All projects
AISoftware

SeaTyrants — Anti-bot AI

Real-time bot detection via ONNX Runtime

An AI that exposes bots before they rot the game.

The client need

Context

The SeaTyrants MMORPG is exposed to automated behaviours that degrade the experience of legitimate players.

Problem

Classic heuristics fall short against increasingly sophisticated bots.

Goal

Detect bot behaviours in real time from gameplay signals, without penalising honest players.

What’s at stake

  • Protect the game's economy and fairness

  • Hold real-time performance on the server side

  • Limit false positives on legitimate players

The proposed solution

What was built

An ONNX Runtime integration in C/C++ that loads trained models and evaluates them on gameplay signals.

Why this choice

ONNX Runtime allows ML inference directly on the server side, natively, with minimal latency.

How it answers the need

Offline training pipeline, ONNX export, C/C++ inference plugged into the server-side gameplay flows.

Key features

  • Server-side ML inference in C/C++

  • Real-time detection on gameplay flows

  • Confidence score per player

  • Models retrained and redeployed without touching the binary

  • Continuous improvement loop

Results

  • Server-side bot detection without degrading the experience

  • Inference latency compatible with a real-time MMORPG

  • Ability to iterate on models independently from the game

Tech stack

Technical choices used on this project.

CC++ONNX RuntimePython (training)

Do you have a similar need?

Let’s talk about your project. I help you adapt this approach to your context.