Deep-Learning Game Server Predicting Puzzle Difficulty
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Evaluating game difficulty for new stage maps in puzzle games requires extensive manual playtesting, which is time-consuming and costly, as it necessitates multiple plays by many users to determine accurate difficulty levels.
Innovation Solution
A deep-learning based game play server that performs reinforcement learning to predict game difficulty by arranging puzzles on a stage map, generating training data sets, and adjusting difficulty levels using binomial regression analysis to induce user interest through dynamic difficulty adjustment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual playtesting is conducted to evaluate game difficulty for new stage maps, then measurement precision of game difficulty is improved, but loss of time and productivity deteriorate due to requiring dozens to hundreds of plays by multiple users
Solution Approach 1:
The patent creates virtual copies of human players through AI agents that can autonomously play the game. These agent players replicate human gameplay behavior through machine learning, eliminating the need for actual human players to perform repetitive playtesting. The agents can evaluate multiple stage maps simultaneously, reducing evaluation time from dozens to hundreds of plays to automated computational processes.
Solution Approach 2:
The patent replaces the mechanical system of manual human playtesting with an automated AI-based evaluation system. The AI agents use reinforcement learning and value network models to automatically assess stage map difficulty, substituting human physical actions with computational algorithms that can process and evaluate game scenarios much faster than human players.
2Measurement precision
If manual playtesting is conducted to evaluate game difficulty for new stage maps, then measurement precision of game difficulty is improved, but device complexity and cost increase due to requiring many persons and multiple plays
Solution Approach 1:
The patent develops a universal AI evaluation system that can assess multiple different stage maps across various game modes simultaneously. The same AI agent framework and value network model can be applied to evaluate different puzzle types, difficulty levels, and map configurations, eliminating the need for separate evaluation systems for each game scenario.
Solution Approach 2:
The patent uses AI agents that copy human player behavior patterns to perform evaluation tasks. These virtual agents replicate human decision-making and gameplay strategies, providing a simplified yet effective evaluation mechanism that doesn't require coordinating multiple actual human players or complex test management infrastructure.
3Reliability
If extensive manual playtesting is conducted to evaluate game difficulty, then reliability of game difficulty assessment is improved, but productivity deteriorates due to the large number of plays required
Solution Approach 1:
The patent implements continuous automated evaluation through AI agents that can play and assess stage maps without interruption. Unlike manual playtesting that requires scheduling, coordinating, and managing human players across multiple sessions, the AI system performs continuous evaluation, maintaining consistent assessment criteria and providing reliable results through uninterrupted automated gameplay and data collection.
Data Source
AI summary
A method and apparatus for predicting game difficulty by using a deep-learning based game play server predict a difficulty of stage maps of a match puzzle game using a deep-learning based game play server that performs the match puzzle game and modify the stage maps. The deep-learning based game play server includes: a communicator configured to receive first stage maps of a first size and second stage maps of a second size; memory configured to store an agent model; at least one processor configured to perform learning of the agent model and perform the match puzzle game using the learned agent model.