Skip to content

Trading

Subtitle: Quantitative Strategy Systems, Risk Management Frameworks, and Survival Under Alpha Decay

The Essence

Trading is fundamentally a decision optimization problem: allocating optimal buy/sell timing, direction, and position sizes under uncertainty with incomplete information.

Mathematically, a trading strategy is an abstract mapping function:

\[f: \mathcal{S}_t \rightarrow \mathcal{A}_t\]

Where \(\mathcal{S}_t\) is the state space at time \(t\) (prices, volumes, order book snapshots, on-chain data, macro indicators, etc.), and \(\mathcal{A}_t\) is the action space (buy/sell/hold × position size × asset selection). Strategy development is the process of finding a state-to-action mapping that maximizes long-term risk-adjusted returns.

What makes crypto trading distinctive:

  • No price limits: 50%+ single-day drawdowns are not uncommon (LUNA collapse). Tail risk management priority far exceeds traditional equity markets.
  • Hyper-developed leverage ecosystem: 125x maximum leverage (Binance perpetuals) means tiny price fluctuations trigger massive cascading liquidations.
  • Unique information propagation paths: Information flows through Twitter/X, Telegram groups, and on-chain transactions — not through Bloomberg Terminals or analyst reports. Retail and institutions access information nearly simultaneously (but processing capability differs vastly).

Core Mechanics

1. Strategy Taxonomy

By signal source and holding period, crypto trading strategies fall into four tiers:

High-Frequency Strategies (HFT, <1 min holding):

  • Market Making: Continuously quoting bid-ask prices, earning the spread. Core challenge is adverse selection risk — when your limit order fills, it's likely an informed trader eating your quote.
  • Latency Arbitrage: Cross-exchange spread convergence. In crypto, this involves API latency, WebSocket push speed, and even physical distance to exchange matching engines.

Medium-Frequency Strategies (1 hour - several days holding):

  • Statistical Arbitrage: Spread mean reversion. Classic methods: Engle-Granger cointegration test, Kalman Filter dynamic hedge ratio. In crypto, BTC-ETH spread and intra-sector spreads (e.g., SOL-AVAX) are common targets.
  • Funding Rate Arbitrage: When funding rates deviate from equilibrium, hold spot long + perpetual short (or inverse), locking in funding rate yield. Risks to monitor: funding rate reversal speed, spot borrowing costs, exchange counterparty risk.

Medium-Low Frequency Strategies (several days - weeks holding):

  • Momentum Factor Strategies: Crypto momentum effects are significantly stronger than in traditional markets. Academic research shows BTC 1-week and 4-week momentum factor t-statistics above 3.0. But momentum factor drawdowns during market reversals are extreme (Momentum Crash).
  • On-Chain Factor Strategies: Factorizing MVRV, SOPR, Exchange Flow, and other on-chain indicators into multi-factor scoring models.

Low-Frequency / Macro Strategies (weeks - months holding):

  • Trend Following: Capturing major trends via moving average crossovers, Donchian Channel breakouts, and other classical methods. In BTC's long cycles (~4-year halving cycle), trend following strategies have historically achieved Sharpe Ratios of 1.0+.
  • Macro Rotation: Rotating between BTC / ETH / altcoins / stablecoins based on macro factors including stablecoin liquidity, US Treasury rates, and DXY.

2. Risk Management Framework

"A trader who doesn't manage risk isn't a trader — they're a gambler."

  • Position Sizing: Kelly Criterion provides the theoretically optimal bet fraction \(f^* = \frac{p \cdot b - q}{b}\) (\(p\) = win rate, \(b\) = odds, \(q = 1-p\)). Full Kelly is too volatile in practice; Half-Kelly or more conservative fractions are standard.
  • Stop-Loss Discipline: Fixed percentage stop, ATR-based stop (Average True Range), volatility-adjusted stop. Key principle: stop levels must be determined before entry — never move stops in the adverse direction during a trade.
  • Portfolio-Level Risk: Correlation matrix monitoring. In extreme crypto drawdowns, all asset correlations approach 1 (Risk-On / Risk-Off effect) — diversification fails precisely when you need it most. Countermeasures: tail hedging, dynamic deleveraging.
  • Maximum Drawdown Constraint: Set a portfolio-level max drawdown threshold (e.g., -20%), triggering automatic position reduction to minimum upon breach.

3. Backtesting Methodology

As Marcos López de Prado argues in Advances in Financial Machine Learning, the majority of published backtest results are artifacts of overfitting. This book strictly follows:

  • CPCV (Combinatorial Purged Cross-Validation): A backtest splitting method that prevents information leakage.
  • Multiple Testing Correction: If you tested 100 strategy parameters, the best one is probably fitting noise. Deflated Sharpe Ratio corrects for multiple comparison bias.
  • Walk-Forward Analysis: Rolling-window backtesting that simulates real "unknown future" conditions.
  • Execution Simulation: Incorporating real order book depth, slippage models, and fee structures into backtests. A backtest without execution costs has zero reference value.

4. Execution Optimization

Alpha leakage at the execution layer is chronically underestimated:

  • TWAP/VWAP Algorithms: Splitting large orders over time to minimize market impact.
  • Smart Order Routing (SOR): In fragmented markets, distributing orders across venues with optimal liquidity. CCXT provides a unified cross-exchange API interface.
  • Slippage Prediction Models: Based on historical order book data, predicting expected slippage for different position sizes. On low-liquidity altcoins, a $1M market order can cause 2-5% price impact.

The Alpha Connection

  • Factor crowding monitoring: When a factor (e.g., momentum) is chased by too many strategies, factor returns decay or even reverse. Monitoring factor crowding is critical for protecting Alpha.
  • Nonlinear factor interactions: Cross-signals between on-chain factors and derivatives factors are often more effective than single factors. Example: Low MVRV + extremely negative funding rate = oversold signal reliability dramatically increases.
  • Execution Alpha: Under identical strategy signals, execution quality differences can account for 5-15% annualized return differential. Optimizing execution paths is itself a form of Alpha.
  • Strategy orthogonalization: Constructing portfolios of low-correlation strategies, using diversification to reduce portfolio volatility. Grinold's Fundamental Law tells us that increasing \(\sqrt{BR}\) (number of independent bets) linearly improves IR.
  • Volatility regime switching: Crypto markets shift from low-volatility to high-volatility regimes with extreme abruptness (regime change). Using Hidden Markov Models (HMM) or volatility clustering detection for adaptive strategy parameter adjustment is a higher-order Alpha source.

Chapter Roadmap

After completing this chapter, you will command: a complete taxonomy of major crypto strategy types with their return/risk profiles; Kelly Criterion and portfolio-level risk management frameworks; López de Prado-grade backtesting discipline capable of identifying and avoiding overfitting traps; core execution optimization methodology. You don't need to find a perfect strategy — you need a complete closed-loop system for strategy development, testing, deployment, monitoring, and iteration. Strategies decay. Systems don't.