Lies, Damn Lies...

[Latrinsorm]

...statistics.

I'm doing some redux research lately on the relationship between redux points and redux factor. Initial findings: there is one.

Also, I had Excel do a logarithmic regression on the data and got a pretty good approximation: most points were within a standard deviation of the line and so on. Excel does logarithmic as A*ln(x) + B. Fine. When I was initially trying to figure out what the regression was, I did 10 ^ (redux factor) because I thought it was going to be logarithmic and sure enough it was roughly a straight line.

However, as I've gotten more data points the 10^ section is matched much better with a polynomial equation of order 2. I'm thinking this means that the actual equation is something like A*ln(x^2 + O(x)) + B, but I'm not sure how to make a regression of that (manually). When it's just x, any coefficient can be absorbed into the B no problem. It can't just be an x^2 term because then the 2 would get absorbed into the A. Do I assume that the two terms have no coefficient?

I have Matlab, but I can't get a license for it again until the fall semester starts, and I don't feel like waiting. If the above problem is manually insoluable, does anyone know of a readily (legally) available program I could use to get the equation I want?

[radamanthys]

legal program? what's that? I don't get it.

[Latrinsorm]

If Faye would describe it as being "procured", I'd just as soon have no part in it.

[Stanley Burrell]

Roll a d100 and just say that there doth be hidden, non-linear roll(s) and call it a day.

You're solving for X like there's no Y, mang :-\

[Latrinsorm]

I figured it out. I was making this way more complicated than it had to be. I'll just use the polynomial regression the spreadsheet spits out and factor out the base.

[Stanley Burrell]

Wouldn't it make sense to ask on the officials if a higher-up can post formulae regarding such matters?

EDIT - That is to say if they will indulge you with such info.

[Back]

Lies lies lies yeah - Thompson Twins

[Kranar]

If you have a crapload of points, by all means, post them.

Good approximation formulas can be achieved using least-squares methods. The points are key though.

Also, there's http://scilabsoft.inria.fr/, which is a free, legal, open-source implementation of Matlab.

[Latrinsorm]

Wouldn't it make sense to ask on the officials if a higher-up can post formulae regarding such matters?Warden said he's not going to make the formula available, but I'm getting a pretty good handle on it.

Yes, I have a crapload of points, but they're all on my laptop. Tomorrow I'll put the most recent version of the spreadsheet on my laptop and post the points up here. Like I said, I'm pretty sure I was just making the problem more complicated than it actually was. I'm very confident I can predict the redux factor of any number of redux points on a level 66 character. I just need to muck around with a high level and lower level char to figure out quantitatively how the level progression works and then do some work with spells.

If anyone has a level 75+ character or a level 30-40 character with a bunch of redux and a fixskills to burn, let me know please! :D

[Bobmuhthol]

I don't have a fixskills to burn :( but I have a level 32 character with a bunch of redux!

[Latrinsorm]

Ok so the excel file is 11MB now, so I don't think I'm going to attach it directly. I started at 110 redux points and went up in 8 point increments, getting the following values for redux factor:

0
.039
.108
.158
.183
.232
.257
.292
.315
.332
.355
.373
.384
.395
.418
.423
.435
.447
.456
.473
.482
.482
.489
.501
.511
.510
.515
.529
.525
.538
.541
.541


.546


.548

with standard deviations ranging from .012 to .020. The gaps are meant to indicate how I did increments of 24 instead of 8 on those two spots.

Now, what I was talking about before about the e^(redux factor) thing is still true; it's mapped very closely by a 2nd order polynomial function. However, as you might guess, this is a 2nd order polynomial function that has a maximum, after which redux factor would go down. That doesn't make a whole lot of sense, so my theory at this time is that there's a fixed maximum redux factor (possibly one for each level) that's approached by a function that looks like ln(Ax^2 + Bx + C). The function is designed so that its apex is the maximum redux factor. I'm going to finish up my subject's training plan later tonight and see if there's any significant change in redux.

[Drew2]

You basically have no desire to lead a normal life, I take it.

[Ignot]

You should calculate the probability of you getting laid.

[Stanley Burrell]

The real freaky bitches like to apply some of that standard deviation from all 30, 60 and 90 degree angles, lemme tell YOU.

[Drew2]

You should calculate the probability of you getting laid.
ROFL.

This guys sees my point.