4 replies to this topic

[Fire Blazer]

So for my final project in my programming class ENGR 131 (with MATLAB) I somehow managed to propose an average stat calculator a la Fire Emblem (though without emphasizing that part ).



It was actually pretty hard, but anyhow, it lets you calculate average or randomized stats and you can input the names of the stats, the base values, and the growth rates. You can also set the base level and the final level and it'll plot it to the right (I'm contemplating asking my partner to make it so that you can plot two of them simultaneously since I thought about it but it seemed like way too much of a pain and I pretty much finished most of the program already... XP).

Now, at the bottom of that screenshot are Eliwood's level 20 growths. If you round them, you'll get the same ones I got (it's a bit impractical to include the non-rounded numbers, IMO).

However, upon repeatedly clicking the "Calculate Randomized Stats", you can get a lot of cool things--pretty much, infinite potential in terms of how good, bad, average, or something in between those Eliwood can be.

...

What shocked me though

and I didn't even spam it a hundred times or anything--I only noticed this because I thought it was a glitch ._.

was this:



Oh my crap Eliwood.

That's hardly one stat per level-up. He literally got +2/20--10%--on EVERY stat except for HP. Which he got 10/20, which is 50%, when he has an 80% growth.

...His growths add up to 325%

The growths off that (hypothetical) run

were more like 110%--almost 3x as bad.

60% vs. 245%--off by a factor of 4--excluding HP

O_O

only an FE fan would ever react like this to something like this

[Snow]

Post this online or something. The Randomized calculator looks awesome BTW, I haven't seen a FE calculator with something at least.

[Bobryk]

This is...actually really cool. Awesome work! Looks like fun too.

holy crap Eliwood is so bad

[kirant]

QUOTE (Snow @ Apr 13 2013, 04:16 AM)

Post this online or something. The Randomized calculator looks awesome BTW, I haven't seen a FE calculator with something at least.

A quick observation: This program requires MATLAB, a professional program. If you want to run it, you'd need either a really expensive license or piracy (and I really can't say I want to you pirate).

I'd post the MATLAB code online though. This is really good stuff.

Did you use statistical modelling to calculate the means of all situations (Such that the average of a stat that rises quickly will slow down as it approaches maximum)? Or did you simply just add the value of each stat as a decimal form of its percentage, up to the maximum? The latter is my quick and dirty method, so if you used the former, I'd love to get my hands it so I can do more intensive modelling.

For example, if we look at how FEwiki's average page for Lilina, we can see they rounded a mean of all the possible results. Typically, the Gaussian shape is retained, meaning the mean increases at the rate at which all events approximately increase (though the probability of each even dwindles slightly since you now have more possible events). However, when the stat begins to cap off, the right hand side of the Gaussian curve begins to "bunch up" at the cap, as it were, meaning the mean is no longer at the centre of the Gaussian curve, but is approaching the cap more slowly.

In simpler terms, let's think of a character at level 20/15 with 28 in a stat which caps at 30. The growth is 50%.

At level 20/16, there are two event's, each equally possible. 28 or 29. So the mean is 28.5.

At level 20/17, there are four possible events, but two are the same. 28, 29, 29, and 30. Each has equal probability again, so the mean is 29.

At this point though, the eight possible events are slightly off. We have 1 incidence of 28, 3 of 29, and 4 of 30. The standard event should have been 28 once, 29 thrice, 30 thrice, and 31 once. As such, our mean, while the simple approach says 29.5, is actually 29.375. That isn't much here, but for characters who come really close to maxing certain stats, that'd be a great metric to show.

I guess, simple question is, what approach does the code take?

(Lilina's stats, as referenced)
http://fea.fewiki.ne...r=lilina&game=6

EDIT - Would it also be possible to make a subplot edit to make trackers for each stat? Sum of total is nice and all, but it doesn't help out a whole lot since some stats aren't as useful as others (ex - Luck vs Speed). As I'm sure you've learned, univariate statistics to describe multiple moving parts grossly oversimplifies things.

[Fire Blazer]

@Snow sadly I could post the source code but it wouldn't do people much good, I can't really compile it into an application or anything without different software that I don't have, to my knowledge XP

And the randomized calculator is indeed fun though it was rather hard to code the graphing part XD

@Bobryk thanks and I know right?

@kirant oh gosh such a long post XD

The code is actually rather simple--it's for my final project in a very elementary programming class (I mean, we're using MATLAB here) and I pretty much did the minimum/most basic stuff here. To start, if I understand right, I did the latter method, adding the stat as a percentage form and whatnot. Meaning, I think you're out of luck here, sorry to disappoint XP

I could definitely split-up the stats but because I didn't want to emphasize the fact that this was for Fire Emblem, I didn't want to make it overly complicated. First of all, I'm working on it with a partner, and I don't want to scare him by making it too complicated, and second of all, the project requirements don't seem that tough, so I didn't want to overexert myself. Tbh, this only took like... 3-4 hours, IIRC, and that was with me getting stuck in a couple places--if I were more experienced with MATLAB, I could have probably done it even quicker, and more efficiently.

I mean, the project requirements were like

"Have a GUI"
"use a loop"
"use a cell or struct array"
"use a function" (besides a callback function for buttons and the such)

I have a relatively complicated GUI (compared to what we've seen in/done for class), have multiple loops, plenty of arrays, and multiple functions, so I had already hit most of those, so yeah, you get my drift

The point is I didn't want to but it could definitely benefit by splitting up stats, putting the average/randomized stats on one plot, comparing one character's data to another (e.g. using interpolation to find the slope of a line and comparing it to another set of data's slope), perhaps even weighting stats accordingly, I'm sure it could be 20x more complex and we could make the ultimate simulator, but this just simulates level-ups--it doesn't even include caps, meaning stats can technically keep increasing forever, which is one of the reasons I didn't use Eliwood's 20/20 promoted stats (that and the fact that it doesn't have the least bit support for promotion bonuses ).

Regardless, my incomplete and not that great code is here if you want to take a peek. It's incomplete because I had to save my partner some work (I pretty much did most of the project before our proposal even got approved--the alpha isn't even do for a while, so we're WAY ahead) but I dunno, if you find this stuff interesting, have fun XD. It might be slightly buggy though and to simplify things it always rounds stats, which isn't super exciting (though most of it really isn't, even if it was fun to make XD), and the input/export data files does nothing yet, I set it up a little for my partner but he has to figure out how to make it actually work lol.