skip to content
meistyr.tech/~luke
case study/001/in alpha

Car Soccer City

A league management platform for an established competitive league. It replaces their spreadsheet-based operations with a database-driven system, built against real stakeholder deadlines.

see what led herein alpha, not yet public
role
Full Stack Engineer · Freelance
period
May 2026 — Present
status
In alpha with league stakeholders
01 — brief

Why they reached out

An established league running on spreadsheets reached out to me after my work on Royalty Series. Everything is manual, their history is spread across different files, and it makes it difficult to scale in future seasons. They need a real platform.

02 — architecture

How the system is set up

platform architecture

Select a part to see what it does and how it connects.

API and web app in TypeScript · Railway
03 — decisions

The decisions behind it

Prisma over hand-written SQL

On Royalty Series the schema lived in SQL and the code’s idea of it lived in TypeScript, kept in step by hand. Here the Prisma schema is the one source. Migrations and types are generated from it, so a schema change that the code doesn’t match fails at compile time.

cost

Less control over the exact SQL that runs, and one more dependency that can hold up upgrades to the rest of the stack.

PostgreSQL over MariaDB

The schema keeps history instead of deleting rows. A player who leaves a roster gets a leave date and the row stays. That means most of the league’s rules only apply to active rows. A player is on one roster per season, a team has one captain, a match has one live report, and a game account belongs to one person. Postgres can enforce each of those in the database with a partial unique index, a unique rule that only covers rows that are still active. MariaDB has no partial indexes, so those rules would live only in application code.

cost

Leaving the database I had run in production for two years. The platform hasn’t launched, so there was no data to move.

Own components over a UI library

I looked at the UI libraries and none of them fit the design I had for the site. Every component is written for the league’s design instead, so nothing has to be overridden to match it.

cost

Every button, table and form is mine to build and keep accessible.

The API as the only way in

The web app and the Discord bot both go through the API, and only the API talks to the database. Nothing outside the system connects to it, and the league’s rules and permissions are checked in one place instead of being copied into the bot.

cost

Every bot feature needs an API endpoint first, the bot is down whenever the API is, and each action takes an extra network hop.

Replay parsing in a separate process, not a separate service

Each replay is parsed in its own process, started by the API. The parser’s output is about 130 MB once read, so keeping it out of the API keeps that memory out too, and a malformed replay can only crash its own process. Each team plays two best-of-five matches a week, so at most ten replays. The number of teams changes each season, but it’s known once registration closes, before any replays come in. The load is known going in and doesn’t need a service of its own.

cost

Parsing shares the API’s server and can’t be scaled on its own.

04 — scope

What it needs to do

01
League and season structure
02
Team and roster management
03
Scheduling and match reporting
04
Standings and statistics
05
Roles and permissions
06
Deployment and release process
05 — status
In alpha with league stakeholders, with regular progress updates.

I’m the only developer, so I make the architecture decisions and I’m accountable for them. The full case study will be posted once the league has run a season on it.

next

Royalty Series

The production platform that came before this: seven competitive seasons, 7,100 users and no unplanned downtime.