Catching VR & PC Game Trends! Building an Automated Ranking Monitoring System
This article was automatically translated from Japanese by AI. Read the original Japanese version →
Table of Contents
Keeping up with new releases and trending titles quickly is super important for gamers, but checking multiple stores every day is a hassle.
So I built a system that periodically monitors rankings on Meta Quest Store, Steam, and other platforms, and automatically notifies me of games that have newly entered the rankings (i.e., games gaining attention right now). It also served as a great exercise for practicing AI coding.
In this post, I'll introduce its design and how it works.
1. What I Wanted to Achieve (System Requirements)
The core idea of this system is "catching changes." Instead of simply displaying the current rankings, it compares them against past history and extracts only the titles appearing in the rankings for the first time.
Target Platforms and Information Sources
This system targets three major game platforms: Meta Quest Store, Steam, and itch.io.
First, for the Meta Quest Store—the primary store for standalone VR devices—I chose Quest Store DB as the data source. The official store has strict scraping restrictions and complex data structures, whereas Quest Store DB has fresh data and makes it easy to fetch pre-filtered rankings like "Paid rankings over the past week," making it ideal for this purpose.
Next, for Steam—the world's largest PC gaming platform—the system monitors the official Top Sellers page. Steam offers detailed search filters, allows locking language and region via URL parameters, and makes it easy to reliably extract each game's AppID (unique identifier) from the HTML response, making it well-suited as a monitoring target. Right now, I'm filtering the data collection specifically to VR game rankings, which I'm especially interested in.
Finally, there's itch.io, a treasure trove of indie games. The official featured games RSS feed and similar lists only reflect current popularity rankings, so by comparing them against my own historical data in this system, I can extract previously unseen titles breaking into the top ranks as "new arrivals." Currently, I'm focusing on games for specific platforms like Playdate and Pico-8.
How It Works (Basics)
- Fetch the current rankings (e.g., TOP 50) from the store (or DB site).
- Compare them against "all past history" stored in S3 to identify games appearing for the first time.
- Notify the user with details of the newly discovered games (title, current rank, URL, etc.).
2. Processing Flow
To give an overview of the entire system, I put together a sequence diagram of the high-level flow. There are several processing steps, and all of them run on AWS serverless services.
sequenceDiagram
participant EB as EventBridge (Scheduler)
participant SF as Step Functions (Workflow)
participant Lambda as AWS Lambda (Logic)
participant S3 as Amazon S3 (Storage)
participant SNS as Amazon SNS (Notification)
EB->>SF: Trigger execution (once daily)
SF->>Lambda: Fetch rankings (Fetch)
Lambda-->>Lambda: Fetch from Quest Store DB
Note right of Lambda: Scraping logic generated by Gemini
Lambda->>S3: Save latest snapshot
SF->>Lambda: Compare data (Compare)
Lambda->>S3: Retrieve past history data
Lambda-->>Lambda: Extract diff by game_id
Lambda->>S3: Update history data
SF->>SF: Check if new games exist
alt New games found
SF->>Lambda: Send notification (Notify)
Lambda->>SNS: Publish notification message
SNS->>User: Email / Feedly notification
end
3. Behind Diff Extraction: Data Management in S3 & Comparison Logic
To accurately determine what counts as "new," I use Amazon S3 for data management.
Data Structure on S3
Two types of JSON data are stored in S3.
Ranking Snapshot: Complete ranking data for each execution. Used for debugging and analyzing historical rank trends.
{ "platform": "meta-quest", "timestamp": "2026-01-17T09:00:00", "games": [ {"game_id": "12345", "title": "Great VR Game", "ranking_position": 1, ...} ] }Game History: Master data of every game that has "ever appeared in the rankings at least once" on that platform.
{ "12345": { "title": "Great VR Game", "first_seen": "2025-12-01", "last_seen": "2026-01-17", "max_rank": 1 } }
Comparison Algorithm
The Lambda function responsible for comparison extracts "new arrivals" using the following flow:
- For each game in the newly fetched ranking, check whether its
game_idexists in Game History. - If it does not exist, add it to the notification list as a "new game."
- At the same time, update Game History: add a new entry if it's new, or update
last_seenand overwritemax_rank(highest rank) if it already exists.
Matching by game_id (the store's immutable unique ID) enables accurate diff extraction without duplicate notifications, even if a game's title changes slightly.
4. Tech Choices: Why This Architecture?
Here are a few reasons behind the technical decisions I made when designing the system's architecture.
Why S3 Instead of an RDB or Cache DB?
Usually, RDS (relational databases) or DynamoDB are considered for data persistence, but I chose Amazon S3 as the data store here.
- Low update frequency: Rankings are fetched only about once a day, so real-time performance isn't required.
- No transactions needed: There aren't high-frequency or multi-user writes requiring complex concurrency control or consistency guarantees.
- Low cost: S3 storage costs are extremely cheap, and with this data volume it fits almost entirely within the free tier.
For the requirement of "a once-daily batch job that only needs to compare against the previous state," snapshot management on S3 was the simplest and most economical choice.
Why Email Notifications Instead of an RSS Feed?
The notification destination, Feedly, is primarily designed for subscribing to RSS feeds. I could have formatted the new game updates as RSS (XML) and hosted them publicly on S3, but I deliberately chose email notifications instead.
- Simpler implementation: Simply sending an email via SNS (Simple Notification Service) involves much simpler logic than formatting RSS feeds and managing publishing/updates.
- Privacy and safety: Hosting an RSS file on S3 means anyone who knows the path could access it, and even for personal use, I felt reluctant about putting any data into a publicly accessible state.
- Good enough for personal use: Since the goal is just for me to catch the info myself, receiving emails from SNS via Feedly's "Email to Board" feature gives a reading experience just as good as RSS.
By leveraging the battle-tested technology of sending emails, I kept development costs and security concerns to a minimum.
5. Collaborating with Gemini & a Safe Development Workflow
In building this system, I collaborated extensively with the AI assistant Gemini. In particular, the scraping scripts for parsing complex HTML structures, the AWS Step Functions definition (ASL), and the CloudFormation templates are all based on code generated by Gemini.
At first, I wrote the code using Antigravity while carefully adjusting the plan, and later handled tweaks and feature additions with Jules.
However, deploying AI-generated code as-is isn't great from a quality or security perspective. So for this project, I set up automated checks using pre-commit.
Quality Control with Pre-commit
Whenever code is committed to the repository, the following tools run automatically to catch issues before they happen:
- Ruff: Static analysis and auto-formatting for Python code.
- cfn-lint: Checks whether CloudFormation templates follow AWS best practices.
- markdownlint: Documentation formatting checks.
- actionlint: Prevents misconfigurations in GitHub Actions.
- asl-validator: Validates the syntax of Step Functions workflow definitions.
- pytest: Ensures logic correctness with unit tests.
This workflow of "having Gemini write the prototype, then verifying safety with static analysis tools and tests without manual intervention" lets me develop both quickly and safely.
6. Aggregating Updates via Email Notifications to Feedly
For the notification destination, I use Feedly's "Follow Newsletters in Feedly" feature.
Notification emails sent from AWS SNS are treated directly as articles inside a designated Feedly feed. This lets me check automatically discovered new games in the exact same way I browse news sites. Since information from multiple platforms (Quest, Steam) is aggregated in one place, keeping up with trends has become much more efficient.
7. Summary
That was a look at the details of the game ranking monitoring system I built using AWS.
By combining low-cost serverless architecture, simple diff extraction using S3, and a safe development workflow supporting AI collaboration, I was able to put together a low-maintenance system.