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 something is grossly wrong when I see it - and this was the case when I saw this thread on BG called "Do people still argue about enmity".
I was surprised by the large percentage of players on BG who just took the information posted as fact without even thinking about how it applied in real situations and if it matched. I think it's tempting to assume something to be true, especially when it looks well written and has some nice charts; however, it has no supporting evidence or real thorough testing behind it.
In this post, I want to go over why I believe that the enmity system the JP WIki and this post assert is incorrect and suggest an alternative take on enmity gain and loss. Because this topic is so hard to test with proper evidence, I'm basing this post on the fact that it's hard to prove something but it's easy to disprove something. More importantly, I will go over why the proper basic outlook on enmity is important and how this leads to some pretty fancy manipulation of game AI in game (from a Black Mage's perspective of course).
This is just based on my own opinion and what I've seen and done in game. Please take it for what you will and feel free to argue with me. While the theory on how this mystery works can be disputed, as long as the outlook one has on enmity leads to better play and strategic planning, I think it serves its purpose. Yes the graphs are really really bad and I did them in paint!
EDIT 11/09/07: This entire post is now outdated and COMPLETELY INCORRECT. I would not trust any information posted in this post regarding enmity. I will still leave it up; however, for an experimentally tested model of enmity (i.e. a correct one), please try here...
http://kanican.livejournal.com/13235.htm
________________________________________
Explanation Versus Observation
There are some basic observations that any endgame player can easily witness regarding enmity. I want to highlight the word "observation", meaning something you can see and claim but offer no explanation as to why it occurs. Whatever hypothesis one comes up with to explain what is easily observed and reproduced is the explanation - this is what I want to disprove (JP Wiki's version of enmity theory). Anyone who has a good Science background should understand how I am going about this discussion.
A theory is an explanation as to why something observable occurs which has a large body of supporting evidence and no evidence countering its explanation. Keep this in mind when I start to summarize what JP Wiki asserts.
Here are the main observations which theories on Enmity gain and loss hope to explain...
- The "Hate Cap" - where 2+ tanks tanking in a long fight will eventually balance hate and easily switch
- Perceived Enmity Loss - players will lose hate over time if they do not perform actions
- Enmity per Action Differences - abilities and spells do not all grant the same enmity
- Spiked Hate Loss - certain actions like getting hit, etc. appear to cause hate loss
These are pretty much universally accepted observations about this game regarding enmity. None of these observations have explanations attached to them, which is important. That is what any attempt as forming a theory has to try to explain.
________________________________________
How JP WIki Explains Enmity
If you don't feel like wading through the JP explanation, I'll try to briefly explain it. There are some basic assumptions which I think the vast majority of players will find to be self-evident and that I will not try to dispute.
- Certain actions grant more enmity than others (eg. Stun gives more hate than Blind)
- Enmity can be lost as well as gained
- Enmity can be reset by certain actions (KO, Zoning, etc.)
What is not as self-evident is their explanation of 2 types of enmity, which they call volatile and cumulative. Both spike hate instantly, but as the name suggestion, volatile hate decays over time while cumulative hate remains static. Here are some graphical representations they use to aid the explanation.
Well this is basically all they really have to explain things. You have 2 types of hate, and some things fall into Volatile and others into Cumulative. This explains some key observations seen in game regarding tanking -
- Why PLD/NIN, RDM/NIN, and NIN/DRK are very good tanks - the hate moves are mostly cumulative
- Why a PLD/WAR that only uses provoke sucks and cannot keep hate off most things
Just right off the bat, I have a somewhat big problem with this. This is because it seems obvious to me that they based this explanation heavily on tanking observations only. It seems clear to me that this came about to explain the success of the tanking styles employed by NIN/DRK, PLD/NIN, and RDM/NIN. It seems a bit backward to take a finished "product" (a tanking style) and go backwards to explain why it works... maybe I'm the only one that thinks this is odd. In the real world, you would never do something like that; I know this is a game and not real life but the same concept should apply I think.
The Most Obvious Flaw - Cumulative Hate is too Strong!
What really irked me about this theory is that this flaw is so damn obvious I don't see why others don't scream it out when they see this theory. This theory states that Cumulative Hate moves do not decay over time. I'll state a quick hypothetical on why this is a huge flaw in the theory.
I put a NIN/DRK to tank something and he does so for 30 minutes... Lets take something like Tiamat - Damage taken by the NIN/DRK is negligable due to Phalanx 2 and a Fire Resist Set (maybe 200 damage every 4 minutes due to touchdown). Pretty easy task not to get hit. Now lets say at the 30 minute mark, I add a second NIN/DRK tank and they both start spamming. The first NIN/DRK has a 30 minute head start on hate. Can the 2nd NIN/DRK catch up and eventually balance?
The answer is of course... and pretty quickly. It'll take maybe 5 minutes at the very very most. I think anyone who is well-versed in NIN/DRK endgame tanking will agree here. But try explaining this with the JP theory... The first NIN had a 30 minute head start on hate and that hate was supposed to not decay. No-one got hit for big damage so there was no real hate spike... This theory is not consistant with this observation. Honestly, that's all I would need to invalidate it.
