BacktestingWalk-Forward Optimization (WFO) rolling
Re-optimize parameters on rolling window then test on next out-of-sample window
Source: blog.quantinsti.com
BacktestingAnchored walk-forward
Anchored start expanding window; test next out-of-sample period
Source: blog.quantinsti.com
BacktestingCombinatorial Purged CV (CPCV)
Cross-validation that purges data leakage near train/test boundary; Lopez de Prado
Source: arxiv.org
Backtestingk-fold time-series CV
Block-based CV preserving temporal order
Source: blog.quantinsti.com
BacktestingBootstrap aggregating (bagging) returns
Bootstrap return series for confidence intervals on metrics
Source: arxiv.org
BacktestingBlock bootstrap autocorrelation aware
Use block bootstrap to preserve time-series autocorrelation
Source: arxiv.org
BacktestingMonte Carlo trade order shuffling
Shuffle trade order to estimate path-dependent risk
Source: quantt.co.uk
BacktestingMonte Carlo permutation test
Permute strategy signals to compute p-values
Source: arxiv.org
BacktestingMonte Carlo path simulation
Simulate alternative price paths via GBM/bootstrap to stress strategy
Source: quantt.co.uk
BacktestingVectorized backtester
Use pandas/numpy vectorization for fast strategy evaluation
Source: blog.quantinsti.com
BacktestingEvent-driven backtester
Tick-by-tick simulation respecting fill prob and partial fills
Source: blog.quantinsti.com
BacktestingRealistic execution model
Include slippage commission spread crossing borrow fees in tests
Source: blog.quantinsti.com
RobustnessOut-of-sample testing
Reserve at least 30% of data as untouched OOS
Source: strategyquant.com
RobustnessWalk-forward efficiency ratio (WFE)
Ratio of OOS performance to in-sample; >0.5 is acceptable
Source: strategyquant.com
RobustnessParameter sensitivity heatmap
Plot Sharpe across parameter grid; require flat plateau not single peak
Source: blog.quantinsti.com
RobustnessMulti-asset robustness check
Ensure strategy works on similar correlated assets not just one
Source: quantt.co.uk
RobustnessMulti-period robustness check
Strategy must work in distinct market regimes (bull/bear/sideways)
Source: arxiv.org
RobustnessDeflated Sharpe Ratio
Adjust Sharpe for multiple-testing bias; Bailey & Lopez de Prado
Source: arxiv.org
RobustnessProbability of Backtest Overfitting (PBO)
Quantify likelihood that backtest selected via overfitting; CSCV approach
Source: arxiv.org
RobustnessGT-Score anti-overfit objective
Embed anti-overfit constraints in objective during optimization
Source: arxiv.org/2602.00080
RobustnessReality Check (White)
Test if best strategy is statistically better than null after multiple testing
Source: arxiv.org
RobustnessStationary bootstrap
Politis-Romano bootstrap for time-series resampling
Source: arxiv.org
Regime DetectionHidden Markov Model 2-state
Fit HMM to returns to detect low-vol vs high-vol regimes; trade differently
Source: blog.quantinsti.com
Regime DetectionHMM 3-state (bull/bear/range)
3-state HMM separates trend up trend down and range
Source: blog.quantinsti.com
Regime DetectionMarkov Switching VAR
Multi-asset switching VAR for regime-aware factor allocation
Source: blog.quantinsti.com
Regime DetectionGaussian Mixture Model regimes
GMM clustering on volatility/return features for regimes
Source: blog.quantinsti.com
Regime DetectionVolatility regime indicator (VIX threshold)
Trade differently above/below VIX 20 thresholds
Source: quantpedia.com
Regime DetectionTrend strength regime (ADX)
ADX>25 = trending regime; <20 = range; switch strategies accordingly
Source: blog.quantinsti.com
Regime DetectionYield curve regime detection
Inverted curve = recession risk regime; defensive allocation
Source: quantpedia.com
Regime DetectionMacro factor PCA regime
Decompose macro factors via PCA; regime by component sign
Source: arxiv.org
Regime DetectionChange point detection (CUSUM)
Online CUSUM detects structural breaks for parameter resets
Source: arxiv.org
Regime DetectionBayesian online change point
Adams-MacKay Bayesian change point for regime shifts
Source: arxiv.org
Regime-AdaptiveRegime-specific Random Forest models
Train separate ML models per regime; use HMM regime label to select model
