or expanding windows, and treat "whether the relationship still holds" as one of the conditions for holding a position. Once the cointegration relationship is no longer statistically significant, or the spread breaks through historical extreme levels without signs of reversion, exiting is more reasonable than "waiting a bit longer."

Costs Will Eat Most of the "Apparent" Spread

Pairs trading requires opening two positions at the same time, so trading costs are double those of a single-direction strategy. In the crypto market, this is not just about fees.

A study reported by KuCoin tested 15 strategies on 4,990 DEX trading pairs. It found that even for the strongest signals, gross returns were about 6.5 basis points per trade, while round-trip costs including gas fees, slippage, and swap fees averaged 230 to 240 basis points. The gap between signal strength and trading costs was nearly 40 times. This study focused on DEXs. CEXs have a different cost structure, but the logic is the same: the spread reversion must be large enough to cover the costs of both legs of the trade.

For ordinary users doing pairs trading on a CEX, fees are calculated as two entries and two exits, plus possible slippage, especially when one of the coins has poor liquidity. The spread needs to revert to a sufficiently significant degree to have positive expected value. If the cointegration test shows that the standard deviation of the spread is very small and the reversion range is limited, then after deducting costs, it may not be worth executing at all.

A Practical Order of Judgment

If you have two coins that "look very similar" and want to judge whether they are suitable for pairs trading, follow this order:

First, run the cointegration test. Use price series, not returns. Use the ADF test on residual stationarity. If the p-value is not significant, give up directly, no matter how high the correlation is.

Next, look at the historical fluctuation range of the spread. After cointegration passes, calculate the standard deviation of the spread. Only if the spread has historically often deviated from the mean by more than 2 standard deviations and reverted within a relatively short time is there trading space. If the spread fluctuation is very small and cannot cover costs, it is also not worth doing.

Then check whether the liquidity of the two coins matches. Pairs trading requires both legs to enter and exit at the same time. If one coin has a thin order book, slippage will eat the profit from spread reversion and may even cause the execution prices of the two legs to deviate severely from expectations.

Finally, set exit conditions for when the relationship breaks down. It is not "the spread will definitely revert," but "if the spread does not revert to level Y within X days, or breaks through historical quantile Z, close the position." These conditions should be written down before opening the position.

High correlation only gives you a candidate list. The cointegration test is the screening threshold. After passing the threshold, costs, liquidity, and relationship stability will determine whether it is truly worth executing. Skipping cointegration and entering only based on correlation turns pairs trading into a simplified version of "two coins rise and fall together so I can just hedge casually." That simplified version does not hold statistically.

OKX Exchange
A leading global cryptocurrency platform,suitable for both beginners and experienced traders.
New user benefit: 20% off trading fees upon registration!！

Crypto Trading Strategies and Risk Management