Rise of the Half Moon Clone

This is a clone of Google's Doodle Rise of the Half Moon, which is a lunar-themed board game. I found this game quite fun so I decided to recreate it.

It is, in fact, not just one Doodle, but a monthly series of Doodles. Every month when the lunar turns into the "Final Quarter" phase, Google publishes a new version of the game. The core game logic, however, is the same in each of these versions. As of writing this, Google has published four versions: October 24, November 21, December 22 of 2024, and January 23 of 2025.

Play the Game in Your Browser!

• Mobile screens are supported.
• Don't know this game yet? No worry, there is a tutorial in-game!

Features

• Core game logic (You don't have to read this! There is a tutorial in-game!)
  • Game board consists of slots and edges that connect the slots.
  • You and the computer take turn placing cards.
  • Each card has one of the eight lunar phases on it.
  • Before each turn you draw one random card so that you have three cards to choose from. You can place the card on any empty slot.
  • The goal is to score by forming patterns when you place the cards:
    • Phase Pair: adjacent cards of the same phase. Worth 1 point.
    • Full Moon Pair: adjacent cards of opposite phases. Worth 2 points.
    • Lunar Cycle: 3 or more consecutive cards that follow the lunar cycle. Point is equal to the number of cards in the cycle.
    • Every card involved in a pattern will be claimed by the player that makes the pattern. The opposite player can claim a card back by making another match using that card.
  • The game ends when the board is completely filled with cards.
  • Each claimed card gives 1 point to the corresponding owner. Unclaimed cards are ignored.
  • The player with a higher score wins!
• AI opponent who plays the game quite well
  • It works by simulating the game for a few turns, trying out different moves and analyzing which is the best. Higher simulation depth means higher difficulty.
  • The difficulty setting is hidden in "Custom Game".
    • Easy: 50% of the time the AI plays greedily; 50% of the time it plays completely at random.
    • Medium: The AI always plays greedily. That means it uses whichever move that gets it the highest score and claims it the most number of cards at this step, without considering next step. I'd say this is about the hardness of the AI in Google's original game.
    • Hard: The AI considers what you are going to do after its move.
    • Harder: The AI considers what it's going to do after you make your move. It also takes into consider the randomness of drawing a card, and what cards will be left after this move. At this level the AI is able to win most of the games against me.
    • If the AI is run natively (instead of in browser) it can be much faster and hence higher difficulty is possible.
  • In a normal (non-custom) game, the AI starts off at the Easy level. It becomes stronger every 3 levels.
  • It does not cheat! The card it draws is completely random, and it does not know what cards you have in your hand, just like how you get random cards and cannot see the computer's cards. If it keeps winning, then it's just smart!
  • It can handle game boards with any shapes!
  • It always assumes that you will make the best move (well, the "best" in the AI's opinion) and you won't play any wildcard.
  • It will consider the effect of wildcards that you've already played.
• Web page UI
  • I spent a long time on animations/transitions!
• All 12 game boards from the October game
• All 16 wildcards from October, November, December and January versions of the game (you get one random wildcard every 2-3 levels):
  • October
    • Hunter Moon: destroy all cards controlled by the computer on the board.
    • Super Moon: makes your Full Moon Pairs worth double for the current level.
    • Scorpio: the computer gets no bonus points at the end of the current level.
    • Leonids Meteor Shower: destroys 2 random cards on the board.
  • November
    • Winter Solstice: your claimed cards can not be stolen for the current level. (My favorite wildcard :)
    • Beaver Moon: flips a card on the board horizontally, or change New Moons into Full Moons and vice versa. If that helps you form patterns, you get points and claim the cards as normal.
    • Sagittarius: the next match you make will be worth triple points. If you make multiple matches in one move, all get tripled.
    • Geminid Meteor Shower: lets you claim half of the computer's claimed cards on the board.
  • December
    • Long Night Moon: the computer gets no bonus points at the end of the current level. (Same as Scorpio from October)
    • Quadrantids Meteor Shower: randomly destroys half the cards on the board.
    • Light of Mars: makes your Lunar Cycles worth +2 points for the current level.
    • Capricorn: lets you place your card on top of another card on the board for your current turn.
  • January:
    • Aquarius: get 10 extra points at the end of current level.
    • Light of Venus: get 1 extra point for each card you've claimed at the end of current level.
    • Moon at Apogee: makes all of the computer's matches worth 1 point for the current level. (A beautiful card!)
    • Wolf Moon: choose 1 of the cards claimed by the computer to destroy that card and all cards that are directly connected to it and claimed by the computer.

Technical Details

The backend of the project is written in C99. It is also written in such a way that it is compatible with C++11. The backend is compiled into WebAssembly using Emscripten so that it can run in a browser, and the frontend is written in JavaScript/HTML/CSS. The backend was not written in JavaScript because an AI would require a lot of computations and I want the speed of C/WebAssembly.

The most technical part of this game is the AI part. It follows the idea of the expectiminimax algorithm. The alpha-beta pruning optimization cannot be directly applied since our game contains "chance nodes" so I applied it partially to only the min/max nodes. Taking existing cards in hand into consideration significantly increased computation time but luckily I figured out a way to cache some of the results that don't depend on the cards in hand to improve the performance. At last it turned out that the pruning optimization was easy and cheap but brings just a little improvement; while the caching, though sophisticated, helped a lot.

Another tough spot is the lunar cycle detection algorithm. Graph searching algorithm is needed. The logic may seem simple at first, but there are many cases that need to be taken care of, especially when multiple cycles are triggered by one card.

Additionally, the build script is written in Python. A little preprocessing was also done in the build script in order to connect the frontend and the backend.

Building from Source

Dependencies:

• Python 3.6 or above
• Node.js
• Emscripten

Follow these steps:

1. Make sure the tools provided by these dependencies are accessible directly in command line. (i.e. add the correct thing to PATH environment variable)
2. cd into src/frontend and run npm install. This requires network connection.
3. cd back to the project root and execute build.py (python build.py or python3 build.py depending on your OS)
4. The game is ready under dist/ directory. Run it using python -m http.server -d dist and play the game in your browser at http://localhost:8000/.

Originality

The game design credit goes to Google. However, all the code and assets in this repo were created by me. You may find, for example, the card textures in my clone quite similar to Google's; but if you look closely, they are quite different as my cards were flexible SVGs while Google's were PNGs. Most wildcard textures indeed have the same design as Google's, but I drew everything from scratch and it took a long time.

Licensing

The code (.c, .inc, .h, .py, .js, .html, .css), config files (.json, .gitignore) and documentation (.md) in this repo are licensed under Apache License version 2.0 (see LICENSE-Apache-2.0).

The assets (.png, .svg, .ico) are licensed under Creative Commons Attribution 4.0 International (see LICENSE-CC-BY-4.0).

However, since Google invented this game and some of the images have similar design as Google's, it's not guaranteed that you can use this project for any purpose.