Enmity Test Results and Conclusions (Translation)

[Rydin]

The problem with that though is that when people have equal hate whoever did the last action takes hate, so the enemy would be bouncing all over the place and targeting whoever took the last swing.

Im pretty sure the tests have proved that when people have equal emnity.. the person who got to that enmity value first has hate

[Estellios]

Hmm you might be right then. I was mostly going off of this from the first post:

First Attack Bonus does not exist!
Also, if two people have the same amount of enmity, the enemy goes after whoever targeted the enemy first.
However, speaking from a GLD's perspective, I guess it's important to get the first strike in order to strike a combo and grab the enemy's attention...

The large enmity gap thing probably won't be that big of an issue at the end of the day and maybe it is there to help tanks get hate back in case they die.

[Rydin]

Hmm you might be right then. I was mostly going off of this from the first post:



The large enmity gap thing probably won't be that big of an issue at the end of the day and maybe it is there to help tanks get hate back in case they die.

I never thought about it like that... thats a very reasonable assumption though... makes sense

[Deathgiver]

Definitely some discrepancies here. When I tank Ifrit I do the following: Everybody gather on tank behind Ifrit, I pull with AA, hit Sentinel, Flash, Provoke, Rampart, sacred Prism and finally cure. Everybody moves out from that point and starts attacking. As they move out I perform the flat blade combo. The total hate generated should be equal too roughly 5k based on his testing, yet about half the time one of the thms pulls hate with his first lightning combo, until Flash and provoke are off cooldown, after which no problem holdig hate for the rest of the fight. Also using sentinel sabatons and +10 enmity belt. I know for a fact none of the mages are outputting 5k damage in a single combo.

[Treach]

8M wasted on a belt I didn't even get just to find out how near useless it would be?........

[Stanislaw]

Definitely some discrepancies here. When I tank Ifrit I do the following: Everybody gather on tank behind Ifrit, I pull with AA, hit Sentinel, Flash, Provoke, Rampart, sacred Prism and finally cure. Everybody moves out from that point and starts attacking. As they move out I perform the flat blade combo. The total hate generated should be equal too roughly 5k based on his testing, yet about half the time one of the thms pulls hate with his first lightning combo, until Flash and provoke are off cooldown, after which no problem holdig hate for the rest of the fight. Also using sentinel sabatons and +10 enmity belt. I know for a fact none of the mages are outputting 5k damage in a single combo.

Let's see if we can break it down.


1.) Auto-Attack. There is no first attack bonus, so Enmity = Damage. Let's assume for a moment that it does 10. Total so far, 10 Enmity.

2.) Sentinel. Activation generates around 14~31 Enmity, so for the moment let's assume 20. Total so far, 30 Enmity.

3.) Flash. Generates about 380 Enmity. With Sentinel's x1.2 boost, this will generate 456 Enmity. Total so far, 486 Enmity.

4.) Provoke. Generates about 550 Enmity. With Sentinel's x1.2 boost, this will generate 660 Enmity. 1146 Enmity.

5.) Rampart. Rampart generates around 190, multiplied by the number of people it hits. No boost from Sentinel. However, it is possible that Rampart generates the multiplier hate ONLY if all 8 members on the party have some sort of hate on the target mob (Ifrit). For the moment let's assume this was the case. 1336 Enmity.

6.) Sacred Prism. 10~31 Enmity. Assume 20. 1356 Enmity.

7.) Cure (/w Sacred Prisim). Let's assume Cure heals for 400, and because 1 Enmity = 0.6 Cure, this means this will generate 240 Enmity. No boost from Sentinel. However, like Rampart, let's assume that it didn't generate any extra hate because nobody has attacked yet (doesn't have hate on Ifrit). 1596 Enmity.

8.) Flat Blade. Flat Blade, in a combo, generates enmity equal to double the damage dealt. Ifrit has high defense, so let's assume it only did 100 damage. 200 Enmity. Assuming Sentinel is still active at this point, 240 Enmity.


So, with the above assumptions in place, your actions will generate somewhere around 1800 Enmity in a worst-case scenario. I suppose it might be possible for the THMs to take hate with a one or two Thunder Combos then.

This is all based on assumptions of course. It might be entirely possible that mobs generate different amounts of hate depending on mob, or Ifrit might have some special battle mechanic in place and enmity is calculated differently.

For example, because Ifrit actually has 5 targets on the battlefield (the four pillars and Ifrit himself), it might thus be possible that the game is dividing the enmity to those pillars even when they're still hidden, thus reducing the enmity generated on several abilities by 1/5th.

[Mychael]

So in theory... if 1 person has 100... and the tank is capped at 27100... as long as the person at 100 never generates any more enmity... no one else in the party can possibly pull hate from the tank???

discuss.....

