Lily's Garden
GameBy Tactile Entertainment · Released Feb 4, 2019
Documents about Lily's Garden
Newzoo Puzzle Games Report: Mobile Game Genre Insights
Newzoo and Pangle's joint genre report examines mobile puzzle games, comparing Western and Eastern markets (the U.S., UK, Japan, South Korea, and China) to help developers and publishers understand regional differences in monetization, marketing, and player demographics.
- The United States is the world's largest mobile puzzle market, followed by Japan and China.
- Western puzzle games rely heavily on in-app advertising (IAA) and simple economies, while Eastern titles—especially in Japan—prioritize deep character collection and gacha mechanics to drive higher in-app purchase (IAP) revenue.
- Western players demonstrate a higher tolerance for frequent ad breaks, whereas Japanese players prefer longer, uninterrupted sessions and show a greater willingness to pay for aesthetic enhancements and additional functions.
Procedural Content Generation of Puzzle Games using Conditional Generative Adversarial Networks
Procedural content generation in match-three puzzle games, specifically within the context of titles like Lily’s Garden, can be effectively addressed through the application of Conditional Generative Adversarial Networks.
- Conditional Generative Adversarial Networks (CWGAN-GP) outperform WGAN-GP-PE architectures in generating functional, symmetrical, and structurally coherent match-three puzzle levels.
- Current machine learning models successfully automate the creation of complex map shapes but struggle to accurately map piece-distribution vectors.
- Procedural generation models currently lack the nuanced understanding of gameplay balance required to produce production-ready content for titles like Lily’s Garden.
Strategies for Using Proximal Policy Optimization in Mobile Puzzle Games
Automating play-testing in mobile puzzle games requires overcoming significant technical hurdles associated with reinforcement learning, particularly the instability and overfitting inherent in Proximal Policy Optimization (PPO).
- The CSResetMaskV2 framework is the most effective configuration for training PPO agents, achieving the highest completion rates and human-like gameplay patterns in match-three titles like Lily’s Garden.
- Integrating action masks directly into the observation space as a partial forward model significantly improves training speed and reliability compared to traditional hard action masking, which often causes algorithmic collapse.
- To prevent catastrophic learning failures and suboptimal loops, developers should utilize a combination of environment resets, color shuffling, and soft action masking.