This is what is supposed to happen according to JP Wiki... hate is never balanced
There is another example on this flawed idea of Cumulative hate I can highlight as BLM on Fafnir, but I will save it for near the end.
Cumulative Hate Cannot Explain Hate Cap and Hate Balance Over Time
This idea of cumulative hate cannot explain the hate cap which is easily observable in any long, straight-tanked fight. There is obviously some sort of hate cap phenomonon in this game and this theory does nothing to explain it. Even harder would be to explain hate balance - where 2 tanks can balance hate relatively easy on an NM. We know in practice that hate balances without that much effort after a short time. Try explaining that when someone supposedly had a 30+ minute head start on cumulative hate gain...
My Explanation - Don't Overcomplicate Things
It is pretty impossible to figure out what the exact formula for hate is in this game due to lack of testable data. Because of this, it is really hard to assert "this is how it works". Instead, I only offer a "model" - something that is not what really happens, but will "ball-park" the answer for you to a relatively high degree of accuracy. It is an oversimplification of how hate really works, which is not a bad thing for a model since it is easier to understand and apply. This model will accurately describe all/most observations seen in game, unlike the JP Wiki with decent reliability to the point where you can accurately predict unseen cases.
My explanation is mostly consistant with what JP claims, with the exception that cumulative hate does not exist. There is only 1 kind of hate gain and you can call it whatever the hell you want.
Different moves and actions grant different levels of enmity, but enmity follows the same basic principles regardless of how it was gained or lossed!
Basically I'm saying if 2 people are at the same hate level, the fact one gained it with a Thunder 4 and the other got it with a Flash + Provoke does not affect how their hate will go down. The next assertion is probably the one that differs the most from the JP Wiki theory...
Natural Enmity Decay is always present and is affected only by your current hate level - the more hate you have, the more hate you will lose per unit time
This is such a simplistic answer to so many observations in game regarding enmity that I find it to be quite "beautiful" (yes that makes me sound horribly nerdy). If you're familiar with Calculus, this is a common concept that can be written something like this...
Where t is time, C is some constant value. This leads to a really commonly seen curve like the one on the top right (I know if you plot the stuff on the left you do not get the stuff on the right, I'm just being extremely general in trying to explain the math). Big thing to get out of this... The more hate you have, the more hate you will lose.
It is really difficult to explain enmity any further with math because of the lack of any formal testing so I won't even bother. All I will assert is again, the more hate you have, the more hate you will lose!
"Hate Cap" is a State of Enmity Equalibirum
If you follow the 2 basic principles I stated above, the concept of a "Hate Cap" becomes pretty intuative. There are 2 things constantly affecting your hate - the hate you gain over time, and the hate you lose over time. The more hate you have, the higher that second value is... Eventually once you reach a certain level of hate, the amount of hate you lose over time becomes so great that you can only sustain your overall level of hate through actions. This point is the hate cap... it's not really some formal cap on hate, but it's just where you happen to hit that equalibrium when hate gained equals hate lost.
This concept also easily explains the issue we ran into earlier with the JP explanation about Tiamat tanking with NIN/DRK. So how does that 2nd NIN manage to balance hate after only 3-5 minutes? The answer is all he had to really do was reach the equalibrium hate level of the first NIN. That 30 minute head start doesn't really help since after the first 3-5 minutes, the hate was pretty much static anyways!
Why is Hate Balance so Easy to Accomplish in Practice?
The answer is again quite simple in this theory. Hate gain occurs in spikes (through actions), but hate loss occurs over time (think smooth decay). Because of this, both tanks (let's think NIN/DRK again) are not actually always at that equalibrium, they just hover around it closely above and below it. Tanks trading hate is something that occurs automatically and naturally due to natural fluncuations in hate around that hate equalibrium. This is a much stronger explanation than that offered by the JP Wiki, where trading hate seems so much hard to do on paper than in practice.
Why is Understanding Enmity Important?
In my opinion, it's really not crucial to completely understand Enmity to the point where you want to model it mathematically; however, it is important to understand how it work if you want to really push the limits of most jobs. It is also especially important to someone attempting to formulate new strategies on HNM. I want to highlight a couple of strategies which can be employed (and that I or TK have employed on a regular basis) that stemmed off this better understanding of hate gain.
A lot of this will be BLM oriented of course, but I will try to make it more relavent to tanking HNMs in general, since that is where enmity is usually highlighted. I want to be very clear that if you follow the JP Wiki train of though on hate, these strategies should not work, which further proves that it is flawed. This is why I really felt it necessary to post something that almost directly attacks another's train of thought. When it starts to inhibit strategy, it needs to be fixed.
I think the biggest backing I have for why this theory is the correct one (or at least more correct) is that I develop ideas on how to do things in this game based on this idea - I forumate ideas based on the concept. In the case of JP Wiki, I feel they made the concept based on the strategies and ideas already used... It's a backwards way of doing things.
Example 1 - Chain Nuking Fafnir
This sounds like an absolutely ridiculous idea, and ok, it's not exactly chain nuking Fafnir, but it is pretty close. I think anyone on Odin Server that has seen me fighting Fafnir and watches me closely knows what I'm talking about. One of the first things any BLM is told when first fighting Fafnir is to not overnuke or you will die. Pretty standard and decent advice considering I still see a ton of BLMs dying on it due to this overnuking - but this leads to a big problem... Black Mages get timid and don't like to nuke as much as they could.
