DocServer

Pseudo mathematics and financial charlatanism: Backtest overfitting and out-of-sample over performance

Bailey, David H. and Borwein, Jonathan M. and de Pardo, Lopez M. and Zhu, Qiji J. (2014) Pseudo mathematics and financial charlatanism: Backtest overfitting and out-of-sample over performance. Notices of the AMS (61). pp. 458-471.

[img] PDF
Download (0b)

    Abstract

    Recent computational advances allow investment managers to me-thodically search through thousands or even millions of potential op-tions for a pro table investment strategy. In many instances, the re-sulting strategy involves a pseudo-mathematical argument, which isspuriously validated through a simulation of its historical performance(also called backtest).We prove that high performance is easily achievable after backtest-ing a relatively small number of alternative strategy con gurations, apractice we denote \backtest over tting." The higher the number ofcon gurations tried, the greater is the probability that the backtest isover t. Because nancial analysts rarely report the number of con gu-rations tried for a given backtest, investors cannot evaluate the degreeof over tting in most investment claims and analysis.The implication is that investors can be easily misled into allo-cating capital to strategies that appear to be mathematically soundand empirically supported by a backtest. This practice is particu-larly pernicious, because due to the nature of nancial time series,backtest over tting has a detrimental e ect on the strategy's futureperformance.

    Item Type: Article
    Subjects: UNSPECIFIED
    Faculty: UNSPECIFIED
    Depositing User: Mrs Naghmana Tehseen
    Date Deposited: 21 Sep 2016 12:01
    Last Modified: 21 Sep 2016 12:01
    URI: https://docserver.carma.newcastle.edu.au/id/eprint/1751

    Actions (login required)

    View Item