Jap... selbst bei 100% Placebo wäre es für mich ein Gewinn.Ingo hat geschrieben: 14. Mai 2026, 12:52Ich finde auch das sie da einen tollen Placebo geschaffen haben, wieviel von ihrem Marketing letzten Endes stimmt, sei mal dahin gestellt aber es funktioniert für mich einfach.Guthwulf hat geschrieben: 14. Mai 2026, 10:06Ich mag Drivatare, weil sie die Gegner (egal ob als Rivals oder KI bei Rennen) personalisieren und man nicht gegen Unbekannte sondern auch gegen bekannte Namen aus der Freundesliste fährt und weil die "realistisch" fahren, da auf Basis des jeweiligen Spielers trainiert (was z.B. Agressivität, Fehleranfälligkeit, Kurvenverhalten angeht - auch wenn da viel Placebo dabei ist). Steigert nebenher auch Erfolgserlebniss und Motivation
Die Zauberworte sind Neural Networks und Machine Learning
Lange bevor die Masse es durch ChatGPT kennengelernt hat. 
Gibt hier und da Infos dazu, aber natürlich alles mit Vorsicht zu betrachten, z.B.:
How Forza's Drivatar Actually Works | AI and Games #60 - https://youtu.be/JeYP9eyIl4E?si=tOHHvKuFwlVWCVTk
Oder Forza Wiki: https://forza.fandom.com/wiki/Drivatar
A Drivatar is an AI-controlled opponent designed to mimic the real-world driving behaviors of human players.
SynopsisHow Drivatars Work
- Realistic driving: Drivatars aim to replicate individual driving styles, leading to noticeable differences in how opponents approach corners, brake, and navigate the track.
- Dynamic behavior: They can exhibit aggressive driving, such as blocking or ramming, based on the habits of the players they've learned from.
- Evolving challenge: As you and other players improve, the Drivatars can adapt, providing a constantly evolving challenge for racers of all skill levels.
Trivia
- Learning from players: The Drivatar system takes data from your driving and that of other players on the Xbox network.
- AI Training: This data is used to train neural networks that form the basis of a Drivatar's behavior.
- Cloud-based profiles: These trained Drivatars are small, cloud-hosted files that can be downloaded by other players.
- Personalized opponents: When you race, you're often competing against these Drivatar profiles, which can be based on your friends or other players you don't know.
- You get paid while you're away: Because your Drivatar can appear as an opponent in other players' races, it can earn you in-game credits even when you're not playing. A message will appear in your Message Center informing you of your Drivatar's earnings and results.
- A history of over-aggression: Early versions of the Drivatar system were trained too effectively, learning the bad habits of real players. This led to overly aggressive, and in some cases, "dirty" AI that would deliberately ram other cars, causing developers to scale back the AI's aggressiveness
- The Rewind feature doesn't affect your Drivatar: To prevent players from "cheating" the system and training a perfect Drivatar, using the Rewind feature does not add to your Drivatar's training data. Only successful, clean laps count toward improving your Drivatar's skill
- Drivatar "rubber banding" is an illusion: While the game does use a form of "rubber banding" (opponents being faster or slower to keep races close), it is applied to the cars, not the Drivatars themselves. The game subtly alters the physics of the AI cars—changing their weight, torque, or tire friction—to adjust their performance. This keeps races competitive while preserving the illusion of distinct, player-like driving styles
- They remember your preferences: When racing, a Drivatar can replicate the exact car, tuning, and custom livery that its creator used on a particular track. If the conditions are different, it will use an approximate version of that player's typical driving style.