So what does knowledge of how Enmity works have to do with this? How does my theory on how this works change how you fight this mob as a Black Mage? The answer is simply this...
You can nuke as much as you want as long as you nuke in a series of small bursts.
Why is this allowed? And why will nuking fewer, larger bursts get you killed? Let's look back at one of the key assertions I made earlier about hate -
Tanks trading hate is something that occurs automatically and naturally due to natural fluncuations in hate around that hate equalibrium.
A Black Mage that nukes a constant volley of smaller attacks is essentially acting as a tank! (I'm talking thunder 3s and 4s, still no 1s and 2s... I just avoid using MAB gear and MBs) The only difference is you are intentionally spacing out when you take hate and when you do not... I will try to better explain this graphically.
Nukes in the 700-1000 range still have a huge jump in hate, but they are smaller enough such that you will only take hate for maybe 10 seconds. On a Fafnir with Elegy and Slow, this is only about 3-4 hits, which is easily countered with Blink, Stoneskin, and a Damage Down Standing Setup. What I essentially do in this fight is get near the top of the hate list along with NIN/DRK tanks and trade hate in volleys.
Thunder 4 > Take Hate > Blink and Skin Rebuff > Thunder 3 > Thunder 4 > Take Hate > repeat
I do not wait to nuke. I know a lot of groups like to "let the tank build hate", which would be the proper thing to do if you follow the JP Wiki version since a build up of cumulative hate gives tanks a "head start" in enmity. Of course, this is flawed and you will actually eventually cap it anyways so it doesn't hurt to start nuking pretty early assuming you can handle those 3-4 hits.
If you follow JP Wiki, it's pretty easy to see the cumulative hate flaw here. Ever come late to a Fafnir fight on BLM and get hate really quickly? It's because you hit that cap so fast due to Fafnir's low magic defense. That 20+ minute head start really doesn't do anything for the tank. I've seen BLMs join in late on their LS's fight, pull off a nasty Burst 2 MB and get hate in like maybe 30 seconds - it's not like the tanks were doing a poor job, that's just how enmity works and doing a 3000 damage Burst 2 is going to take hate.
So why avoid using fewer, bigger nukes? If you get too high above the hate equalibrium provided by your tanks (we use NIN/DRK usually), you will have hate for too long to survive. Using smaller nukes keeps you balanced with the tanks so that you only take 3-4 hits at a time. If you put me in a party with a Ghorn, I will literally solo DD Fafnir from 100% in about 35 minutes just straight chain nuking - my only limitation is MP, not hate.
For BLMs who want to see this for themselves, you don't even need great gear to do this. I don't nuke in MAB gear, but Elemental Skill gear. I'm intentionally lowering my damage to balance hate better. All you really need as far as gear is a decent Elemental Skill set and a good Standing Gear setup. This is a great example of how strong a player one can become with just a basic understanding on how hate really works in this game.
Example 2 - Thundaga 3 Trains in Dynamis
Sometimes you get big trains in Dynamis and the puller brings back like 20 mobs. If you're in a smaller group like TK usually runs in, usually what happens is you sleep them as they come in, but realize the instant this stuff wakes you'll be wiping (unless you pre-timed your sleeps and layer). The solution I usually come up with on this is pretty odd and really looks like suicide but it works, in large part due to how hate works in this game.
I'll use the example of Dynamis-Buburimu. Anyone who's ever had to pull nightmare mobs in this zone knows how damn annoying links are (rabbits and mandras especially...). Sometimes you end up pulling like all 20 mobs and if you go in with like 18, you can just sense the wipe coming. In order to prevent this, I take those slept mobs and I actually Thundaga 3 them, then train the mobs to another area. The reasoning behind this is you gain tremendous hate off damage and the mobs will go after you and not the sleepers. I then run them out and am able to gain distance due to terrain abuse + Herald Gaiters.
When I gain sufficient distance, I can sleepga or sleepga 2 them without getting hit (distance plugin and /RDM fast cast help here). With elemental seal, the timer on all these mobs is now layered and 1 good BLM can now hold 10+ mobs easily - this means 1 good, heady BLM can save a wipe from 20+ mobs in Dynamis. I've done this multiple times just in the last month since TK has started doing more Dynamis-Buburimu and our pulls have been shaky. Still, we rarely wipe because I can seperately maintain 10+ mob links easily using the Thundaga + run trick.
But what does this have to do with Enmity? Cool trick but doesn't seem like it has to do much with Enmity gain. Well, I needed the huge burst of damage to spike hate off the sleepers initially, but according to JP Wiki, damage is cumulative hate that does not shed... Of course it actually does shed (especially if you put distance between you and the mobs by kiting). By the time the alliance is done killing the first half of the mobs, the mobs I have have shed most of the Thundaga 3 damage hate to the point where a simply provoke gets it off me! That is why this works! If hate never shed, I would be dead at this point just because Thundaga 3 is such a large hate boost, the tanks would take too long to surpass it without the shedding factor.
This actually does work! You can try it if you have the right gear and practice. If you just assumed the JP Wiki version of hate gain and loss to be true, I think most would not ever try this knowing it to be suicidal. It's just another example of why it's important to really understand how it works.
