Asymmetric z-thresholds
Entry RulesNeutral
Different entry levels for long vs short side based on historical asymmetry of spread distribution
Source: Quant Lit
Bollinger Band on spread entry
Entry RulesNeutral
Enter when spread crosses outer Bollinger Band; close when crosses MA
Source: Common
Cointegration re-test exit
Exit RulesNeutral
Re-test cointegration weekly; close pair if test fails (relationship broken)
Source: Best Practice
Conditional value-at-risk threshold
Entry RulesNeutral
Set entry levels to control CVaR of spread tail loss
Source: Risk Quant
Copula-based entry signal
Entry RulesNeutral
Use copula CDF to detect joint tail extremes; no normality assumption
Source: Liew Wu 2013
Correlation-then-cointegration filter
Pair SelectionNeutral
First filter pairs by Pearson rho >0.7 then test cointegration; reduces compute on huge universe
Source: QuantInsti
DBSCAN clustering on returns
Pair SelectionNeutral
Density-based clustering on standardized returns finds candidate pair groups before cointegration testing
Source: ML Quant
Distance method (Gatev Goetzmann Rouwenhorst)
Pair SelectionNeutral
Sum of squared deviations of normalized prices over 12-mo formation period; trade top-20 pairs next 6 mo
Source: Gatev 2006
Engle-Granger 2-step pair test
Cointegration MethodsNeutral
OLS regress Y on X then ADF test residuals; if stationary p<0.05 pair is cointegrated; trade z-score deviations of residual
Source: Hudson Thames
Fundamental-similarity pair filter
Pair SelectionNeutral
Match stocks on size sector P/E leverage; trade only fundamentally similar pairs
Source: Value-Quant
Graph-theoretic minimum spanning tree pairs
Pair SelectionNeutral
MST on correlation distance reveals natural pair candidates as adjacent nodes
Source: Network Quant
Gregory-Hansen test with structural break
Cointegration MethodsNeutral
Cointegration test allowing one structural break in level/trend; needed for pairs that survive regime change
Source: Academic Lit
Half-life of mean reversion (OU calibration)
Cointegration MethodsNeutral
Calibrate Ornstein-Uhlenbeck speed lambda from spread; half-life = ln(2)/lambda; filter pairs HL 1-30 days
Source: ArXiv 2109
Hatemi-J 2-break cointegration
Cointegration MethodsNeutral
Extends GH to two structural breaks; for pairs spanning crisis periods
Source: Academic Lit
Hurst exponent mean reversion filter
Pair SelectionNeutral
Spread H<0.5 indicates mean-reverting regime; H near 0.5 random walk; H>0.5 trending
Source: Hurst Lit
Johansen test multi-asset cointegration
Cointegration MethodsNeutral
VECM-based test that finds multiple cointegration vectors; rank of Pi matrix tells how many independent stationary spreads exist
Source: Hudson Thames
Kalman filter dynamic beta
Cointegration MethodsNeutral
State-space model with hedge ratio as latent state; smooths through time without lookahead bias
Source: QuantInsti
Kapetanios-Snell-Shin (KSS) nonlinear test
Cointegration MethodsNeutral
Tests cointegration with nonlinear ESTAR adjustment; useful when reversion speed varies with spread magnitude
Source: ArXiv 2109
Mean reversion exit at z=0
Exit RulesNeutral
Close position when spread crosses zero; standard exit
Source: Common
Optimal threshold via sup-Wald (Bertram)
Entry RulesNeutral
Closed-form optimal trading bands for OU process maximizing return per unit time
Source: Bertram 2010
PCA-based cluster pair selection
Pair SelectionNeutral
Project returns to top-k PCs reduce noise then cluster; pairs from same cluster high cointegration prob
Source: Academic Lit
Pesaran ARDL bounds test
Cointegration MethodsNeutral
Tests cointegration with mixed I(0) and I(1) variables; useful for fundamentals-driven pairs
Source: Academic Lit
Phillips-Ouliaris cointegration test
Cointegration MethodsNeutral
Alternative residual-based test more robust to autocorrelation than ADF; use when EG residuals show serial correlation
Source: Academic Lit
Rolling OLS hedge ratio
Cointegration MethodsNeutral
Refit beta over rolling N-day window; simpler than Kalman but lookback window choice critical
Source: QuantConnect
SSD ranking with industry filter
Pair SelectionNeutral
Hybrid: distance method scoring within sector buckets; better economic rationale
Source: Academic Lit
Same-sector restriction
Pair SelectionNeutral
Limit pair search to same GICS sub-industry; reduces spurious cointegration false positives
Source: Academic Lit
Stochastic spread model entry
Entry RulesNeutral
Calibrate Vasicek/CIR to spread; trade when spread far from theoretical mean
Source: Academic Lit
Stop-loss at z=+/-3
Exit RulesNeutral
Cut loss if spread widens past 3 sigma; protects against broken cointegration
Source: Common
TLS (orthogonal regression) hedge ratio
Cointegration MethodsNeutral
Total least squares accounts for measurement error in both X and Y; better than OLS when both noisy
Source: Academic Lit
Time-stop exit at half-life multiple
Exit RulesNeutral
Force exit if no reversion within 2-3x half-life; avoids stuck positions
Source: Risk Mgmt
Trailing exit on z-score peak
Exit RulesNeutral
Lock profit by exiting when z retraces 50% from peak instead of waiting full reversion
Source: Discretion
VECM dynamic hedge ratio
Cointegration MethodsNeutral
Vector error correction model for time-varying hedge ratio; updates beta as relationship evolves
Source: Portfolio Optim Book
Variance ratio test on spread
Pair SelectionNeutral
Lo-MacKinlay VR<1 indicates mean reversion; complement to Hurst
Source: Academic Lit
Volatility-adjusted z-score
Entry RulesNeutral
Normalize z by realized vol of spread instead of unconditional std; faster regime adapt
Source: Quant Lit
Z-score entry +/-2 sigma
Entry RulesNeutral
Open spread when |z|>2; long-short on direction; classic Gatev style
Source: Gatev 2006
An encyclopedia of publicly documented strategies for education. YCAI does not recommend, endorse, or trade any of these. Not investment advice.