Enmity testing - Still Going!

[nekoshoonen]

I'm just simply not fond of Forums run by Gillsellers that have had keyloggers in their pop-unders...

[Foustian]

The discussion there isn't all that different from the discussion here. Anyway, I started the thread because educating people is generally more effective than lol'ing at them. I would think any intelligent person could appreciate that (not directed towards any individual, just the lolzam mentality that sanctimoniously permeates these forums). I find that the proportion of quality posters is rather similar across sites, having frequented them both over the years. The real difference is in the cliqueness less so than substance.

How can you say that after watching them immediately derail the thread into "Third party apps are bad! BAD! BAAAAAD!" followed by disputing facts that are clearly established by easy to understand tests? :/

This would be as opposed to the mightily superior "someone is stealing your ideas!" posts?

Seriously, all these attempts to try to privatize data about a game that you have no f*cking rights to crack me up. It's one thing to directly plagiarize a post word-for-word, but if you seriously believe that game data posted on this forum can somehow be prohibited from being mentioned at another web site (including, but not limited to, Alla, KI, FFXIO, and various wikis), I have a sizable tract of oceanfront property in Nebraska that you may be interested in.

Wow, learn to fucking read. Your reply is 100% invalid to anything that was said in the blocks of text you quoted.

Not really. He was comparing the people here (albeit only like 1 or 2 of them) that brought up the wiki's pages on enmity modeling to the people on alla who derailed the thread into a discussion on plug-ins. He then went on to expand his thoughts about said people, but his original statement was still relevant.

[shirizaan]

Then perhaps in the future he should learn to quote material that actually demonstrates the things he's directly bitching about instead of some vague reference.

[rath]

I personally do not know anyone that has read these tests and still insists it is wrong. Most people I know that read it are very impressed with how thorough these tests are. I can only imagine those that do refute it just skim it and do not actually understand hypothesis testing.

[Elipse]

Hey Kaeko, (or anyone that can answer)

I'm a bit worried that people are jumping to conclusions here without enough testing to say one way or another on an issue. There seems to be a vast number of possibilities, but people are focusing on 1 or 2.

1) Has anyone checked or tested to see if the amount of hate generated from a cure is related to % of hp cured instead of level of the player? I haven't seen anyone say one way or another but has anyone grabbed say a pld with 1500 hp and cured him for 100 vs curing a blm with 700 hp for 100? Then going and checking the enmity levels? I haven't really seen any test at all other than someone curing a low level player (which has lower hp in general) and a high level rdm. Which most of the 75 rdm's testing this could in fact have very similar hp's. It could be tiered for all we know (1-5% = .6x, 6-10% = .65, 10-15% = .7275 etc, just throwing out numbers).

2) Has anyone checked it for damage as well? I haven't seen any tests other than the ones we ran and everyones jumping to the conclusion that its mob level. First test we came with with 13.3333 as the multiplier. Thats with doing like 20 damage to a 30hpish mob. This was done with hand to hand and no weapon equiped by a rdm.

Second test we came up with about 3.8x as the multiplier. Thats with doing something like 95 damage on a 1900 hp mob. This was done with a katana and by a ninja. I didn't run this test with nearly as tight controls as the first one. It was merely a quick test just to see if the 13.3333x multiplier was universal. Which quickly proved not to be the case then off I went to salvage.

Job changes, weapon type changes, mob level, mob level/player level difference (IE level 50 mob has 25 level difference therefore x modifier) but so does the % of damage dealt. Same with curing it could be tiers or possibly other factors we're missing. In otherwords, while mob level is likely, without any other tests I've seen run every other possibility is just as likely and shouldn't be ignored. I'd say % of damage done is just as likely as mob level in terms of figuring out the multiplier.

If there are more tests that have been done and I've simply overlooked them, can you guys point me in that direction? Otherwise I think we shouldn't just count everything out without more testing. I'm sorry if I sound like a jerk or something. I might have missed something simple that invalidates everything I said above, so go easy on me!


Taj: I think you were talking about how you can find the mobs level during the check process? Are you thinking about making a windower plug in that would allow players to check a mob, see its level and have that level shown on screen somewhere? This would likely help a great deal in the process because not only can we check for level differences between player/mob for hate, we can also find mobs with exact same hp levels to help with controlled damage testing/hp% tests. If you happen to be making such a plug in, or if you know of one that exists, then thank you very much in advance!