I don't know if this would ever come into play on anything short of Pandemonium Warden. 27000 appears to take a long time to gain, especially considering that the cap should keep going up unless someone is leeching. ^^;

My question though, is what happens when someone dies and gets on the hate list again--does everyone above 27000 immediately drop to 27000 cap plus whatever amount of emnity the revived player gains? If an ability is used that lowers emnity for the person with the least emnity, does the cap go back down?

[urhryu]

I'm not too sure about the +/- enmity materia gear mentioned but I do find that with -12 enmity belt, I could almost cast additional set of combo on Ifrit before hate gets to me. If it's 0.1% as mentioned, -12emity would just be only 1.2% which is far too low a value to make any difference isn't it?

[Kaeko]

http://lodestone.finalfantasyxiv.com...entry?e=307111

Sorry the image is hard to read. I had to post quickly to head off to work. Glad to see some other testing going on. I see a lot of discrepancies in the table so I'll have to dig at the methodology differences when I get back.

Here's a more readable version of the table I posted this morning.



I sense that there is likely a difference in methodology that leads to slightly different values. I produced this table using "rock throws" to find the lowest enmity values (Protect, Sanguine Rite, then a couple of the "19"s), then used these abilities to match hate. I assume that the JP testers used pure damage. Most of the values are either within the JP table or within the realm of reason (for instance, Flash and Voke are off, but look close).

Couple of notes just eyeballing:

(1) First Attack Bonus: I agree, when I tested this pre-1.20 this was the case as well. This differs from XI, which had a significant pulling bonus. Be aware that there is actually a tiny "bonus" for aggroing the mob though (1 damage worth).

(2) Enmity Decrease over time?: Agree, there is no time-decay component of enmity. Again true pre-1.20. This was not the case in XI however.

(3)Provoke on GLA vs. MAR?: I agree it was the same when I tested this patch, though our value for voke differs. However, the rank of the player matters significantly so it's more about having equal ranks as opposed to equal jobs.

(6) Cure enmity: The blogger uses the 1.6 cure : 1 damage ratio whereas when I tested I merely inversed it. I got the modifier to be roughly 0.625 damage = 1 hp cured, which comes to the same thing.

(12) Enmity from gear: When I tested this, it really threw me off because I thought I was going crazy seeing it do almost nothing. I actually wrote it off as broken, but then a member of my LS swears it makes an tangible difference so I thought there was a high probability of testing error. It's very nice to see he reaches near the same result (that gear enmity does almost nothing); however, I would not write off the use of this completely as we both reached this conclusion with similar testing methodologies and they BOTH may be flawed.

(14) Hate Cap: obviously very important for endgame hate management but I have not looked at this in 1.20. In 1.18 I saw absolutely no evidence of a hate cap though, tested up to 30000+ damage worth.

Enmity Table: most of my values are either 4, 19, 114, 55, 363, and then some oddballs. I think they mostly fit within the ranges the JP table has so I'm confident that they are in the right ballpark. However, the precision may not be up to the point as I had them in XI.

**********

Finally, I want to talk about the very adept poster who mentioned that these numbers do NOT seem to hold any weight on Ifrit fights, noting that a GLA can spam an initial flurry of actions and then have hate stolen in 1 THM combo. This was quickly noticed by my LS after we completed similar testing in 1.20 and we're honestly still lost on it.

For Stan's explanation, I think it's fairly accurate in terms of #s, but I would just add that yes, the players that are hit by prism cure and rampart need to be on the hate list for the tank to get any benefit. So if you open with a prism cure and no one else has done anything yet, you waste it.

But the reason I feel that something is off when trying to apply these hate studies to an Ifrit fight is you will often-times see a pickup group Ifrit fight go 10+ minutes and the tank will have been cured 50,000+ HP by a healer. If we convert this value to 'damage" with the 0.625 ratio, we get 31,250 damage. If you then take a look at the damage parser, you'll see THMs doing maybe upwards to 12,000-13,000 damage. Poll the players and the THMs are always the ones complaining about hate, not the healers. Shroud of Saints is only a temporary enmity decrease so it shouldn't alter the results. Also, the main tank in my LS swears enmity gear works and I have to say, eyeballing (as much as I hate using it to drive research) it it does seem to do something; however, from both my own and the JPs' tests, it should be near unnoticeable.

I do think that the current look into the enmity system for XIV is probably only hitting the surface as we're missing at least 1 very large concept here (given that the "on paper" math does not translate to an actual endgame scenario). I am really excited to see others do enmity research though. Thanks to Stan for the awesome translation (and for Guildwork )

[Treach]

inb4 monster level/rank affects enmity gained from abilities but not damage and possibly cures.....