Modeling Enmity

[kanican]

Overview

I don't claim to know how everything in this game works.  There are so many complex equations dealing with just about every part of how this game functions that it's really overwhelming to just think about it.  There are much better, more "famous" players to look to for equations like Aurik or VZX.  I do, however, believe that I can tell

Comments

[anonymous]

i work by a slightly different theory for a few reasons:

1. the SE employee who works on their calculations isn't a brilliant mathematician or is limited by the hard/software.
-this is evident from the lack of foresight in the role of lvDIF in various calculations (leading to the lv cap being stuck at 75), also leading to the caps on haste/subtle blow/etc. that is equippable because it is calculated as fractions.

2. calculus takes a little more processing power than just doing -+*/
-knowing that its harder to do calculus and less efficient the most likely function would be to tier the enmity (over 100, over 150, over 200) and then set the reduction from that (at tierX losing 1 shadow = -30enm, 3seconds passing = -10enm) however this is more difficult than the method i work with.

3. more enmity requires more volatile memory space.
because having a decay curve as you described leads to a cap on the enmity you can gain which is a function of abilities available/spells+haste/etc. you end up with an absolute maximum anyway, this would be when hate loss over time is greater than or equal to any possible hate gain function. so why not just cap it at a definite value which takes up minimal number of bytes, and dont even bother tiering loss rate.

therefore i work on the basis of a specific maximum value, when you reach that cap you immediately become top on hate list (because anyone else has just decayed below cap until they perform another action). however as you noted, its easy to lose hate very quickly - avoiding dying in the process is the only challenge.

its actually the same thing in terms of effect, but the way i calculate it is easier.

i can draw some nice mspaint graph showing how hate loss rate would vary with hate level according to you theory and lead to an assymptote if you want~

mikazuki (ciz)

[kanican]

I think as someone mentioned earlier, you don't really need calculus to do this, I just happened to write it in that form (was pretty much the simpliest type you could do too).

I think the key here is I'm not trying to decipher how SE actually calculates hate and enmity - I'm only trying to find a suitable model that can account for 99% of all cases you would see in game. I think the best metaphor for this I can give is how people still learn Newton's Laws in Physics even though they are technically incorrect (they still work 99.999% of the time). Whether or not SE actually does enmity a certain way really doesn't matter to me; I just look towards a model which can grasp how it works and is accurate enough to let me predict cases.

The biggest piece of evidence I can give for there being a lack of a formal Hate Cap is how in extreme cases, I have seen a tanking combo of NIN/DRK and RDM/NIN side by side, and the RDM/NIN was unable to take hate at all unless the NIN was actually hit with a big physical attack. According the idea of a formal hate cap, this should not be possible.

There's always going to be lots of debate on this, so I enjoy talking about any altnerative theories or observations which de-bunk what I've written. Like I wrote in the post, it's much easier to disprove something than to prove it.