[Mashu]

Has someone tested what Foe Sirvente does to the VE decay rate-- which I'd imagine is what the decrease in enmity loss is referring to? Can't seem to find BRDs that have it merited :<

Awesome flippin' blog btw >.> Well appreciate the work

[Elipse]

I posted this on Kaeko's LJ and I'll add it here as well:

Okay I did some testing on this. Enmity does not cap at 50.

Testing set up:

Ninja using +45 enmity in gear
+4 in merits
+3 in food (dorado sushi)

Total +52 Enmity.

Test:

Player 1 pulls.

Player 2 casts Hojo: Ichi 3 times. Total hate value 240 CE.

Nin with +52 enmity casts Hojo: Ichi twice.
(80E X 1.52) x 2 = Theoretical enmity 242 (not counting decimal)

Player 1 zones out and stays out.

Now player 2 in theory only has to cure 1 for 0hp 2 times in order to pull hate. However, the bee doesn't turn until cure #4.

This implies two things:

First enmity does not cap at 50, but likely 100 in the sentinal example.

Second is that something in this calculation is incorrect.

-Either gear is mislabled (see thf AF2 pants for an example of how this could happen).

-Dorado sushi gives 4 enmity instead of 3 or perhaps even a % of your total enmity?

-Food enmity is calculated after gear enmity in the calculation. Example using this case:

Enmity from gear and merits: 49
Enmity from food: 3;

First calc with gear: 80E x 1.49 = 119.2
Second calc adding in food: 119.2 x 1.03 = 122.776

- The last possibility I can think of is that 1 enmity is not equal to exactly 1%. That its just very slightly more than 1%.


Anyhow the point of all this is that enmity does not cap at 50. Any ideas where the calculation went wrong?

[Liselle]

Anyhow the point of all this is that enmity does not cap at 50. Any ideas where the calculation went wrong?

Well, everyone discusses things as if they were in base 10 to make it easier to talk about, but the game's internal calculations are different. Getting 10% physical damage reduction doesn't mean 10%, it's closer to 9.7%, or 25/256 (if I remember correctly, anyway). I'll bet 10 gil that precision and rounding of the internal calculations has at least contributed to you getting strange results, if it's not the sole reason.

I think it's moot in practice, personally, since wearing that much +enmity gear is not likely to happen. The only important bit of data is that +enmity is roughly equivalent +% enmity, and additionally that it has no effect during Sentinel.

[Valyana]

Any ideas where the calculation went wrong?

How long did you wait for the VE from Hojo to decay?

[barber2006]

This is exactly why I have your LJ bookmarked. Fantastic work. I think it could really open a lot of new possibilities for the future when the tables are completed. For example, a ninja subbing Drk can:

sleep twice a minute for 640 CE
aspir once a minute for 320 CE
stun 1.33 times a minute for 239 CE

for a total of 1199 CE


OR that same ninja could sub bst and charm 4 times for a total of 1280 CE, all without needing MP.


Of course, this doesn't include things like souleater, but you see where I'm going. It's quite possible that there are some mazurka-esque subjobs out there just waiting to be discovered. Heck, dispel has a 10 second recast. That's 1920 CE from dispel alone. Add the CE from 2 sleeps per minute and you have double the CE of a /drk along with a nice 7.5% fastcast and a MDB trait too. Of course, casting 2 sleeps and 6 dispels a minute is very MP heavy, but lets just see how this shakes up. I bet some new doors are about to be opened.

[OctavoGilgamesh]

3) Distance relative to the mob at which an ability or spell is used does not affect enmity gained.

I've found this to be false. Often times I will run away to Convert (usually about 30 distance) and get no hate, whereas if I do it at 20, I'll get a ton of hate. However, if I do it at 10, I will get the same hate as doing it at 20, so it is partially true.

Over 30 is completely outside of hate range and won't affect your enmity at all, as any BST will tell you.

I'm not sure if I missed anyone who might have elaborated on these points, but I was close to posting a reply on Kaeko's LJ on this issue. The only thing they tested for in relation to distance was direct actions on a mob, not indirect actions that still cause hate like bst and smn pet commands or a simple erase. Your right in saying anything you perform beyond 30 distance causes no hate towards the player performing the action. And since every ja or spell that will be used against a mob or a player engaged to a mob will be cast within 21 distance or so depending on terrain and models, well within the 30 distance threshold, they will cause hate.

[Ashira]

Has someone tested what Foe Sirvente does to the VE decay rate-- which I'd imagine is what the decrease in enmity loss is referring to? Can't seem to find BRDs that have it merited :<

Awesome flippin' blog btw >.> Well appreciate the work

Yes, Kaeko and I tested Sirvente and Dirge the other day as well; but we don't have accurate numbers I believe because of using the rather imprecise method of counting seconds.

