If you’ve read much of our amiibo content on-site, you’ve probably heard the word “hard-coded” a lot. It’s a really important word to know, too. The short definition of hard-coded is “a behavior you can’t change”. We’ll go into a bit more detail in just a moment, but for now, know that if we say a behavior is hard-coded, it means that you can’t fix it with any amount of training. There are hard-coded behaviors in both Super Smash Bros. 4 and Super Smash Bros. Ultimate, but for this post, we’re going to be talking about the relevant game (Ultimate, though some might argue otherwise).
It’s no secret that Figure Players as a whole are kind of disappointing, especially to new trainers. As the premier amiibo training website (or something of the sort, anyway), it’s our job to disappoint new players by breaking the bad news: FPs can’t learn combos, they can’t recognize super armor, and they can’t tell if team attack is on! They also only “see” the startup of their opponents’ attacks, so lingering hitboxes such as Incineroar’s Alolan Whip prove extremely effective against Ultimate’s AI.
So, if you can train amiibo, why can’t you train these bad habits out of them? Well, that’s because Figure Players use Ultimate’s CPUs as a “base AI” of sorts. That base AI is then modified by the training data saved to the amiibo. And said base AI is what suffers from those problems we just listed — being unable to learn combos, unable to recognize super armor, and so on. We can’t train the base AI, so the only way to “fix” these problems would be if Ultimate’s developers patched out these flaws themselves. It’s unfortunate, and it definitely removes a bit of freedom from amiibo training, but we have to work with what we can get.
Here are some other hard-coded behaviors that can’t be changed. For one, amiibo can’t tell if their opponent is within kill range, so they won’t “save” kill moves for the right moment. This means you can’t teach an amiibo to use tilts when their enemy is at low percentages and smash attacks when they’re at high percentages. You have to teach the FP to use tilts and smash attacks often no matter what the situation is, which is kind of a bummer. Another issue that we talk about a lot here is that FPs don’t react well to multi-hit moves. After they shield the first hit, they think the attack is over and drop their guard… and then get hit by the rest of the attack. There’s no way to change this, either. Hard-coded.
Then there’s character-specific hard-coded behaviors, which… we could talk about those for a long time. One infamous example is Mario’s down air. The AI is known to spam this move, and that may be because it’s hard-coded to use an up throw to down air combo at low percentages. When the AI successfully connects that down air, it becomes more likely to use a down air on its own. So since that combo is hard-coded and can’t be changed, a Mario FP is seemingly doomed to use down air no matter what you do. In this case, all you can do is teach it to never grab, which is a shame!
You may think hard-coded behaviors are a bad thing, then, but that’s not necessarily the case. Sure, there are a lot of examples of hard-coded behavior gone wrong, but Joker is a perfect example of hard-coded behavior gone right. His AI is known for one of the most complex combo games in amiibo training history, and it’s very easy to teach him to use these combos! That wouldn’t be possible if it weren’t for all of his hard-coded combo routes. At the same time, though, so many behaviors being hard-coded makes amiibo training tough. If you’re new to amiibo training and don’t know anything about the AI, you’re probably going to be confused by some characters’ behaviors. Luckily, that’s why we write guides, and if you’re figuratively itching to train a new character, we have full how-to posts available on every single one!
If you would like to read more informational posts, please follow this link.