Source: blog.quantinsti.com
Regime-AdaptiveRegime-conditional position sizing
Scale position by regime: full size in trend; half in range; flat in chaos
Source: blog.quantinsti.com
Regime-AdaptiveRegime-conditional asset allocation
Different allocation per regime: equity-heavy in growth; bond-heavy in recession
Source: quantpedia.com
Feature EngineeringLagged returns features
1/5/10/20-day lagged returns as ML features
Source: arxiv.org
Feature EngineeringRealized volatility features
Multi-window realized vol features (5/20/60-day)
Source: arxiv.org
Feature EngineeringRange/spread microstructure
Daily range high-low close-open as microstructure features
Source: arxiv.org
Feature EngineeringVolume features
Volume z-score volume ratio dollar-volume features
Source: arxiv.org
Feature EngineeringCross-sectional rank features
Universe-wide percentile rank features for relative-value models
Source: arxiv.org
Feature EngineeringTechnical indicator features
RSI MACD BBands as ML features (be careful of multicollinearity)
Source: arxiv.org
Feature EngineeringTerm structure of vol features
VIX-VVIX spread VIX/VXV ratio as risk-on/off signal
Source: quantpedia.com
Feature EngineeringCross-asset features
Bond-equity correlation gold-USD beta as macro features
Source: quantpedia.com
Feature EngineeringFractional differentiation
Lopez de Prado fractional diff to remove memory while preserving statistical properties
Source: arxiv.org
Feature EngineeringTriple-barrier labeling
Lopez de Prado labeling with vertical/horizontal barriers
Source: arxiv.org
Feature EngineeringMeta-labeling
Train secondary model to filter primary signal; reduces false positives
Source: arxiv.org
Alternative DataSatellite imagery oil storage
Track oil tank fill levels via satellite for inventory surprises
Source: quantt.co.uk
Alternative DataSatellite parking lot count retail
Count cars at retail parking lots to predict same-store sales
Source: quantt.co.uk
Alternative DataSatellite shipping/port congestion
Track container ship congestion for trade flow surprises
Source: quantt.co.uk
Alternative DataCredit card transaction data
Aggregate card data predicts company revenues days ahead
Source: quantt.co.uk
Alternative DataApp download/MAU data
App store rankings predict user engagement metrics
Source: quantt.co.uk
Alternative DataWeb scraping product prices
Scrape competitor pricing changes for margin forecasts
Source: quantt.co.uk
Alternative DataWeb traffic SimilarWeb
Web traffic data correlates with online ad revenue
Source: quantt.co.uk
Alternative DataGlassdoor employee reviews
Decline in employee sentiment predicts negative earnings
Source: quantt.co.uk
Alternative DataJob postings (LinkUp/Burning Glass)
Spike in job postings indicates expansion intent
Source: quantt.co.uk
Alternative DataSEC EDGAR text mining
NLP on 10-K/10-Q risk language predicts forward returns
Source: arxiv.org
Alternative DataEarnings call transcript NLP
Sentiment in call Q&A vs prepared remarks predicts post-earnings drift
Source: arxiv.org
Alternative DataTwitter/X StockTwits sentiment
Real-time sentiment scoring drives short-horizon predictions
Source: arxiv.org
Alternative DataReddit WSB mention spike
Spike in retail mentions often precedes mean-reversion (post-2021 GME effect fades)
Source: arxiv.org
Alternative DataGoogle Trends search volume
Search interest correlates with retail demand
Source: arxiv.org
Alternative DataNews article sentiment
RavenPack/Bloomberg sentiment scores on news flow
Source: arxiv.org
Alternative DataWeather data energy demand
HDD/CDD forecasts drive nat gas and power futures
Source: quantpedia.com
Alternative DataUSDA WASDE leak monitoring
Detect anomalous trading patterns hours before USDA reports
Source: quantpedia.com
Alternative DataEsoteric satellite (mall traffic)
Track mall foot traffic via satellite for REIT revenue forecasts
Source: quantt.co.uk
Reinforcement LearningQ-learning trading agent
Tabular Q-learning over discrete state-action grid
Source: quantt.co.uk
Reinforcement LearningDeep Q Network (DQN)
Neural-network Q-function for continuous state space
Source: quantt.co.uk