Essentially +Enmity gear raises the spike on VE usage; -Enmity lowers it. Dirge works like -Enmity gear, but Sirvente works entirely differently and instead slows down the decay of VE. It was rather interesting, but we only did the beginning tests to set our controls and precedents, didn't really go after the numbers yet.

We want to get working on Cure numbers, like Elipse just mentioned, it's just been a matter of making sure we have everything really nailed down before going on to the more complicated formulas. Testing really can't be rushed or you end up forgetting variables.

Kae wanted to ask if anyone could confirm the precise spike for Flash, Stun, Souleater, and Last Resort value so we can test something else tonight. Stun/Flash are likely 1200, possibly 1280... and SE/LR are around 1300. It shouldn't be too hard to figure out but if you really need the methodology, PM... well, PM Kae because I'd probably try to explain too much and make someone more confused. >.>

[Elipse]

Any ideas where the calculation went wrong?

How long did you wait for the VE from Hojo to decay?

Err more than enough. Don't forget its like 30 seconds or something inbetween hojo ichi recasts.


Edit: I understand its completely impractical to know some of these things. Still curiousity gets the best of me.

For those that also get curious, guess what? While I had already said sentinel was 100 enmity or at least had the effect of 100 enmity, it wasn't confirmed if it was in fact 100 enmity in the calulation or simply a doubling of base hate. Sentinel grants 100 enmity. Its not a doubling of base hate. I confirmed it with a novia today.

Double edit: This proves the enmity cap is 100.

Who cares? probably no one except me and my curiousity

[Ashira]

More precious from Alla:

After I made a remark about now we know precisely why PLD/WAR tended to fail so badly as an HNM tank --

I do hope this is a joke, because PLD/WAR most certainly does not "suck so bad." Maybe the player sucks, but a smart PLD will still hold hate off of everything short of stun spam.

One of the biggest issues I am having with this whole "testing" is it is making this unfounded assumption that holding hate is the ONLY part of being a tank. PLD sub war for the VIT and DEF bonus, not just Provoke, NIN for the added damage which in turn generates hate (which your study is failing to include) There are many factors that are just not being included in this study and its unfortunately leading people to use one thing over another over the idea of "hate generation" without taking into account the added benefits, or LOSS of benefits certain combinations cause.

I mean the simple fact is whats the use of a NIN/DRK if once the shadows drop they get their ass kicked harder than a BLM thanks to the weak DEF and HP DRK has since NIN certainly doesnt have any VIT.

I'm not sure if I'm more amused by his comment about NINs getting hate generation from doing damage (can't remember the last time I used katanas on Tiamat....), or the way he makes it sound like the NIN/DRK, PLD/NIN, RDM/NIN idea just spawned in this last week with our tests, or by the fact one of his fellow Allakhazammers told him /NIN and /DRK give a higher VIT boost to his PLD.

Speaking of... I know someone mentioned that they tried it on JoL with some success, but I'd like to see some more PLD/DRK testing... :D

[Liselle]

You definitely could have shitcanned that entire post once you saw the bit about subbing /WAR for VIT and Defense Bonus. If I were reading that thread, I probably would have posted a snarky reply suggesting that the OP look at a nearby calendar and note that it's no longer 2004. Also, that hitting Goblins in the arm doesn't make them drop the bomb.

[VZX]

Maybe I missed this because I didn't read the whole topic throughly :
Sometimes when I pull a monster and my ranged attack misses; Before anyone voke/flash/etc the monster I pulled, the monster will attack the /healing mage first instead of myself or anyone that doesn't /heal when I got back to camp.

How this model explain it?

[Ashira]

Maybe I missed this because I didn't read the whole topic throughly :
Sometimes when I pull a monster and my ranged attack misses; Before anyone voke/flash/etc the monster I pulled, the monster will attack the /healing mage first instead of myself or anyone that doesn't /heal when I got back to camp.

How this model explain it?

/healing or /sitting in sight range of sight aggro/linking or around sound aggro/linking mobs provides some spike of VE hate, and if you remain sitting, it does not *increase* and continue to build, but instead counts as a spike for each hate cycle that the mob figures. It is something Kaeko and I started just preliminary tests on yesterday and need more time to figure out how to quantify it more accurately. It was pretty interesting though; I healed behind a bunny and got nothing, but when he turned it around, it would begin to pivot towards me.

