Proposal: Align Three-Cushion Ball Size with UMB Standards (61.5 mm)

[osmane]

Hello @tailuge & team,

First of all, thank you for the Billiards simulation project.

I’d like to propose a modification specific to the three-cushion (three-rail) mode:

Current Issue

From visual inspection of the three-cushion view in the demo, it appears that the balls are slightly smaller in proportion to the table than what the UMB (Union Mondiale de Billard) regulations prescribe. This difference is subtle but noticeable to anyone who carefully compares it with real-world carom billiards measurements.

UMB Official Measurements

According to the UMB regulations for three-cushion billiard:

• Ball diameter: 61.5 mm
• Table playing area (inner dimensions): 2840 mm × 1420 mm

Therefore, the correct ball-to-table ratio (ball diameter ÷ long side) is approximately 61.5 / 2840 ≈ 0.02165, i.e., about 2.17 %.

[osmane]

I fixed these issues and a few more in the project with chatGPT 5 and Gemini 2.5 Pro and published it on my own github. For those interested, the url is https://github.com/osmane/billiards

Improving the realism and visual quality of the 3-cushion game mode. The table dimensions have been updated to the official 2.84m x 1.42m measurements. To support this, a dynamic scaling feature was implemented to ensure the visual table model accurately matches these new physical dimensions.

Additionally, there were several visual enhancements, including a two-step fix to the ball view resolution and the activation of anti-aliasing for smoother graphics.

The public folder was added to house static files. The .gitignore file was also updated to exclude the dist folder from version control.

The README.md file was updated for the active fork.

A new version of the project was deployed.

[tailuge]

This is great. Thank you so much for doing this.

It is definitely less daunting and a better feel with the bigger balls. I based my values off some of the original papers on the physics so happy to shift to these new values. I will aim to update my code and handle the scaling - I think I have to rework the table blender model to match. I might keep the old school graphics for now, it helps it run on phones and raspberry pi.
I notice you git ignore the /dist folder, but that is the part that deploys onto github so you will have to update it manually (not important if you run it locally)

The physics for cushion bounce is pretty good, but the constants used really need updating to better match reality. A guy from NightCafeBilliard has captured real table positions that can be used to tune the simulation e.g. https://tailuge.github.io/billiards/dist/diagrams/gemini.html?shot=545600

There is some discussion of this on the PoolTool forums https://github.com/ekiefl/pooltool

[osmane]

I'm glad you liked it.

Thanks for sharing the details, that's a solid approach. Referring to the original papers adds a nice foundation to the project. Keeping the old-school graphics for performance on phones makes a lot of sense. What caught my attention is that the project is built on JavaScript and touches on something less common like 3-cushion billiards. Curious to see where you take it next.

I wanted to clarify the changes to clear up any confusion. I set up a new GitHub Actions workflow that automatically builds the /dist folder from the source files and deploys it to GitHub Pages on every push to main.

Because of this new automation, the dist folder no longer needs to be tracked in the repository. Adding it to .gitignore helps keep the repository clean. So, the deployment is now fully automated. :)

I tried out these changes to the project's build and deployment process, with plenty of pointers from LLM's.

[tailuge]

Hi @osmane , I finally got around to making the change and resized the 3c table. I also added a bot player (I knew the right way so just told the LLM how to go about it without too much change). I see your fork has gone, hope you can use this one for now.