Example 3 - NIN/DRK Tanking with Enmity Equipment
I get a lot of questions about this with regards to NIN/DRK tanking gear. A lot of opinions differ here on BG forums so it is somewhat of a hot topic. I've heard that you only want to use enmity gear to reach the hate cap, but after that, it does not matter - this could not be further from the truth in practice and it's very obvious if you stack a lot of enmity once reaching your hate equalibrium.
So what does it help once you hit the hate cap? This was stated earlier in the post...
Eventually once you reach a certain level of hate, the amount of hate you lose over time becomes so great that you can only sustain your overall level of hate through actions
The key thing to focus on is that the hate cap is not some magic number... it depends on where your hate gain equals your hate loss. Adding enmity gear increases hate gain per action or unit time, so the equalibrium is actually higher! This makes enmity gear useful even once you hit the cap, because the cap will be higher! The most noticable thing I've seen personally thusfar that illustrates this is our Ouryu fights. When I was first testing gear for NIN/DRK out in LS, I noticed that an Enmity setup allowed full hate control; whereas, a standard damage down only setup was unable to do this even with the same BLM nuking and same fighting style.
What happens to allow a single NIN/DRK to maintain hate no matter what the BLMs try to do to take it? The hate equalibrium reached by a NIN/DRK in full enmity gear is so high that Black Mages just cannot reach it by simply nuking. Because damage is halved on Wyrms like Ouryu, the hate decay prevents Black Mages from getting enough hate at any time to overtake the NIN/DRK without using things like bind, sleep, and stun spam.
________________________________________
Conclusions
There are probably more examples to cite, but this post has started to run on kind of long so I'll stop at this point. Again, my main purpose here in this post was to try to present a "better" take on enmity and how it works. I don't claim to have "discovered" or "made" this idea, but I did try to show how it could be used to better strategies in game. In the end this is what it's about to me.
The model I have suggested is not going to be the actual method hate is calculated in game and I do not contend it to be. Instead, I see it as something that "best-fits" hate gain and loss and can be used as a prediction model for strategy. It is simply an explanation on why observatable events in game regarding hate occur in the way they do.
I know this idea probably has its flaws as well and I welcome anyone who wants to try to find them. At the end of the day, it's about using the knowledge of game mechanics to develop strategy. I just felt that bad information leads to bad strategy, which is why I felt it necessary to directly counter another source, JP Wiki. There are still a lot of unknowns about hate and a lot of things I failed to cover. A big one I can think of is how distance plays a role in shedding hate, which I think most players will agree it plays a big role. I think players already understand how this can be used in strategy development on kited fights, so it's not something I think needs to be talked about in depth.
________________________________________
Pic of the Day
Seeing as how the post this time was pretty math oriented and "boring" if you will, I'll post a couple of more pics this time to even things out.
Two Tarutarus in Morrigan's Robe!
"Kraken Club Do you need it? Gaijin Price 70M, JP Price 50M and will debate price or trade items!"
N. Legs Drop from Tier1 Einherjar - first confirmed "good" drop in Einherjar by any group.
Couple of JPs in TK made a GM call about harassment after doing Ballista and killing some poor guy about 50 times. I think Saku was pretty mean to him or something and he started harassing people, got temp-banned, then got on another account to harass some more. If you're on Odin, you can probably figure out who the guy was...
Me getting Novio Earring! (had to put this on here)
Drama Thread of the Day
http://www.bluegartrls.com/forum/viewtop
http://www.bluegartrls.com/forum/viewtop
Obvious pick this time... FFXI WIki gets caught cheating donors.
The idea of BG Wiki sounds exciting!
________________________________________
________________________________________
Page Summary
- (Anonymous) : "beauty" [+0]
- (Anonymous) : (no subject) [+1]
- (Anonymous) : (no subject) [+1]
![[info]](http://l-stat.livejournal.com/img/userinfo.gif)
kainsin : !! [+1] - amlaruil : (no subject) [+2]
- aurik_ffxi : enmity [+3]
- (Anonymous) : (no subject) [+3]
- (Anonymous) : (no subject) [+1]
- kanican : Post Edit and Changes [+6]
- sylphet : (no subject) [+4]
- (Anonymous) : fantastic [+1]
- (Anonymous) : (no subject) [+1]
- (Anonymous) : Resist gear [+1]
- bhoreas : (no subject) [+3]
- ono_valefor : (no subject) [+0]
- tacrozar : (no subject) [+0]
- (Anonymous) : (no subject) [+1]
- (Anonymous) : I agree [+0]
- genomeffxi : hate decay [+0]
chasethestars
Comments
"It must be beautiful"
Nice work, anything with functions and graphs gets my thumbs up :D
-random LJ fan.
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)
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.
