Strategy classification
Quantitative strategies can be classified by forecast target, instrument, horizon, and implementation. The same broad factor can be traded in different ways. Momentum can be cross-sectional, time-series, intraday, or multi-month; each has different data and cost requirements.
Trend following
Trend strategies take positions in the direction of recent price movement, often across futures, currencies, rates, commodities, or equities. They can benefit from persistent macro moves and struggle in choppy reversals. Position sizing and volatility targeting are central.
Mean reversion and statistical arbitrage
These strategies expect short-term deviations among related instruments or from estimated fair relationships to reverse. Research can use pairs, residuals, baskets, order flow, or cross-sectional models. Failure occurs when the relationship changes, the position is crowded, or financing and liquidity disappear.
Value, carry, quality, and defensive signals
Value relates price to fundamentals or other anchors. Carry reflects the return earned if conditions remain unchanged, such as yield, roll, or forward premium. Quality and defensive factors use profitability, stability, leverage, or risk characteristics. Definitions differ and can overlap; a factor name isn't a full specification.
Market making
Market makers quote prices, manage inventory, hedge, and earn spread while bearing adverse-selection and market risk. Performance depends on microstructure, latency, queue position, fill probability, and the behavior of informed traders.
Volatility and options
Volatility strategies trade implied versus realized volatility, skew, term structure, dispersion, or event risk. They require option pricing, Greeks, hedging, transaction cost, and tail-risk analysis. A strategy that earns small premium regularly can hide rare large losses.
Event and text signals
Models can process earnings, news, filings, corporate actions, and other events. Natural-language methods still require clean timestamps, document versioning, and an understanding of whether the signal is available before the simulated trade.
Alternative data
Transaction, web, satellite, mobility, supply-chain, and other datasets can support research. Evaluate legal rights, privacy, representativeness, revision, survivorship, vendor stability, and whether the data remains exclusive enough to matter.
Machine learning
Machine-learning methods can model nonlinear relationships and high-dimensional data. They don't remove the need for economic reasoning, careful validation, and robustness. Complex models can overfit more subtly and be difficult to diagnose under regime change.
Failure-mode checklist
A complete strategy description includes leakage, overfitting, multiple testing, selection bias, data revisions, transaction costs, capacity, crowding, factor exposure, leverage, liquidity, and production risk. The strategy is the entire system, not the predictive formula.