In briefConditional evidence weights for gradient-boosted credit models: estimation, additivity, and adverse action reasons
SummaryThe paper defines conditional weight of evidence, which measures how observing a credit characteristic changes default log odds once the rest of an applicant's credit record is known. Its sign shows whether the information is adverse or favorable for that applicant, and its size shows its strength. The authors estimate it with gradient-boosted trees and test accuracy in a bureau-style simulation. The measure is not an adverse action reason.
In briefOptimal trading filters and how much predictive signals are worth in trading
SummaryThe paper solves a quadratic trading problem with transient price impact using Wiener-Hopf factorization. The optimal policy is expressed as a conditional forecast composed with an explicit causal linear filter. Results are given for exponential resilience and power-law impact decay, and the net value under trading costs is characterized.
WhyApplies Wiener-Hopf factorization to derive optimal trading with transient impact and fractional filters, a clear novelty; results are analytic and mainly impact quantitative finance theory.
In briefRachev-Ratio Portfolio Allocation Using TimeGAN-Generated Scenarios: Replacement Versus Supplementation
SummaryThis study tests how well scenarios from a TimeGAN model work in portfolio allocation using 12 exchange-traded funds. It compares historical, replacement, and hybrid policies across five training seeds, with delayed execution and transaction costs included. The generated scenarios understate volatility and tail losses and distort dependence between assets, and longer training improves reconstruction loss without improving correlation accuracy. Some hybrid configurations earn higher average net returns, but the confidence intervals for generated-versus-historical differences include zero at the reference block length.
In briefDynamic portfolio selection using range VaR and log returns
SummaryThis paper studies continuous-time portfolio selection in which risk is measured by Range Value-at-Risk (RVaR) and performance is evaluated using logarithmic returns. The dynamic problem is reduced to a static quantile problem, and the optimal terminal wealth is characterized through a convex envelope. The RVaR solutions reduce to their VaR and Expected Shortfall counterparts in limiting parameter cases.
In briefDynamic Programming for Value-at-Risk Decision Processes in Continuous Spaces
SummaryThis paper studies Markov decision processes whose objective is Value-at-Risk (VaR), a quantile-based risk measure. Earlier work gave dynamic programming equations for finite state and action spaces, but the general case was unresolved. The authors show these equations are valid in continuous spaces, for both finite and infinite horizons. They apply the framework to an optimal execution problem with a VaR objective and characterize the optimal dynamic strategies.
In briefChoosing Trade Duration and Execution Schedule Under Concave Impact and Hurst Scaling
SummaryThis paper studies order execution under concave power-law market impact and Hurst-scaled price fluctuations. It selects the execution horizon by maximizing the Sharpe ratio of the trade's realized P&L, then derives the execution schedule shape via convex programming. The authors report that an interior optimal horizon exists only under certain sign conditions, and the objective is nearly flat over a wide window. Under Hurst-scaled cost, the schedule is front-loaded and tapers smoothly to zero, saving 10-17% of envelope cost versus a constant-rate (TWAP) schedule.
In briefModelling Cross-Product Liquidity in Intraday Electricity Markets with Hawkes Processes
SummaryThis paper introduces a multivariate State-Dependent Hawkes Process to model how Quarter-Hourly and Hourly intraday continuous electricity contracts influence each other's liquidity. Trading sessions are split into 'Normal' and 'Stressed' regimes using bid-ask spread, mid-price volatility and queue imbalance. The model is calibrated on level 3 limit order book data from the Dutch EPEX SPOT market for the first month of each quarter of 2024. Results show roughly 95% of orders are reactions to earlier orders, and under stress the leadership of the Quarter-Hourly contract weakens while the cross-contract lag shrinks sharply.
Ayoub Jhabli, Tarek Alskaif, Kwabena E. Bennin et al.