Or the Redmage for that matter.
every time period do: hate = .9*hate
One thing you don't address is the possibility that actions a mob does to you reduce hate. A specific example of this is Ghrah; if you aggro one (but don't cast on it) it'll lose aggro after 2 actions that fail to hit you. Of course this doesn't seem to match any of the normal observations about enmity, so it may be a different mechanism altogether. Other evidence that taking damage reduces hate is difficult to differentiate from enmity decaying over time.
another problem with a decay function like that is how hard it is to maintain higher levels of hate (which nin/drk generally shows can be maintained), if this was the case a blm nuking would lose hate almost immediately unless the unit time was very long (ie. around the 5-10s mark) which we can observe is unlikely.
overall it would be difficult to get the correct function (0.954, 0.831, 0.9206....) compared to giving static hate loss/unit time.
it could be done that way of course, i personally dont think it is.
To me, the importance is to develop a model that can somewhat accurately predict what would happen (for strategy), not necessarily figure out what SE really does to calculate hate. Even if said model does not work, if it accurately describes the vast majority of cases, I think it still has tremendous use.
Your proposed system has elegance, but I think yours is not completely accurate either.
Here is a situation that yours fails to explain:
I was camping rapido with whm, 2 rdm, me on blm. I was keeping a cactuar slept so that rapido would link when he ran by. We had some bad luck getting the link, so I ended up sleeping, binding and graving this cactuar solid for about 20 minutes. At that point, the cactuar was glued to me.
I stopped casting anything and the red mages spammed hate spells at it (sleep bind blind grav etc) for about 5 minutes, but it continued to hate me. I even had the white mage teleport me to Altep; he targetted the red mages while I was at the telecrystal, but as I re-approached our camp he made a beeline for me and once again was impossible to pry off of me.
At no time did I take any damage from this cactuar: he was too low level to hit me reliably, and could not break my stoneskin--though he took plenty of swings that missed or hit for 0. We didn't do any damage to him either.
If as you suggest hate decays purely based on current hate accumulated, then at some point my non-action should have allowed the red mages to eclipse my hate. However, regardless of what they did they were unable to. I had reached some nirvana where the only way that anyone could pull hate was to take damage.
I guess my goal here isn't to figure out what SE is really doing to calculate enmity, but to try to find the model which can most accurately describe all cases. Perhaps there were some oversimplifications if it couldn't cover this case. If someone had told me the situation you described, I would not have guessed what you said actually happened (so the model fails in this case).
As tanking has evolved for us as a linkshell, one of the most critical things we've done it to always make sure that the tank party firmly establishes hate (ie @ the "hate cap") before starting damage. Good gear and skilled tanks help to reach the "cap" faster. However once that has been established, it has been my observation that almost nothing can take hate off of the tanks. We've killed Tiamat in less than an hour and in two hours + depending on our setup.. and regardless, people don't have to log any more because our tanks keep very solid hate.
This is even evident in XP. If a tank uses provoke followed by enmity generating spells/JAs, a melee can usually pop out a big WS and not have the mob turn to them. However if the melee opens up with a big WS before the tank has had a change to establish enmity, it's much more difficult for the tank to get it back despite spamming JA/spells.
As stated, a good tank can still walk in to a situation mid-fight and gain enmity quickly. A BLM could too if they decided to chain nuke > manafont. Yet, as you mentioned, a BLM can do more damage by using a combination of tier III and tier IV nukes spaced out and never pull hate. On the other side, a weak tank can try to come in to a fight late and will never attain the enmity level of the "good" tanks.
Bahamut V2.0 BC is a good example of enmity management. (We don't Kraken DRK it). We typically use a single tank on Bahamut and he/she always establishes hate while the rest of us just sit back and support while DDs wait. This is always the riskiest tanking situation with us for enmity because there's a single tank who inevitably will take some sort of damage due to flares. However the next person to get hate is inevitably the RDM who is in charge of dispelling (and sometimes stunning) Bahamut, despite the BLMs nuking him for massive damage. If you're going to use the JP idea of enmity, the BLMs apply "spike" damage later in the fight but the RDM is consistanly applying "cumulative" enmity on a regular basis. I'm not ready to exclude a bipartite system for enmity generation/loss but I really like your ideas as well.
In summary, the best strategy I've seen for the maintenance of hate on tanks is two fold:
1: Early establishement of hate on tanks.
2: Temporally spaced out damage from DD.
It's clearly not an easy topic to tackle, so great job on a nice analysis of it!.
1) Reaching the idea of the "Hate Cap"
2) Hate Decay
3) Maintaining "Hate Cap"
Both you and Aurik have mentioned the idea that establishing hate first leads to a much better stranglehold on hate by tanks (ie. letting the tank reach the cap first before having others scale the enmity ladder is more effective than having others scale the ladder at the same time while reaching cap).
I must admit that this is something that this is something definitely true and readily observable/reproduced, yet the model I've given does not predict this should be the case; this is a big flaw in how it works since this is quite an important observation.
I think at this point, I want to take a step back and just tackle one part of the enmity issue instead of attempting to address it on the whole.
You posted something later on which I really liked so I'll post more specifics on that one!
Also there are other issues like your assumption that all hate is equal. In this I believe the JP information is correct and that there are different kinds of hate actions and how the hate gained from them dissipates. Of course all hate dissipates, I think anyone that believes other wise is grossly misreading the information and it is easily denied by observation, but that doesn't mean that different kinds of "hate gets" do not have different kinds of decay.