Reinforcement LearningDouble DQN
Reduce overestimation bias in Q-learning with double network
Source: quantt.co.uk
Reinforcement LearningPolicy Gradient (REINFORCE)
Direct policy optimization via gradient ascent on expected reward
Source: quantt.co.uk
Reinforcement LearningActor-Critic (A2C/A3C)
Combine value and policy networks for stability
Source: quantt.co.uk
Reinforcement LearningPPO trading policy
Proximal Policy Optimization with clipped surrogate objective
Source: quantt.co.uk
Reinforcement LearningDDPG continuous action
Deterministic policy gradient for continuous position sizing
Source: quantt.co.uk
Reinforcement LearningSAC entropy-regularized
Soft Actor Critic for exploration-exploitation balance
Source: quantt.co.uk
Reinforcement LearningFinRL ensemble agent
Open-source ensemble across PPO/A2C/SAC for trading
Source: arxiv.org
Reinforcement LearningMulti-agent RL
Multiple agents trade different markets with shared learning
Source: arxiv.org
OptimizationGenetic algorithm parameter search
Evolutionary search over strategy parameters
Source: strategyquant.com
OptimizationParticle swarm optimization
Swarm-based search for high-dim parameter spaces
Source: strategyquant.com
OptimizationBayesian optimization
Gaussian-process-based hyperparameter tuning
Source: arxiv.org
OptimizationDifferential evolution
Population-based optimization for noisy objectives
Source: arxiv.org
OptimizationSimulated annealing
Probabilistic acceptance of worse solutions to escape local optima
Source: arxiv.org
OptimizationMean-Variance optimization (Markowitz)
Optimize portfolio weights to minimize variance for target return
Source: quantt.co.uk
OptimizationHierarchical Risk Parity (HRP)
Lopez de Prado hierarchical clustering for stable allocation
Source: arxiv.org
OptimizationBlack-Litterman views integration
Combine market-implied returns with views for stable optimization
Source: quantt.co.uk
OptimizationRisk parity equal risk contribution
Each asset contributes equal risk; popular in CTAs
Source: quantt.co.uk
ExecutionAlmgren-Chriss optimal execution
Optimal execution trading off market impact vs timing risk
Source: quantt.co.uk
ExecutionImplementation Shortfall
Minimize total cost vs decision price
Source: quantt.co.uk
ExecutionVWAP/TWAP smart slicing
Algorithmic time/volume slicing to track benchmark
Source: quantt.co.uk
ExecutionAdaptive arrival price
Adapt aggressiveness to liquidity and price relative to arrival
Source: quantt.co.uk
Risk ModelsMulti-factor risk model (Barra-style)
Decompose risk into style+industry+country factors
Source: quantt.co.uk
Risk ModelsEWMA volatility forecasting
Exponentially weighted moving average for vol updates
Source: quantt.co.uk
Risk ModelsGARCH(1 1) volatility
Standard volatility persistence model for risk forecasting
Source: arxiv.org
Risk ModelsDCC-GARCH dynamic correlation
Dynamic conditional correlation between assets
Source: arxiv.org
Risk ModelsCopula-based dependence
Gaussian/t/Clayton copulas for tail-dependence modeling
Source: arxiv.org
Risk ModelsExtreme Value Theory tail risk
EVT/GPD modeling of tail of return distribution
Source: arxiv.org
Anomaly DetectionIsolation Forest market anomaly
Detect price/volume anomalies for risk-off triggers
Source: arxiv.org
Anomaly DetectionAutoencoder reconstruction error
Train autoencoder; flag periods with high reconstruction error
Source: arxiv.org
Network ModelsLead-lag information flow
Granger causality / transfer entropy to find lead assets
Source: arxiv.org
Network ModelsCorrelation network clustering
Build correlation graph; cluster stocks by community structure
Source: arxiv.org
Network ModelsMinimum spanning tree of stocks
MST highlights central stocks driving market dynamics
Source: arxiv.org
Causal InferenceSynthetic control causal effect
Causal effect of policy/event by constructing synthetic counterfactual
Source: arxiv.org
Causal InferenceDouble ML for treatment effect
Chernozhukov double ML to control confounders in alpha discovery
Source: arxiv.org
An encyclopedia of publicly documented strategies for education. YCAI does not recommend, endorse, or trade any of these. Not investment advice.