AI Strategy for Game Launch(2)

AI Guide for Engagement & Retention

ℹ️ Redesigning Game Structures to Reduce Churn

The churn prediction and automated rewards described above focus on catching players right before they leave. But sometimes, the more effective strategy is to redesign the very structure of the game to prevent churn from happening in the first place.

Supercell’s 《Brawl Stars》 made such a bold move in 2023 by completely removing its loot box system. This was not just a monetization model change, but a strategically prepared decision backed by data and ML-driven simulations.

The decision stemmed from three key factors. Player frustration was mounting as many failed to obtain the characters they wanted, regulators were intensifying scrutiny, and doubts about the long-term sustainability of random monetization models were growing.

To address this, Supercell introduced the ‘Starr Drops’ system, allowing players to unlock desired Brawlers directly with points earned through gameplay.

By removing the need for random box openings, the system gave players clear goals and tangible rewards, restoring trust. Premium currency usage also shifted from chance-based draws to predictable value offerings such as time-savers, exclusive skins, and battle pass unlocks.

The game economy was overhauled in full — growth speed, unlock pacing, and reward loops were repeatedly simulated and fine-tuned through dozens of A/B tests. ML-based retention and monetization prediction models were applied in advance to forecast the structural impact of the shift.

The results were positive. Players gained confidence that “I get rewarded for the time I put in,” onboarding barriers for new users dropped, and churn rates for existing players decreased. While monetization initially dipped due to the loss of loot box excitement, ARPPU (average revenue per paying user) eventually rose as players shifted spending toward predictable, value-driven products. Operationally, user feedback became easier to anticipate, legal risks were reduced, and global market adaptability improved.

This case demonstrates that churn prevention is not only about offering rewards. It is about how AI and data are used to redesign the fundamental structures of a game. Titles with fast early growth but repeated churn, those with fatigue from random monetization, or those with widening pay-to-play gaps and community conflict can draw practical insights from this case. Structural innovation powered by data and AI is proving to be the most effective way to ensure long-term sustainability in the game business.

ℹ️Guarding Fairness: AI-Driven Anomaly and Cheating Detection

The foundation of any game service is ‘fairness.’ Hacks, macros, and farming accounts that flood in during launch can destroy player trust in the early stages. The challenge is that these accounts often behave almost like normal players.

For example, a VIP who plays all day looks similar immediately to a bot running macros. AI distinguishes these “abnormalities that appear normal” through pattern analysis.

AI-based anomaly detection systems generally operate in two ways:

First, unsupervised learning models automatically detect accounts whose behavior deviates significantly from most normal players.

For instance, if normal click intervals average above 300ms but a certain account clicks hundreds of times at 50ms intervals, it is flagged as likely a bot.

Second, supervised learning models classify behaviors by comparing current users to data from previously identified cheaters or farming accounts. Abnormal win rates in PvP or repetitive patterns such as breaking down and reassembling the same item dozens of times are examples of behaviors flagged by the model.

Nexon has reduced operational risk by applying AI-based anomaly detection to its games. For example, accounts with unusually high win rates in certain maps were automatically clustered, enabling operators to investigate potential hacks, bugs, or exploits early — even without user reports.

NCSoft uses AI to detect farming accounts in real time by analyzing trading patterns, device usage, and movement routes, then clustering accounts with similar behaviors. This approach makes it possible not just to block single accounts, but to disrupt entire farming networks at once.

AI anomaly detection is not just a technical safeguard but a strategic pillar of maintaining a fair game environment. Once players feel “this game is fair,” trust itself becomes more powerful than content.

ℹ️ When NPCs Start Talking Back: AI-Powered Content Generation

In the past, NPCs repeated the same static lines. Today, NPCs can change their tone and dialogue depending on a player’s behavior and situation — level, recent quest success rate, play patterns, even emotional response logs. At the heart of this shift is generative AI (GenAI).

Ubisoft’s Ghostwriter project enables an automatic generation of NPC dialogue tailored to context and emotional nuance. Writers no longer need to craft dozens of variations manually. With prompt-based inputs, dialogue style and content are generated automatically. As a result, content production time was cut by more than half, freeing writers to focus on creative storytelling.

NCSOFT, meanwhile, has experimented with its proprietary LLM, VARCO, to create personalized in-game messages based on player login patterns, playstyle, and guild activity. Compared to generic event messages, these AI-generated messages doubled click-through and quest acceptance rates.

Generative AI does not replace writers or operators. Instead, it alleviates repetitive content creation tasks while enabling real-time, emotion-driven personalization. This expands the sense of “this game speaks to me” for players.

♻️ Final Thoughts: AI Is Not Just Technology, but a Practical Strategy

The success of game operations ultimately hinges on player experience. That experience depends on ‘how the game presents itself, how quickly it responds, and how personalized it feels.’

Technologies that interpret data, predict behavior, and generate content are now available in SaaS form, ready for immediate adoption. Game business professionals — PMs, planners, marketers, and operators — can directly design “what kind of experience to give to which players” without needing to be engineers. At the center of this evolution stands AI.

In the next article, we will explore how AI enables real-time operational automation that improves stability and quality after launch.

※ Disclaimer: This content reflects the author’s personal views and includes only publicly available examples. It does not represent the official position of any company mentioned