-ringthree
Basically, with maybe resists on 1/6-8 nukes, I would blow 1000MP, and even Manafont on Ouryu chain nuking it and it would not look at me. This is with repeated attempts. The exact same circumstances were used in a seperate attempt where the NIN did not use enmity gear, but the standard haste+phy damage gear and the results were quite drastic and I would pull hate relatively easily w/o 2hr usage. This observation suggests that enmity did play a factor.
Again, like I have said in some of the other replies, it's about finding a suitable model to predict unseen cases, not to figure out what exactly SE is doing with hate. If a model can fundimentally describe most unseen cases, I think it has tremendous use. This model I outlined is something I personally used a great deal in formatting strategies on this blog.
The Ouryu and Tiamat strategies which are now so popular due to this blog contained observations which backed this theory; which we then used to figure out how to format a tanking scheme for fights like Proto-Omega and Proto-Ultima. I feel it is quite successful in modeling unseen cases, which is why I eventually felt comfortable enough to post it.
The JP Wiki is a terrible model in my opinion because it attempts to overcomplicate how it works in order to cover more and more obscure cases. In doing so, it still fails to cover many ideas and common observations. I do not know of any action that appears to have a seperate decay rate. I know provoke gets this alot, but it's usually because they tried to only use provoke (which is doomed to failure in just about any model), or got hit many times in the process (PLD/WAR style).
I do not see how this is a viable argument because the point of contention is an observation based of what is essentially a controlled test. We had 2 sets of fights with multiple trials, the same tanking setup and style, the same buffs, the same "Nuker" (me), the same nuking gear (meaning the BLM's gear is controlled and is not variable), the exactly same circumstances when nuking (when it's airborne, at any part of the fight). The only variable was the selection of tanking gear, which is why I feel this is a pretty good observation to bring up.
Based on this observation, the model (which I do not propose to be the way it actually works, but something which can accurately model most cases seen in game) fits well and can describe why this observation was seen. I do not think of this from a Mathematical point of view because I am not trying to prove anything. Think of it from more of a Sciencific point of view, where my goal is to simply find a model which accurate describes observations - which mine does in this case, and rather simply.
The mathematical explanation is very elegant and it still is incredibly simple because it asserts that all forms of hate follow the same rule with little/no exception. It is probably not how it really works, but it "ball-parks" the answer if you will to the point where you can accurate describe unseen situations (the point of scientific testing). I think your beef with this is the "testing" process in the controlled experiment on Ouryu, not whether the method I describe fits the observation (ie. you have issues with the accuracy of the observation, not the model's explanation of it).
As far as enmity, many very highly regarded NIN/DRKs use this enmity setup on switches. I think the best source of if it works or not is just their experiences and choice to use it. I believe Rukenshin also follows an enmity setup similar to those described here as well (not that he necessarily is correct 100% of the time).
Again, I cannot offer anyone "proof", but only that this model works for the vast majority of situations seen in game. Math was only used to help describe the nature of the model, but I think people have confused it as something I asserted to be what SE does with hate. The actual formula is probably way different, but you can think of this model as a "best fit" version of it - simplified and easily processed.
firstly i cast bind and blind (10+ and 30+ times respectively) while ghalleon was on his way.
ghalleon then used meditate/provoke/warcry/berserk/hasso/se
i still had hate at this point.
we then waited for a few minutes, after X time, the mob turned to ghalleon, neither of us had taken any action for a while. the mob hit me for 0dmg a few times (stoneskin) but this didn't seem to relate to when it turned away.
i then cast bind/blind until hate returned to me, at this low lv of hate ghalleon used his abilities and the mob quickly returned to him (then died because he had macroed in a blaze spikes piece...).
next ghalleon aggroed a mob, this was especially interesting to watch:
i cast blind and the mob came to me, missed me once~ and returned to him. i cast blind 5 times more and the mob would come to me, miss once and return to him.
either my hate decayed or evading attacks reduces enmity. at this point i wondered if who took initial hate was relevant so:
i aggroed a mob, took some hate, then ghalleon used his abilities, the mob went to him of course, i then used blind to bring it to me - again it would return to him. proving to me that who had initial aggro was not a factor in this case. incidentally i was on 37whm, he was on 75sam.
all this leads me to draw 3 conclusions:
1. level is a factor in who has hate when hate has totally decayed otherwise on players of different level.
2. bind hate decays. and therefore most likely all hate joins the same pool and is part of the same decay process.
3. the rate of decay is measureable and from max hate to minimum takes a period with a magnitude of minutes.
i screenshotted most of this naturally, the only reason i cant really define laws about hate rather than say this is strong evidence is that taking 0dmg on stoneskin or misses from the mob might be the major factor here and i will when i have time kite a mob to prove otherwise.
-ciz
(and that stuff about how its unlikely to be a fractional decay above is also me).
I never considered level difference playing a factor since I focus so much on endgame activities where everyone is 75. I think the model I've outlined probably is an oversimplification considering the complexity of how enmity works, but I don't see that as being a bad thing necessarily.
Models are not meant to be 100% exactly how something actually works, it just has to be able to accurate describe what actually happens in the simplist way possible.
I think this model does that and is still pretty accurate - to the point where you can count on it being correct if you try to predict enmity and hate in a strategy or job setup. The fact that you can accurately model most cases in game with such a simple answer to the enmity question makes it useful in my opinion.
