Stationarity
Published Last updated
Key Takeaway
Statistical property where cryptocurrency price series fluctuate consistently around a stable mean without trending, enabling predictable mean-reversion trading and reliable statistical inference.
Learn These First
What Is Stationarity?
Statistical property where cryptocurrency price series fluctuate consistently around a stable mean without trending, enabling predictable mean-reversion trading and reliable statistical inference.
How Stationarity Works
Frequently Asked Questions
How do I test whether my cryptocurrency pair is stationary enough for mean-reversion trading?
Use Augmented Dickey-Fuller test on your cryptocurrency pair spread. Run ADF test on the series representing the spread between your two correlated assets. If p-value returns below 0.05, conclude stationarity with 95% confidence; if p-value exceeds 0.05, conclude non-stationarity. Stationarity p < 0.05 suggests the pair is suitable for mean-reversion trading. Additionally, examine autocorrelation function (ACF) plots confirming spread autocorrelations decay toward zero (stationary signature). Visual inspection of spread charts helps: stationary spreads oscillate around constant mean; non-stationary spreads drift without mean reversion. Combine formal ADF testing with visual inspection for robust stationarity assessment.
Why can't I just trade individual cryptocurrency prices if they don't pass stationarity tests?
Individual cryptocurrency prices like Bitcoin are non-stationary: they trend upward over years without reverting to historical means. Non-stationary series violate statistical assumptions underlying regression and hypothesis tests, producing false signals. Trading systems built on non-stationary data estimate parameters unreliably—backtests appear profitable while live trading produces losses. Additionally, non-stationary prices lack mean-reversion characteristics, making mean-reversion strategies inappropriate. Momentum strategies sometimes work on trending non-stationary data, but most mean-reversion approaches fail. Professional traders avoid direct price trading, instead using pairs where spreads achieve stationarity, enabling reliable statistical inference and profitable mean-reversion strategies.
What happens if market conditions change and my previously stationary cryptocurrency pair becomes non-stationary?
Stationarity is regime-dependent; shifts in correlation structure, volatility, or fundamental relationships cause stationary pairs to transition to non-stationary behavior. Walk-forward ADF testing detects this: re-run ADF tests on rolling windows (quarterly updates) monitoring p-values. Increasing p-values signal deteriorating stationarity. When p-values exceed 0.05, your pair has transitioned to non-stationarity; mean-reversion strategies built on the relationship should pause or cease. Many traders discover this too late, trading non-stationary pairs unknowingly. Vigilant traders continuously monitor stationarity through periodic ADF retesting, retiring pairs before stationarity fully collapses. Market regime monitoring enables proactive pair management rather than reactive recovery from failed strategies.
Common Misconceptions About Stationarity
Individual Bitcoin prices are stationary because they fluctuate around an equilibrium price.
Bitcoin prices are non-stationary; they trend without fixed equilibrium. What appears like equilibrium are localized ranges (sideways consolidation) between trends, not stationarity. Stationarity requires constant mean over entire period; Bitcoin's mean increases (trends upward) over years. Individual non-stationary prices fail Augmented Dickey-Fuller tests (p-values exceed 0.05). Confusing price oscillations with stationarity causes traders to build failed strategies on non-stationary foundations. Use pairs trading where spreads achieve stationarity rather than individual price trading.
If my ADF test shows p-value 0.03, my cryptocurrency pair is definitely stationary and will remain so indefinitely.
ADF test results are time-dependent, not permanent. P-value 0.03 confirms stationarity on tested data period only. Cryptocurrency market conditions shift; previously stationary pairs can transition to non-stationary as correlations break down. A pair stationary from 2022-2023 might become non-stationary by 2024. Professional traders retested stationarity continuously on rolling windows, detecting deterioration early. Static ADF results from past analysis provide no guarantee about future stationarity. Treat stationarity as provisional ongoing assessment requiring continuous monitoring, not permanent discovery.
High correlation between two cryptocurrencies means their spread is stationary and suitable for mean-reversion trading.
Correlation and stationarity are distinct properties. Two assets can be highly correlated (move together) yet both contain unit roots, making their spread non-stationary. Correlation measures relationship strength; stationarity measures whether deviations revert to equilibrium. Test spreads directly through ADF testing for stationarity confirmation. Don't assume correlation implies stationarity; verify through formal statistical testing. Many traders confuse these concepts, building trading systems on correlated non-stationary pairs that fail immediately.