Mathematics > Optimization and Control
[Submitted on 20 Dec 2018 (v1), last revised 10 Jun 2021 (this version, v3)]
Title:Distributional Robust Kelly Gambling: Optimal Strategy under Uncertainty in the Long-Run
View PDFAbstract:In classic Kelly gambling, bets are chosen to maximize the expected log growth of wealth, under a known probability distribution. Breiman provides rigorous mathematical proofs that Kelly strategy maximizes the rate of asset growth (asymptotically maximal magnitude property), which is thought of as the principal justification for selecting expected logarithmic utility as the guide to portfolio selection. Despite very nice theoretical properties, the classic Kelly strategy is rarely used in practical portfolio allocation directly due to practically unavoidable uncertainty. In this paper we consider the distributional robust version of the Kelly gambling problem, in which the probability distribution is not known, but lies in a given set of possible distributions. The bet is chosen to maximize the worst-case (smallest) expected log growth among the distributions in the given set.
Computationally, this distributional robust Kelly gambling problem is convex, but in general need not be tractable. We show that it can be tractably solved in a number of useful cases when there is a finite number of outcomes with standard tools from disciplined convex programming.
Theoretically, in sequential decision making with varying distribution within a given uncertainty set, we prove that distributional robust Kelly strategy asymptotically maximizes the worst-case rate of asset growth, and dominants any other essentially different strategy by magnitude. Our results extends Breiman's theoretical result and justifies that the distributional robust Kelly strategy is the optimal strategy in the long-run for practical betting with uncertainty.
Submission history
From: Qingyun Sun [view email][v1] Thu, 20 Dec 2018 23:10:46 UTC (234 KB)
[v2] Sat, 6 Jul 2019 23:22:18 UTC (640 KB)
[v3] Thu, 10 Jun 2021 16:15:16 UTC (456 KB)
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.