Optimal Blackjack Betting Strategies Through Dynamic Programming and Expected Utility Theory
Lucas Bordeu, Javier Castro · 2025
Read Optimal Blackjack Betting Strategies Through Dynamic Programming and Expected Utility Theory by Lucas Bordeu, Javier Castro (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-00724.txt.
Introduction
Lucas Bordeu and Javier Castro compute best play and best betting in blackjack with formal math, in a 2025 paper. They model the game as a Markov decision process, where each state calls for a choice. Expected utility theory then scores choices by the player's taste for risk, from steady to bold.
Dynamic programming yields the top policy for both card play and stake sizing. Simulations show the best play policy only edges out the classic basic strategy, which speaks well of that folk system. Betting by exact deck makeup slightly beat the Hi-Lo counting scheme. The authors supply both brain-friendly versions and full computer versions. Rule changes and more players remain open ground.
Catalog record
| CREATOR | Lucas Bordeu, Javier Castro |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 154 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-00724
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Optimal Blackjack Betting Strategies Through Dynamic Programming and Expected Utility Theory Lucas Bordeu, Javier Castro Abstract This study presents a rigorous mathematical approach to the optimization of round and betting policies in Blackjack, using Markov Decision Processes (MDP) and Expected Utility Theory. The analysis considers a direct confrontation between a player and the dealer, simplifying the dynamics of the game. The objective is to develop optimal strategies that maximize expected utility for risk profiles defined by constant (CRRA) and absolute (CARA) aversion utility functions. Dynamic programming algorithms are implemented to estimate optimal gambling and betting policies with different levels of complexity. The evaluation is performed through simulations, analyzing histograms of final returns. The results indicate that the advantage of applying optimized round policies over the “basic strategy” is slight, highlighting the efficiency of the last one. In addition, betting strategies based on the exact composition of the deck slightly outperform the Hi-Lo counting system, showing its effectiveness. The optimized strategies include versions suitable for mental use in physical environments and more complex ones requiring computational processing. Although the computed strategies approximate the theoretical optimal performance, this study is limited to a specific configuration of rules. As a future challenge, it is proposed to explore strategies under other game configurations, considering additional players or deeper penetration of the deck, which could pose new technical challenges. keywords Blackjack Policy Optimization, Markov Decision Process, Dynamic Programming, Expected Utility Theory, Risk Aversion Collaboration Note: This work was carried out in collaboration with Prof. Andrés Abeliuk (Department of Computer Science, Universidad de Chile), who provided valuable feedback on the structure and academic content of this manuscript. 1 Introduction The card game “Blackjack” has captured the imagination of millions of people around the world because of its apparent simplicity, but underlying complexity. Over the years, strategies to increase the chances of winning at this card game have evolved from mere tips and tricks, to complex mathematical and computational systems. Motivated by curiosity and passion for Blackjack, this paper studies optimal strategies for the game, modeling the problem as a particular case of the dynamic portfolio optimization problem, highlighting the game’s connection to financial theory. The player’s complete strategy is divided into a ”round policy” and a ”betting policy.” The round policy consists of decision-making during each round of play and has been extensively explored by the community. Its optimization is considered partially resolved Peytaví Machado (2018) ; “MGP” () ; “k_c” () , culminating in good approximations limited by computational costs. In contrast, the optimization of the betting policy remains a less explored area. Although some studies have examined the problem in a simplified manner Vince and Zhu (2013) ; Werthamer (2005) ; Jensen (2014) , none have formally modeled it or addressed it rigorously. In this paper, a rigorous mathematical approach is proposed for optimizing both policies, employing the theoretical framework of Markov Decision Processes (MDP) and expected utility theory. The developed model considers the relevant variables influencing the player’s returns, assuming that the round policy maximizes the expected return of an individual round, while the betting policy seeks to maximize the expected utility of an entire betting session. This approach is versatile enough to capture known strategies, such as the Basic Strategy Shackleford () and betting strategies based on the Hi-Lo card counting system Wikipedia (2024) . The optimization of the round policy is performed through dynamic programming, using efficient algorithms based on vector structures to minimize the required computational costs. As a result, a semi-optimal round policy is computed, which includes a simplification in the case of the ”Splitting” option. Additionally, the probability mass function of the return for each round is estimated, both under the Basic Strategy and the semi-optimal policy. On the other hand, the betting policy is also optimized through dynamic programming, considering a discretization of the state space and a simplification of the transition function. Betting policies based on the Hi-Lo counting system and full knowledge of the deck composition are considered. The study concludes with the derivation of optimized betting policies for different levels of risk aversion, along with analytical estimates of the return distributions. This is the first work to rigorously address the problem of betting optimization in blackjack, and it is conjectured that the obtained results are close to the theoretical optimum under the assumed simplifications. Therefore, this study represents a formal solution to the game under a specific configuration of rules and conditions, and it opens the possibility of solving it for other configurations and generalizations of the game in future research. The article is structured to ensure logical continuity and facilitate understanding. First, a comprehensive theoretical framework is presented that mathematically models the optimization problem, including optimality conditions under certain assumptions. Then, the methodology employed to compute the policies and evaluate their performance is detailed. Finally, the obtained results are discussed, and the study is concluded by highlighting its implications and suggesting possible lines of future research. 1.1 Game Configuration There are multiple aspects that fully define a session of Blackjack, so it is essential to be precise when referring to the game as the object of study. This section describes the ”game dynamics” and the ”rules and conditions” assumed throughout the research. Game dynamics refer to a description of what the game involves and how it unfolds. The rules and conditions refer to a specific set of rules and assumptions considered for this particular study. 1.1.1 Game Dynamics Blackjack is a card game played at a casino table or through online platforms, where multiple players can participate simultaneously, though each competes individually against the house, represented by the dealer. The essence of the game lies in strategy and decision-making based on the value of the dealt cards, with the goal of achieving a score close to or equal to 21, without exceeding it. 1. Start of the Round and Betting: Before the round begins, each player at the table places an initial bet. This bet is required for the player to participate in the upcoming round. Bets are placed in a designated space in front of each player. 2. Dealing the Cards: Once the bets are placed, the dealer proceeds to deal the cards. Each player receives two face-up cards, while the dealer receives one face-up card and, depending on the game variant, may receive another face-down card (known as the ”hole card”). 3. Card Values: • Number cards (2 to 10): Worth their face value. • Face cards (J, Q, K): Worth 10 points. • Aces: Have a dual value, worth either 11 points or 1 point, depending on what is more advantageous for the player’s hand. An Ace counts as 11 unless that value would cause the hand to exceed 21 points, in which case it counts as 1. 4. Round Progression: After the initial deal, players evaluate their hands and proceed with their actions, choosing from the following options: • Hit: Request an additional card to try to get closer to 21. • Stand: Keep the current hand without receiving additional cards. • Double Down: Double the initial bet in exchange for receiving only one more card. • Split: If the two initial cards have the same value, the player