There is something with hate cycling as it works the equations, as Kae thinks. I thought it sort of figured it at the beginning of each attack round but that doesn't prove out because sometimes it would take a swing at me and other times it just barely pivoted to look at me - so it depends where the decay is in the hate cycle at the beginning of each attack round.

It's simple and yet annoying. :D

In any case, the spike is higher than pulling CE, so when the enmity is determined by the mob however it cycles as you pull to camp, the /healing spike will figure higher than your current CE in one of its processes.

Maybe Kaeko can explain it more clearly.. I think very artsy-like, and he thinks very mathy-like.

[VZX]

Maybe I missed this because I didn't read the whole topic throughly :
Sometimes when I pull a monster and my ranged attack misses; Before anyone voke/flash/etc the monster I pulled, the monster will attack the /healing mage first instead of myself or anyone that doesn't /heal when I got back to camp.

How this model explain it?

/healing or /sitting in sight range of sight aggro/linking or around sound aggro/linking mobs provides some spike of VE hate, and if you remain sitting, it does not *increase* and continue to build, but instead counts as a spike for each hate cycle that the mob figures. It is something Kaeko and I started just preliminary tests on yesterday and need more time to figure out how to quantify it more accurately. It was pretty interesting though; I healed behind a bunny and got nothing, but when he turned it around, it would begin to pivot towards me.

Could you be more specific on the method?

There is something with hate cycling as it works the equations, as Kae thinks. I thought it sort of figured it at the beginning of each attack round but that doesn't prove out because sometimes it would take a swing at me and other times it just barely pivoted to look at me - so it depends where the decay is in the hate cycle at the beginning of each attack round.

It's simple and yet annoying. :D

In any case, the spike is higher than pulling CE, so when the enmity is determined by the mob however it cycles as you pull to camp, the /healing spike will figure higher than your current CE in one of its processes.

Wouldn't it be a contradiction if that's really the model?

Also on your /Kaeko tests, whenever you guys say "player x takes hate". Does it imply player x really locked the hate on them as long as no other actions that decrease/increase enmity? If it's locked, then there's no little VE spike during hate cycling. Therefore, it'd be a contradiction, wouldn't it?

[Ashira]

At even hate.. i.e., third party pull, I cast dispel, Kaeko casts dispel - we are even but it faces Kaeko.

If I /heal, it would glance at me and then turn back and continue to swing at Kaeko. It would pivot slightly every couple seconds to glance at me while remaining on Kaeko... though sometimes the timing in that was such that when it turned to me its attack time was also up so it would swing at me.

Kae really has to explain this, I'm horrible at it because I try to give all the details and end up too confusing.

[Kaeko]

Regarding Elipse's Enmity Test on Hojo:
The fact taht it took 4 cures instead of 2 means that each Hojo was 80x1.52=121.6>>Round>>122. So the truncation may not work the way I had posted. My best guess would be that Enmity works on some base 2 scale like haste. Would require some fancier tests to show that though. The model we have is definitely incorrect for this little inconsistancy though.

Regarding Foe Sirvente and Adventurer's Dirge:
Dirge at level 1 is simply Enmity - 10 (the equivalent of putting on -10 in gear). Foe Sirvente does not "alter" your enmity, but lowers the decay rate from roughly -60VE/sec to somewhere in the realm of 54-56 VE/sec (I'm just guessing it brings it to -55 VE/sec but I can't prove it at this point). Provoke with Sirvente on is still 1800 VE, but it takes about 3-4 more seconds to fully decay. This was using level 1 Sirvente. I would be curious to know what level 2 and 3 did. I think Dirge's is pretty obvious since it says "-3 enmity per level" so I'm guessing a full Dirge is -16 enmity. Anyone who has a level 3 Foe Sirvente I would really appreciate help on it.

Regarding Aggro While /healing or /sitting:
There are a couple of basic "rules" we've established for this which are a bit odd.
1) They produce NO CUMULATIVE ENMITY, only VE (roughly 120ish? we aren't sure)
2) You only get that VE if the mob can "sense" you, i.e. if you have a sight mob and /heal right behind it, nothing happens
3) The range of "sense" is exactly 10' Distance. Rest a 9.9 and you get VE, rest at 10.1 you get nothing.
4) The hate seems to be applied in some sort of odd cycle count - we can't explain it right now.

Regarding "takes hate" terminology:
Yea this one is a bit ambigious on my part when I wrote it. Typically though, when I say "takes hate", it means the mob turned but probably turned back. If a mob turns then stays turned indefinitely, I'll usually say something to the effect of "took hate permanently". I know some of the parts are inconsistant with this, but generally, I'll specify if something only turns for a bit and when something turns and stays turned.