arXiv subjects

Steven E. Pav

Publications and source records attributed to Steven E. Pav.

13 recordsLinked to original sources

arXiv2026

Post Selection Estimation of Sharpe Ratios

We consider the problem of estimating the true Sharpe ratio of an asset selected for having the highest observed in-sample Sharpe ratio among many assets. We discuss estimators based on the polyhedral lemma, James Stein shrinkage, debiasing the expected maximum Sharpe ratio, thresholding and empirical Bayes. We test these estimators in simulations, computing bias and root mean square error across different values of sample size, number of assets, and spread and shape of population Sharpe ratios. We also compute rank correlation of the estimators against the underlying quantity, simulating how these estimators might be used to compare or rank the output of different teams which perform this selection process. We find that the James Stein estimator provides the best performance across many different realistic values of the relevant parameters, followed by the GMLEB estimator of Jiang and Zhang. These results are fairly robust to correlation of asset returns, with some caveats.

q-fin.PM↗

arXiv2026

Conditional inference on the asset with maximum Sharpe ratio

We apply the procedure of Lee et al. to the problem of performing inference on the signal-noise ratio of the asset which displays maximum sample Sharpe ratio over a set of possibly correlated assets. We find a multivariate analogue of the commonly used approximate standard error of the Sharpe ratio to use in this conditional estimation procedure. We also consider several alternative procedures, including the simple Bonferroni correction for multiple hypothesis testing, which we fix for the case of positive common correlation among assets, the chi-bar square test against one-sided alternatives, Follman's test, and Hansen's asymptotic adjustments. Testing indicates the conditional inference procedure achieves nominal type I rate, and does not appear to suffer from non-normality of returns. The conditional estimation test has low power under the alternative where there is little spread in the signal-noise ratios of the assets, and high power under the alternative where a single asset has high signal-noise ratio. Unlike the alternative procedures, it appears to enjoy rejection probabilities monotonic in the signal-noise ratio of the selected asset, and actually maintains near-nominal rejection rates under the conditional null.

q-fin.ST↗

arXiv2026

The Sherman-Morrison-Markowitz Portfolio

We show that the Markowitz portfolio is a scalar multiple of another portfolio which replaces the covariance with the second moment matrix, via simple application of the Sherman-Morrison identity. Moreover it is shown that when using conditional estimates of the first two moments, this "Sherman-Morrison-Markowitz" portfolio solves the standard unconditional portfolio optimization problems. We argue that in multi-period portfolio optimization problems it is more natural to replace variance and covariance with their uncentered counterparts. We extend the theory to deal with constraints in expectation, where we find a decomposition of squared effects into spanned and orthogonal components. Compared to the Markowitz portfolio, the Sherman-Morrison-Markowitz portfolio downlevers by a small amount that depends on the conditional squared maximal Sharpe ratio; the practical impact will be fairly small, however. We present some example use cases for the theory.

q-fin.PM↗

arXiv2025

Inferring Piece Value in Chess and Chess Variants

We use logistic regression to estimate the value of the pieces in standard chess and several chess variants, namely Chess 960, Atomic chess, Antichess, and Horde chess. We perform our regressions on several years of data from Lichess, the free and open-source internet chess server. We use the published player ratings to control for the confounding effect of differential player skill. We adjust for the attenuation bias in regressions due to the noise in observed ratings. We find that major piece values, relative to the value of a pawn, are fairly consistent with historical valuation systems. However we find slightly higher value to bishops than knights. We find that piece values are smaller, in absolute value, in Atomic and Antichess than standard chess. We also present approximate values of the pieces to equalize odds when players of varying skill face off. We briefly consider self-play experiments using the Stockfish engine, which give a contrasting view of piece value.

stat.AP↗

arXiv2024

An Iterative Algorithm for Regularized Non-negative Matrix Factorizations

We generalize the non-negative matrix factorization algorithm of Lee and Seung to accept a weighted norm, and to support ridge and Lasso regularization. We recast the Lee and Seung multiplicative update as an additive update which does not get stuck on zero values. We apply the companion R package rnnmf to the problem of finding a reduced rank representation of a database of cocktails.

cs.LG↗

arXiv2021

Inference on the Sharpe ratio via the upsilon distribution

The upsilon distribution, the sum of independent chi random variates and a normal, is introduced. As a special case, the upsilon distribution includes Lecoutre's lambda-prime distribution. The upsilon distribution finds application in Frequentist inference on the Sharpe ratio, including hypothesis tests on independent samples, confidence intervals, and prediction intervals, as well as their Bayesian counterparts. These tests are extended to the case of factor models of returns.

