Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Nil Eyeglazed]

You're right, the "off by 3" thing to a nice round number is suspicious.  The problem is that if you start playing around with it, you see that values change rapidly at the high end of the scale-- 3800 should be faster than 3797, if it was available.  The 200 lower end cap doesn't matter though.

Using a coefficient of -0.0003 is not accurate enough.  If you solve for the constant at 200, you're 5.5 off at 3800.  In reality, the slope should be a little steeper than the fraction I got, I think.  I've been playing with slopes around -0.0003008, which should give you an idea of the kind of precision you need.

Currently playing around with values.  I especially can't explain that 3-large jump in advspeed from 3796 to 3797.  That has to mean that it's rounding sometime earlier in the alogorithm, and then figuring advspeed from those rounded values, which it then rounds again.    I guess no lives depend on the math though.  Double rounding could maybe allow a cap of 3800, too.

I've edited wiki with some updated info, especially the idea that strength affects equally with agility (which, yeah, has some interesting ramifications).  If you want to put a formula on some page, go for it; I was going to wait until I had something I was more confident about, although I guess there's always talk pages for nearly-correct information, which is sort of where the figuring is at now.

EDIT: Real world experiments (erm, fortress mode experiments) continue to provide surprises.  I found one goblin at sum=2267 moving between 10 and 11 while another moved consistently at 10.  Curious if I could isolate the problem, I set their attributes exactly equal to each other in every way.  Movement differences persisted.  Both shared the same flags.  Editing obvious traits did not appear to solve the difference either, but runesmith doesn't permit setting traits numerically to make the two goblins identical.  No idea what is going on, just that the model is not yet complete.

[dree12]

11 is very slow for 2267 sum according to the forumla! They should get 10 most of the time. Maybe the goblin is wounded?

[Nil Eyeglazed]

Sum 2267 is right smack dab in the middle of the range (by arena experiment) that gives delay 10 even.  So I've got one sum 2267 goblin who's either doing delay 10 even or very, very close, and another one who's doing ~10.5.  Neither goblin shows any wounds or scars.

I don't think I'm going to bother working on the formula further-- it should give accurate results plus or minus one, and if there are rounding artifacts, its hard to say where exactly in the algorithm the rounding is happening.  It would take testing a lot of different hypotheses and comparing a lot of data for very little payoff.

Probably in a couple of days I'll finally get around to making a wiki account (have just been signing with IP) and post the formula in some relevant talk page-- need an account to make a talk page.  Certainly, if you do any work on making a better formula, write it up!

I got some likely sums for integer delays, and I plan on using that to try and figure out some goblin-based clock generators, see if I can do anything useful with it.  There are a few other problems to sort out with this-- like why they fall down when a long path is suddenly blocked off but not when a short path is suddenly blocked off.  I'll be able to see if there are other goblins that don't match anticipated delay during these tests, and maybe I'll find the common feature.  In the meantime, the 10.5 delay goblin seems like an anomaly.  I'll post more if I figure something out.

(In case you're curious, likely integer delay sums are 421, 790, 1160, 1529, 1898.5(could go either way), 2267, 2637, 3006.5, 3376, 3745.  Could confirm some of those in arena, most provide non-integer adventure speeds.)

Thanks for your help, dree!  I don't know about you, but I benefit a lot from people doing parallel experiments-- important to compare data, verify and see what doesn't work, and other people help set me straight when I make a stupid mistake.