Every trader eventually picks a side: mean reversion or momentum. The mean reversion camp believes prices oscillate around fair value and the edge is in fading extremes. The momentum camp believes trends persist and the edge is in following them. We tested both rigorously across five years of crypto market data and found that the debate itself is the wrong framing.
The Numbers
Our mean_reversion_bb strategy on 13 high-beta altcoins produces Sharpe ratios between 9 and 19.25, depending on the symbol and parameter configuration. These are exceptional numbers by any standard. Our momentum_rsi_macd strategy on the same altcoin basket produces Sharpe ratios between 3.5 and 7.8. Still strong, but roughly half the mean reversion performance in aggregate.
On first glance, mean reversion wins and the discussion is over. But aggregate numbers hide the regime-dependent behavior that matters for real portfolio management. The question is not which strategy produces a higher backtest Sharpe. The question is which strategy is making money when the other is losing.
Five Regimes, Two Strategies
We validated both strategies across five distinct market periods: 2021-2022 (Bull Run to Crash), 2022-2023 (Bear Market and Early Recovery), 2023-2024 (Recovery to New Highs), 2024-2025 (Consolidation and Range), and 2025-2026 (Recent). Each period has a different market structure, and the two strategies respond to those structures differently.
During ranging and consolidating regimes (2024-2025 especially), mean reversion dominates. Prices oscillate predictably between support and resistance, Bollinger Bands capture the boundaries accurately, and the strategy generates consistent returns with low drawdown. Momentum struggles in these conditions because there is no sustained trend to follow. MACD crossovers fire frequently but lead to whipsaws rather than profitable trends.
During strong directional moves (parts of 2021-2022 and 2023-2024), the dynamic flips. Momentum captures the trend and rides it. Mean reversion fights the trend, fading rallies that keep going and buying dips that keep dipping. Mean reversion still ends these periods positive on altcoins because the oscillations within the trend provide enough tradeable bounces, but its drawdowns are deeper and its equity curve rougher.
Asynchronous Drawdowns
The most valuable property of running both strategies simultaneously is that their drawdowns rarely overlap. When mean reversion is struggling because the market is trending hard, momentum is capturing that same trend. When momentum is getting whipsawed in a range, mean reversion is collecting steady profits from the oscillations.
We measured the drawdown correlation between the two strategies across the full five-year validation period. The correlation coefficient was 0.18, which is remarkably low for two strategies trading the same asset universe. This means that portfolio-level drawdown when running both is significantly lower than either strategy in isolation. The maximum portfolio drawdown was 11 percent, compared to 18 percent for mean reversion alone and 22 percent for momentum alone.
The regime_adaptive Meta-Strategy
We built a regime_adaptive strategy that attempts to delegate to the best strategy per regime automatically. It uses our quantitative regime detector to classify the current market state, then routes signals to mean reversion during ranging conditions and momentum during trending conditions.
The results were mixed. In backtests where regime classification was accurate, regime_adaptive outperformed both individual strategies. But regime detection has inherent lag, and misclassification during transitions caused worse performance than simply running both strategies in parallel. The transition periods between regimes are exactly when strategy selection matters most, and they are also when regime detection is least reliable.
Our current deployment takes the simpler approach: run both strategies on overlapping symbol baskets and let the portfolio effect handle regime rotation naturally. The 13 mean reversion bots and 5 momentum bots (15m) plus 6 momentum bots (4h) trade independently. The portfolio-level risk manager caps total exposure at 50 percent and monitors correlation between positions.
The Case for Multi-Strategy Portfolios
The data is clear: neither strategy wins in all conditions, and attempting to switch between them introduces timing risk that offsets the potential benefit. The optimal approach is diversification across strategy types, letting each strategy play its natural strengths while the other provides ballast during its weak periods.
This extends beyond just mean reversion and momentum. Our full deployment includes derivatives strategies (leverage_composite), macro strategies (correlation_regime, nupl_cycle_filter, stablecoin_supply_momentum), and multiple timeframe variants. Each strategy has a different regime sensitivity profile, and the portfolio benefits from every uncorrelated return stream we can add.
The practical takeaway: do not choose between mean reversion and momentum. Run both with proper position sizing and portfolio risk management. The real edge is not in any single strategy. It is in the portfolio construction that combines multiple imperfectly correlated strategies into a whole that is smoother and more resilient than any individual component.