Reflex Lab
itzik123.github.io- Category
- Entertainment
- Rank
- No. 301Apps board
Previous survey · No. 285 ·
- Type
- GAME
- Date
About
Reflex Lab is an interactive browser demo in which a small neural network learns to defend a tower against Clash Royale attacks. A Clash Royale simulator written in C++ and compiled to WebAssembly scores each try, and the network learns with the REINFORCE policy-gradient method. Users can first defend by hand and compare their score with the network's. It is an unofficial fan project, not affiliated with Supercell.
At a glance
Reflex Lab is an interactive demonstration of a neural network learning where and when to place a defender in a small Clash Royale simulation. It is worth opening if you want to see training attempts and their scores change in real time.
Good fit
Players curious about timing and placement, and learners who want a visual example of how a simulator can score an agent’s choices.
Consider something else when
Look elsewhere if you want to practice full matches or judge how an agent would play the live game. The lab uses one attacker, one defender and no elixir.
Start here
Start the lab and skip to the AI if you want to watch first. Train briefly, pause and open See results. Compare damage prevented on the five attacks with the random and best-placement baselines.
Cost and requirements
The demo trained in a desktop browser without a sign-in, payment or local installation. The project describes the scoring engine as C++ compiled to WebAssembly.
Limits to know
Results apply to the five attacks and the selected matchup. The short test showed a result for each attack. It did not measure the full-match agent, performance across devices or how the same training run would turn out again.
Sources checked 2026-11-05. Source-based assessment.
Tags
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.