Guess my RGB

Original Article ↗ Hacker News Discussion ↗

A simple web game that asks players to guess a hidden RGB hex color has turned into both a casual challenge and a mini laboratory for color perception and search strategies. Commenters swap scores and tactics, reverse‑engineer the scoring function, and even script automated solvers, while others point out how the game highlights the limitations of RGB versus perceptual color spaces like HSL, Lab, or OKLCH. Alongside bug reports and feature wishes (expert modes, social sharing, camera input), people link similar tools and games, suggesting this kind of color-guessing puzzle is useful for teaching color theory as well as just being unexpectedly fun.

Overall reception

  • Many commenters found the RGB guessing game unexpectedly fun, addictive, and nostalgic (e.g., recalling hand-picking hex codes in early web dev).
  • People shared scores and number of guesses, often getting “Splendid!” in 4–20 tries, with some one-guess perfect matches (sometimes via dev tools).
  • Several described it as a good casual/party-style game and something they’d play with friends or students.

Gameplay & strategies

  • Common human strategy: adjust one channel at a time, watching score changes, effectively hill-climbing in RGB space.
  • Some users treat it as a puzzle: first get a rough hue (e.g., ROYGBIV), then refine by channels.
  • A few compared it to ear-training / equalizer tuning: iteratively “notching out” error.

Scoring function & math exploration

  • The scoring code was posted and analyzed: Euclidean distance in 4-bit RGB, normalized by maximum possible error per target color, then scaled to 0–100%.
  • Several users argued local optima (stuck at 94%) should not exist mathematically; apparent stalls were blamed on rounding or input imprecision.
  • Others explored optimal search strategies:
    • Per-channel hill-climbing, binary / ternary search along each axis.
    • Exploiting the numeric score to triangulate the true color in 3–5 guesses on average.
    • Brute-force strategies using precomputed buckets and variance minimization to choose the best next guess.
  • It remains unclear whether a perfect 3-guess strategy exists; some argue 4 guesses are needed in the worst case.

Bugs & UI issues

  • Reported bug: “New Game” sometimes didn’t change the color and subsequent guesses showed NaN%; this was later claimed fixed.
  • Another issue: some colors (e.g., #952 vs. #951) appeared visually identical while giving non-100% scores; unclear whether due to rendering, perception limits, or a subtle logic issue.
  • Mobile users noted fat-fingering sliders and difficulty moving by exactly one step.

Design ideas & feature requests

  • Requests included:
    • Expert mode with 0–255 channels and alpha.
    • Variants for HSL, OKLab, OKLCH.
    • A forfeit/reveal option.
    • Better tick visibility for sliders.
    • Wordle/GeoGuessr-style social sharing and competitive modes.
    • Camera-based color selection.
  • Several linked similar games/tools and teaching uses (code clubs, recruitment tasks, mobile color puzzles).