Poker Chipper

Optimally select poker chip denominations for cash games using constrained, nonlinear optimization.

About

Imagine hosting a small poker game. You know how many friends will attend, how many chips you have, and the buy in. How do you pick chip denominations?

Without Poker Chipper, picking chip denominations is manual and error-prone. You first try to pick values that divide each other nicely and are easy to remember. (Alternatively, struggle to recall the values from the last time you played.) You next try to find a way to make the chosen values sum to the buy in. Then, you backtrack whenever there aren't enough chips for everyone, or when there are too many chips left over. Finally, you settle for a suboptimal chip allocation so that you can actually start playing.

With Poker Chipper, on the other hand, picking chip denominations is fast and easy. Input the number of players, buy in, blinds, and number of chips. After waiting a moment for the mathematically optimal results, you can save them for next time, or share them with others. If you want to modify the results, adjust the "advanced options," and the chip denominations will be recomputed accordingly.

How Poker Chipper Works

Poker Chipper uses mixed-integer nonlinear programming (MINLP), which is a form of constrained, nonconvex optimization, to optimally pick poker chip denominations.

Optimization is performed using a "solver" – in this case, SCIP. Some inputs to Poker Chipper are translated into solver constraints, which affect what denominations are admissible. Other inputs influence the solver's objective function, which affects how candidate denominations are evaluated, ranked, and chosen. In the user interface, inputs affecting constraints, and inputs affecting the objective function are described as influencing the "requirements" and "score," respectively, of candidate solutions.

Poker Chipper is a fully static web application. In other words, all of the processing involved in performing optimization occurs client-side – entirely in the user's browser. The SCIP solver is designed to run natively, so Poker Chipper bundles it for the browser by compiling SCIP to WebAssembly (WASM) with Emscripten.

How to Read the Code

Code Table of Contents

The links below are listed in the order the code should be read to understand the application from the highest to lowest level.

• src/App.svelte – main high-level application
• src/*.svelte – UI components
• src/solve.js – convert user inputs to MINLP model and solve using WASM SCIP
• src/solveWorker.js – web worker to run solver in non-blocking thread
• public/serviceWorker.js – service worker for caching and offline functionality
• public/* – PWA manifest, global stylesheet, favicons, etc.
• experiments/* – exploration of initial concept in Python (with both Z3, and later, SCIP), and Dockerfile for compiling SCIP with Emscripten
• src/compiled/* – SCIP compiled to WASM via Emscripten, plus associated support files

Application Architecture

In the diagram below, labeled arrows represent asynchronous fetch requests and message passing.

flowchart LR
    A[Svelte Front End] -->|Solve Requests| B["Web Worker
    (non-blocking thread)"]
    B -->|Solutions| A
    A <-->|Static Assets| C[Service Worker]
    B <-->|Solver Static Assets| C
    C -->|Cache| C
    C <-->|Static Assets| D[Web Server]

Project Status & Contributing

Poker Chipper is actively maintained. If there are no recent commits, everything is running smoothly! As of the initial release, the code is stable, and there are no major, outstanding features that remain to be added.

Bug reports and feature requests via GitHub Issues are encouraged. Pull requests with more than 20 lines of code are unlikely to be merged quickly, unless associated with prior discussion or accompanied by substantial, explanatory English prose. In other words, pull requests containing code without context may be merged after much delay, or may not be merged at all.

Since Poker Chipper is a fully static web application with no server-side processing (outside of serving unchanging files), it is extremely scalable, and has a very low maintenance burden. As such, even if something were to happen to me, and I could not continue to work on the project, the public version should continue to remain functional and available online as long as my GitHub account is open and the jstrieb.github.io domain is active.

Support the Project

The best ways to support the project are to:

• Share the project on sites like Twitter, Reddit, and Hacker News
• Report any bugs, glitches, errors, or shortcomings that you find
• Star the repository and follow me on GitHub
• Host a version of the code translated into another language

If you insist on spending money to show your support, please do so in a way that is widely beneficial. In particular, donations to the following organizations help me, in addition to the general, Internet-using public:

• Electronic Frontier Foundation
• The Internet Archive
• Signal Foundation
• Mozilla

Acknowledgments & Greetz

Poker Chipper would not have been possible without help and feedback from:

• Logan Snow
• Amy Liu
• Chris Cherian
• Ani Chowdhury
• Will Hooper