q-fin.ST↗

arXiv2020

Inference on Achieved Signal Noise Ratio

We describe a procedure to perform approximate inference on the achieved signal-noise ratio of the Markowitz Portfolio under Gaussian i.i.d. returns. The procedure relies on a statistic similar to the Sharpe Ratio Information Criterion. Testing indicates the procedure is somewhat conservative, but otherwise works well for reasonable values of sample and asset universe sizes. We adapt the procedure to deal with generalizations of the portfolio optimization problem.

stat.ME↗

arXiv2020

Asymptotic distribution of the Markowitz portfolio

The asymptotic distribution of the Markowitz portfolio is derived, for the general case (assuming fourth moments of returns exist), and for the case of multivariate normal returns. The derivation allows for inference which is robust to heteroskedasticity and autocorrelation of moments up to order four. As a side effect, one can estimate the proportion of error in the Markowitz portfolio due to mis-estimation of the covariance matrix. A likelihood ratio test is given which generalizes Dempster's Covariance Selection test to allow inference on linear combinations of the precision matrix and the Markowitz portfolio. Extensions of the main method to deal with hedged portfolios, conditional heteroskedasticity, conditional expectation, and constrained estimation are given. It is shown that the Hotelling-Lawley statistic generalizes the (squared) Sharpe ratio under the conditional expectation model. Asymptotic distributions of all four of the common `MGLH' statistics are found, assuming random covariates. Examples are given demonstrating the possible uses of these results.

q-fin.PM↗

arXiv2019

A post hoc test on the Sharpe ratio

We describe a post hoc test for the Sharpe ratio, analogous to Tukey's test for pairwise equality of means. The test can be applied after rejection of the hypothesis that all population Signal-Noise ratios are equal. The test is applicable under a simple correlation structure among asset returns. Simulations indicate the test maintains nominal type I rate under a wide range of conditions and is moderately powerful under reasonable alternatives.

stat.ME↗

arXiv2015

Safety Third: Roy's Criterion and Higher Order Moments

Roy's `Safety First' criterion for selecting one risky asset from many is adapted to the case of non-normal returns, via Cornish Fisher expansion. The resulting investment objective is consistent with first order stochastic dominance, and is equal to the Sharpe ratio for the case of normal returns. An investor selecting assets via this objective is not universally attracted to positive skew, rather the preference for skew depends on term, the expected return and the disastrous rate of return.

q-fin.ST↗

arXiv2015

Moments of the log non-central chi-square distribution

The cumulants and moments of the log of the non-central chi-square distribution are derived. For example, the expected log of a chi-square random variable with v degrees of freedom is log(2) + psi(v/2). Applications to modeling probability distributions are discussed.

stat.AP↗

arXiv2014

Bounds on Portfolio Quality

The signal-noise ratio of a portfolio of p assets, its expected return divided by its risk, is couched as an estimation problem on the sphere. When the portfolio is built using noisy data, the expected value of the signal-noise ratio is bounded from above via a Cramer-Rao bound, for the case of Gaussian returns. The bound holds for `biased' estimators, thus there appears to be no bias-variance tradeoff for the problem of maximizing the signal-noise ratio. An approximate distribution of the signal-noise ratio for the Markowitz portfolio is given, and shown to be fairly accurate via Monte Carlo simulations, for Gaussian returns as well as more exotic returns distributions. These findings imply that if the maximal population signal-noise ratio grows slower than the universe size to the 1/4 power, there may be no diversification benefit, rather expected signal-noise ratio can decrease with additional assets. As a practical matter, this may explain why the Markowitz portfolio is typically applied to small asset universes. Finally, the theorem is expanded to cover more general models of returns and trading schemes, including the conditional expectation case where mean returns are linear in some observable features, subspace constraints (i.e., dimensionality reduction), and hedging constraints.

q-fin.PM↗

arXiv2013

SRCEK: A Continuous Embedding of the Channel Selection Problem for weighted PLS Modeling

SRCEK, is a technique for selecting useful channels for affine modeling of a response by PLS. The technique embeds the discrete channel selection problem into the continuous space of predictor preweighting, then employs a Quasi-Newton (or other) optimization algorithm to optimize the preweighting vector. Once the weighting vector has been optimized, the magnitudes of the weights indicate the relative importance of each channel. The relative importances are used to construct n different models, the kth consisting of the k most important channels. The different models are then compared by means of cross validation or an information criterion (e.g. BIC), allowing automatic selection of a `good' subset of the channels. The analytical Jacobian of the PLS regression vector with respect to the predictor weighting is derived to facilitate optimization of the latter. This formulation exploits the reduced rank of the predictor matrix to gain some speedup when the number of observations is fewer than the number of predictors (the usual case for e.g. IR spectroscopy). The method compares favourably with predictor selection techniques surveyed by Forina et. al.

stat.AP↗