Figuring out how the RNG works in this game

[ruadath]

I've mentioned this before but the RNG in this works in an extremely strange manner. I haven't seen this occur in any other FEs (let me know if I'm wrong), but the game uses the RNG differently for short-range attacks vs. long range attacks... most of the time. And yes, I've tested this many many times in different scenarios to ensure this isn't just a result of me neglecting to account for skill procs, supports, or something silly like that. The most glaring evidence for this is a test I did with the a tome-user (with 1~2 range) who attacked from 3 different squares and got different results.

I don't really know too much about exactly what is going on here (hence the reason for creating this topic), but I do know that close-range attacks are always "stable;" they work in the normal way that we expect them to (the first 2 RNs are averaged for the hit, and the next for the crit, and then skill procs or whatever). The long range attacks, however, do not seem consistent at all; in some rare cases (which seem fixed for particular tiles on a map, although I have not found any conclusive patterns), it works just like a short-range attack.

The majority of time however, it seems to use some alternative mechanism, or perhaps a combination of multiple alternative mechanisms (more details on this to follow). I have not done any testing to figure out if these same mechanisms apply for longbows or siege tomes (and probably won't touch that until we resolve this simpler issue first).

I have (in fact, I've had) some proposed tools/methods to do testing on this and maybe gather information, sort of like with the Enemy AI Targeting thread I made here a while ago. Just like in that case, however, I've been too lazy to actually put effort into this and get a detailed understanding of how this system works, and I'm hoping this thread might inspire someone else (or maybe me) to actually put effort into this and figure out what is going on.

Feel free to post here with any questions/comments related to the above.

[VincentASM]

I'm not really sure, but might it be possible to use Dolphin for debugging?

The different behaviour at short and long range interests me though. If it is true, it makes me wonder why they implemented it. To nerf long-ranged attacks somewhat?

[ruadath]

Yeah, Dolphin debugging is how I figured this out in the first place.

The different behaviour at short and long range interests me though. If it is true, it makes me wonder why they implemented it. To nerf long-ranged attacks somewhat?

It's not a matter of nerfing, since (as far as I'm aware) the statistical distributions of hits/crits are not any different for the two. It's just that the game does something different with the RNs it is given to get the numbers for the long range attack. For example a seed that yields a 3% crit on any short range attack may not yield a crit on a long range attack.

What is really bizarre about this, however, is the fact that it's not consistent. Sometimes, a long range attack will function exactly like a short range attack. Most of the time, it will not, and there usually isn't any clear reason for the difference.

[ruadath]

Ok some additional testing revealed the following (even more bizzare results):

Having Leonardo attack a unit from 2 different positions (SW and 2S) an leveling up off the same RNG seed resulted in 2 different levels. This seems to rule out the possibility of different interpretations of the same RNG seed, and instead suggests that we have either two (or more) separate RNG seeds (unlikely) or some weird method for determining the number of RNs to be burned... However, the results from Leonardo's levels appear to be somewhat inconsistent with this theory, and furthermore, using Wind Edge!Edward in his place also produces inconsistent results... not sure how to interpret this.

I will look more into this and report back if I get any more interesting results. If anyone is interested in helping me figure this stuff out, let me know, because it's been bugging me for a while and I have no idea how to approach this problem.

[Gradivus.]

I'm curious, what would that knowledge be handy for? You can already RNG abuse this game as far as humanly possible with the tools you have. It might not be worth your time

Wild guess on my part: the camera moves differently depending on where you attack from -> more frames are loaded -> RNG produces different results

[ruadath]

It's not really useful for anything, but it's just extremely weird and doesn't seem to have any reasonable explanation (and the mechanic doesn't show up in FE11, not sure about FE12 but I doubt it), although I'm sure that one does exist. I'm curious because I can't seem to figure out what might have inspired the game programmers to add this... "feature..." to the RNG, and the fact that it doesn't affect close-range attacks at all, but only sometimes effects long-range attacks (and in different ways) is really weird...

EDIT: I don't think it has anything to do with the camera stuff since this occurs on PP too, but I'll runs some tests regardless. The game doesn't seem to burn RNs in any time-related things anywhere else, so it would surprise me if this was the case.

Edited August 14, 2016 by ruadath