I wanted to further stress that this is a model (an oversimplifaction) that attempts to best fit how enmity reall works, but does not attempt to actually describe how it is calculated in reality. It is meant to be simple, elegant, and easily applicable when discussing strategy and predicting cases in game - which I feel it does extremely well in most situations.
Thank you for all the comments thusfar and I look forward to hearing more takes on this!
My explanation of this is the speed at which hate is lost is different for every action you take (Cure, damage, JA, etc). Apparently, doing damage is pretty much the slower kind of enmity loss (maybe it's just nuking, who knows), AND you can never cap hate: you can reach your own hate cap depending on what you do (all the NIN/DRK hate tools), but in my previous example, it's clear that a BLM can cap hate higher than a NIN/DRK given time.
I don't know how long you're talking about for Tiamat fights, but I've been through 3 hour snore-fests low-manning this using the standard NIN/DRK + BLM Nuking in air strat laid out on this blog. I'd be pretty much one or two of the only sources of DD the entire time, so I'm doing roughly 50,000-60,000 damage in a fight like this (which I would consider an extreme).
I do not experience this drawn in or getting hate that you claim occurs so I don't know what is going on in your fights. Perhaps a Horrid Roar goes off or NIN/DRK afks in the air too long? I think if I'm not taking hate after 60,000 Damage, that's reason enough for me to believe I won't be getting it given unchanging conditions. My main observation which backs this model is the Wyrm fight, because I asserted so strongly back then that I found it basically impossible to get hate purely off damage (I can off damage, stuns, and sleeps if I really wanted probably).
Unfortunately, the biggest issue I'm finding is that this is based off anecdotal observations for the most part since clear testing for this is difficult to document without going to video. Your Wyrm experience completely contracts everything I have ever documented about Wyrm fights in all of my posts regarding Tiamat pretty much (which is also really anecdotal). I would like to be able to say that there was some flaw in the tanking method or some other uncontrolled factor, but it's really hard to tell based off observations like this.
I think one issue with your possible interpretation of hate types is when you take 2 BLMs joining in at different times (let's say on Fafnir, 1 joins 20 minutes late). If there is some type of residual hate build up like you suggest, then the BLM that started early should have some type of built up hate and will take hate easier than the one that just joined (given they both reach near the hate of the NIN/DRK). From my experiences this isn't really true. Once you are near that cap and start to surpass it after a nuke, I've found myself in the exact same hate situation as others who do this but started earlier.
Anyways, good job on the post. One thing you really didn't comment on is how serious BLMs (especially) should be about enmity- gear. I know some people take it way too seriously and some don't give a damn. Did you ever do some careful tests of whether or not (for instance in your Faf chain nuking example), wearing Errant Cape instead of say, Prism Cape, makes you "tank" less, or maybe even not at all at some point?
As for wyrm fights, the BLMs usually aren't stunning. We just sit down and rest the entire down phase unless dealing with elemental adds. I have gone through fights where I had to stun a lot though (such as when we solo tanked it thoroughout) and still no hate produced. Usually BLMs have to rest most the time on the ground I think, unless we have access to Ghorn that day.
getting a working model is obviously all you need to succesfully strategise, everything you describe and theorise is correct in this respect and i don't think anyone disputes that; we only talk about the maths because at least in my case i wouldn't care about being a horrible tank if i could know the numbers behind it.
incidentally mechanics is correct in how it describes motion and kinetic forces on the macro-scale, statistical and quantum mechanics are applied to things where the mechanical component doesn't make all the others insignificant (for example if you lift a cup from the table, the van de waals forces involved are generally negligible), classical mechanics is still applied because its the best model for that scale.
likewise quantum mechanics works well for the scale it is used on, but its actually all about as realistic an interpretation of what is happening as classical mechanics, a lot of times i've heard "this part tends to zero so we just ignore it to make the calculation simpler".
-ciz
I would praise anyone who really attempted to figure out what actually happens when people calculate hate, but it's really of no concern to me personally since I don't think it provides a huge benefit strategically over a simplistic model. Most people cannot grasp whatever is really going on (myself included probably), and it would really serve no use due to its complexity unless someone took the equations and made it into some automatic plugin calculator. A model is superior in this I think due to its simplicity.
This is like classical mechanics versus how it "really works" (aka Einstein's equations on relativity). You don't bother introducing the complexity of "how it really works" when a much simpler model gets you high accuracy anyways.
This is about the most simple model you could possibly come up with as far as enmity modeling in my opinion. I think most people agree enmity decays, and I've used what is essentially the most basic and common form of decay you can find (first order). You see this type of decay and function a ton in nature and in life (RC circuits, cell growth, finance, etc.) and it's not like its something I magically pulled out of nowhere. I just found it quite interesting that the most basic model possible was able to explain just about all my observations on hate in this game.
This could all be random garbage with no meaning whatsoever! /disclaimer on.
I've always wondered exactly how + enmity and -enmity gear works. We know it helps! but in what way? I created a hypothetical scenario with the assumption that everyone starts out at with a base hate ceiling. + enmity gear raises that ceiling and -enmity gear lowers it. Therefore, someone the lowest enmity would be incapable of reaching the cap of someone with high enmity. This does NOT mean that someone with low enmity can't pull hate, clearly they can but it would mean that it happened because the tank was not at the enmity cap. However I think it could help explain certain phenomina.
