← All work
Mobile Product / Gaming · Flagship

Morabaraba
Africa's chess.

Category
Mobile Product / Gaming · Flagship
Year
2025
Role
Founder, TME Group

A modern digital implementation of the traditional African strategy game Morabaraba — built with the ambition of making it a globally recognised strategy game: Africa's chess. The project combines cultural heritage, software engineering and modern mobile product development.

Context

Morabaraba is a traditional African strategy game, played in Southern Africa for centuries: boards carved into rock at sites such as Mapungubwe are estimated to be 800 years old. It is recognised in South Africa as an indigenous sport, governed by Mind Sports South Africa, and has had World Championships since 1997. Its digital versions had not kept up: a handful of generic Mill-game apps, playable but with weak opponents, no ratings and no competitive play.

Challenge

Make a traditional game compelling enough to be taken seriously as a global strategy game — which means a strong engine as well as a polished mobile product. There was a catch: Morabaraba is solved. With perfect play the first player can force a win (Gévay and Danner, 2014), so an engine that simply plays perfectly would be unbeatable and no fun. Strength had to be calibrated for people, from first-timers to tournament players.

Approach

Treat Morabaraba the way the best chess platforms treat chess: as a game whose value grows through ratings, competition and learning, not a one-off app. The engine came first, then was shaped into named South African opponents, each with its own rating and personality, so every player has a fair game. The app was built in Flutter for the mid-range Android phones most players use, with the search running in the background so the board never stalls.

Architecture

  • Engine: minimax search with a multi-term evaluation function
  • Openings: custom opening book
  • Opponents: 16 AI characters, from Lesedi at 400 ELO to Zandile at 3000 ELO
  • Neural tier: the four strongest opponents evaluate positions with a small neural network (NNUE) trained on a million self-play games
  • Online: ranked matches, live ELO ratings, tournaments and leaderboards
  • Distribution: App Store and Google Play (10K+ installs on Google Play, rated 4.7)
  • App: Flutter
  • Search: alpha-beta minimax with an opening book and Zobrist hashing, run on a background isolate
  • Backend: Firebase: Auth, Firestore and Realtime Database, with Cloud Functions for matchmaking, tournament scheduling and rating integrity

Outcome

Live on the App Store and Google Play, with 10K+ installs on Google Play and a 4.7 rating. Players can take on 16 AI opponents from 400 to 3000 ELO, or play ranked online matches with live ratings, tournaments and leaderboards.

Lessons

To be updated.

FLUTTER · MOBILE DEVELOPMENT · GAME LOGIC & AI ENGINE · PRODUCT DESIGN

Let's build something
meaningful.

I enjoy conversations around data, artificial intelligence, technology leadership, digital products and the future of enterprise technology.