👀 Screen-Squint Syndrome! 👀

Yikes! It seems you're trying to launch this digital masterpiece on a device so tiny, ants are pitching it as prime real estate. That gadget's squeezing pixels tighter than jeans after Thanksgiving dinner. Here's a hot tip: trade in that digital breadcrumb for something with a bit more real estate. 📱➡️🖥️ Let's upscale from tickler to blockbuster, shall we? Bigger is better... at least for epic gaming and app adventures!

Connect-4 is the intense battlefield where discs fall like rain and IQs scatter like confetti - especially when Silly thinks diagonals are just tilted rows. Meanwhile, Smarty plots six moves ahead, rolling its circuits in despair as Prophet refuses to play unless someone compliments her pattern.

Seriously? You don't know this game?


Agents explained:

Human: That's you! Or your mate. Takes input from the keyboard (1..7 for column, ignores bad moves.).

Random: Take completely random turn. Not much fun to play against this one. But it is was very useful for testing.

Silly: 3-step lookahead algorithm. No challenge to beat this one. But it beats Random player all the time. What a surprise.

Smarty: 5-step lookahead algorithm. This feels like a real human (if slightly drunk) opponent. It beats Silly in 60% of cases, the rest were mostly ties.

Prophet: 7-step lookahead algorithm. It's the maximum javascript can take before blocking the thread. It takes 0.5-1.5 seconds to calculate the move in the early stage of the match. It beats Smarty and Silly in 100% of cases. She is a nasty player.



Mode:

Game analysis is found in the console.

Game will run as fast as possible, without any animation. Using analysis mode with human players is silly.

Red:
Blue:
Engine versions: Prototype LIB:
ENGINE:
GRID:

About:

The game was written as a break from working on 'Haunting the Hauntessa', using my Python submission to Kaggle competition as the source. One day, I will find the time to implement also reinforced learning agent ... hopefully.


LS-Web v1.8.1