n-Player Stochastic Duel Game Model with Applied Deep Learning and Its Modern Implications.

Area of Science:

• Game Theory and Computational Modeling
• Stochastic Processes and Operations Research

Background:

• The theory of stochastic duels has evolved, with limited exploration of modern applications.
• Existing combat assessment models often lack realism for contemporary scenarios.
• Video games offer a rich source of data for analyzing combat dynamics.

Purpose of the Study:

• To provide a conceptual foundation for stochastic duels and explore game models.
• To introduce a novel evaluation model for realistic combat scenarios, including player intent and teaming.
• To investigate the application of stochastic duel theory in modern contexts, particularly through video game data.

Main Methods:

• Literature review and timeline of stochastic duel theory development.
• Introduction of a new pair-wise duel evaluation model with a trust mechanism.
• Development and training of a deep-learning model using data from shooter video games.

Main Results:

• The proposed model, unlike conventional ones, handles pair-wise duels throughout the game.
• A deep-learning model trained on video game data supports the novel evaluation model's efficacy.
• The new model demonstrates improved realism by considering player intent and teaming.

Conclusions:

• The novel stochastic duel model offers a more realistic approach to combat assessment.
• Video game data provides a viable resource for training and validating combat models.
• This research lays groundwork for innovative applications of game theory in solving real-world problems.