![]() |
|
AI INSANE + XOI - Printable Version +- Little Fighter Empire - Forums (https://lf-empire.de/forum) +-- Forum: Little Fighter 2 Zone (https://lf-empire.de/forum/forumdisplay.php?fid=7) +--- Forum: AI Scripting (https://lf-empire.de/forum/forumdisplay.php?fid=56) +--- Thread: AI INSANE + XOI (/showthread.php?tid=11730) |
AI INSANE + XOI - Xoi23122000 - 10-30-2026 The original Little Fighter 2 has four computer levels: Easy, Normal, Difficult and CRAZY!. Its computer player is a set of rules written for the original fighters, keyed by their ids. Put it in a mod (One Piece GrandWorld, Naruto: The Setting Dawn, KT-LF2, Sx2…) and it has no idea what the new moves are, so it mostly punches, kicks and throws the odd D+A at random. INSANE! is a fifth level, selected right after CRAZY! on the character selection screen (and in Tournament and Battle mode). It contains no line about Luffy or Naruto: everything it knows about a fighter comes from playing each of that fighter's moves itself, in a private copy of the match, before the match begins. That is why it can use the moves of mods nobody has ever written an AI for. I have developed an entirely new version of an AI capable of self-learning; it can recognize its own skills and utilize them appropriately. Furthermore, it can scan and anticipate the enemy's incoming moves within a 16-frame window, allowing it to adapt dynamically. It is not yet complete, however; I am currently expanding the AI’s awareness to include the opponent's capabilities, which will enable it to respond even more effectively. I am currently continuing to refine how the AI perceives information about its opponents, with the goal of making it increasingly sophisticated. Developers are welcome to access my open-source code to contribute further—this AI script is designed to be compatible with almost any mod (I have successfully tested it with about six different mods, and it works well). New AI model, which I’ve named "XOI." Its learning mechanism is similar to the SPSA (Simultaneous Perturbation Stochastic Approximation) method used in world-class chess AIs. It has proven highly effective; in tests conducted using the "KT" and "HELL" mods for the game *Little Fighter 2* (LF2), it achieved a 65.4% win rate against the "INSANE" difficulty AI (in the HELL mod) across 45,120 test matches, and a 54.9% win rate (in the KT mod) across 20,358 test matches.My new AI, XOI, utilizes a training mechanism similar to that of the chess engine Stockfish—specifically, SPSA (Simultaneous Perturbation Stochastic Approximation). This allows the AI to learn from past mistakes through countless simulated matches, refining its parameters for specific behaviors tailored to various scenarios. At present, its level of learning remains quite rudimentary, with significant room for further development. My AI project is designed to be applicable to all types of mods available, rather than being limited to a single mod or character. I am currently building upon the strengths of "AI SAUCE"—the most powerful existing AI for *Little Fighter 2* (LF2), though it is currently restricted to about 15 characters. I intend to have my AI spar against the SAUCE AI for approximately 500,000 matches, allowing it to learn from its mistakes and refine its performance using the Simultaneous Perturbation Stochastic Approximation (SPSA) method. Once it reaches its peak potential, I will apply these learned parameters across all available mods. All source code is available on my GitHub: https://github.com/anmen5000/OpenLF2 Visit the demo page here: https://anmen5000.github.io/OpenLF2 I developed this based on the open-source code from: https://github.com/OpenLF2/OpenLF2 RE: AI INSANE + XOI - Xoi23122000 - 11-07-2026 After integrating the core strengths from SAUCE's open-source AI, my AI build is now smarter than ever with melee characters—it actually managed to beat SAUCE! Right now, it only gets crushed by ranged characters like Henry, Bat, Firen, Firzen, and Freeze. I'm currently redesigning my AI based on SAUCE's strong points with ranged combat. Rudolf is a special case: I lost badly with an 81 - 189 score (a 30% win rate) because SAUCE's Rudolf is just designed too flawlessly, so I won't be developing this particular character any further. All in all, huge thanks to SAUCE for giving me this incredible inspiration!
|