Here's a graphical representation of a hypothetical situation. Assume that base enmity is at 5. The WHM has -1 enmity bringing enmity cap to 4 and the tank has +4 enmity bringing their cap to 9. As long as the tank maintains an enmity level above 4 the WHM shouldn't be able to pull hate, however if they do drop below 4 and the WHM is at their personal hate cap, the WHM will pull hate.
Clearly the real situation is not as simple as this because not all actions are done with a static amount of + or - enmity gear on but I think it's a possible explanation as to how a tank could maintain solid hate over the course of a long fight as well as why some tanks never seem to be able to attain another's hate level.
Basically, the concept I want to test is can enmity gear seperate the respective hate caps of tanks and non-tanks enough so that they can never overlap when the tank is played well (represented graphically in the chart you showed). I'm trying to see if you can manipulate gear and playstyle to reach an "idiot-proof" strategy on HNM fights where (assuming people play correctly), it is impossible to get hate off the tank (similar to how Tia and Ouryu were outlined on this blog). I think this may be actually testable to an certain extent; and it is certainly possible on certain fights in my experiences with Wyrms.
The reason I think this is so important is due to fundementally why tanking schemes like NIN/DRK, PLD/NIN, and now RDM/NIN have become so popular. Why did these styles overtake PLD/WAR and NIN/WAR at most long endgame fights? I think the answer is that that hate is so secured using these styles that most/all others in the group do not need to log or even worry about hate. I remember doing Tiamats NIN/WAR and the BRDs almost always drew hate and we were afraid to use anything but SMN and TA WS DD for this reason. The breakthrough of using these styles now give way to drastically different DD styles such as BLM ganking, which would have been suicidal in the past.
So my big question is how to increase the chance or ease of a tank reaching this hate "nirvana" (as Aurik called it) - this would be like the example I posed in the post where the tank holds hate so well I could nuke 60,000 in damage w/o it looking at me. Reaching this opens up an incredible number of possibilities on how groups do fights - this "nirvana" is probably the single biggest breakthrough in Wyrm fight strats (requiring a good deal of seperatedly discovered breakthroughs like fire resist gear, TP limitation rules, and /DRK to eventually reach).
If enmity gear can raise the cap and increase the chance of being able to reach that hate "nirvana" state (and someone could prove it), I think it would be a tremendous breakthough. Likewise, if using enough - enmity gear can lower or negate the chance of a non-tank reaching the state (like the WHM in your example), it could be just as important.
Thanks for your interest in this and that nice chart! (I may use it in some future posts if this turns out well if that's ok with you) Outside of the concept of "kiting", reaching this hate "nirvana" (I really like that term Aurik) and proving it even exists is, in my opinion, biggest breakthrough in endgame strategy.
I remember Tiamats the same as you described. WHMs had to log for hate. Tank pt BRD and RDM would always pull hate. RDM/DRK would always have to log. BLM would get drawn in, shit we'd make people log every 2k damage they did. I think one of the problems with /WAR during extended fights is the limitations of using a 30s recast JA as your main hate tool, with not much to do but lose hate in the remaining 30 secs. NIN and PLDs are not effective DDs on HNMs so that doesn't contribute anything to their hate maintaining capabilities. PLDs have Flash, and NINs have enfeebles but it still is just not enough to compete. There are just more opportunities to lose hate in between vokes than there are opportunities to gain hate.
PLD/NIN and NIN/DRK work very well as a team. PLDs rarely have MP issues any more with the plethora of new gear and job adjustments given to them in the past year or so. NIN/DRK can basically spam spells for hate as long as they have refresh and haste. If PLD lags behind in hate, NIN can souleater while PLD spam cures them. Both are happy, amd magically, the mob never manages to turn it's head from the tanks unless something goes wrong.
I think I can come up with a controlled experiment that's easy to pull off for the whole how does enmity affect hate cap question. I'll try posting results once I get it set up and done.
I really like recalling really old fights on these mobs. CoP wyrms were especially huge when they came out. I remember everyone really getting excited anytime any LS beat these for a good 6+ months. I actually kept some really old videos I found of various groups doing these back in the day as like reference. Looking back at them now it's pretty cool to see how the endgame community has progressed.
It's still a flawed idea based on some of the responses I've gotten so needs some work. I'll try to put some formal testing to it later on. The biggest dispute is this idea of hate cap. My biggest concern is how enmity + and - afffect this (since it actually leads to something useful in fights).
TK is a JP and NA mix. It's roughly 75% JP 25% NA so what is seen on this blog can sometimes be misleading since it's in english. The screenshots can also be misleading since they are taken in NA times for the most part and NAs talk more in general in here (talk as in spam the LS/party chat). I personally love it not for the accomplishments but really the atmosphere and all that good stuff. Much of the membership hasn't really changed in the 3-4 years it's been active, which I think is rare and a testiment to the members.
We still do a lot of kings. Rest of the stuff can change depending on current needs. We'll survey the entire LS every couple months to check wishlists and determine what mobs to focus on depending on which things people want most. It's not really too structured really; I think you have to be flexible to last this long since the game changes.
The NIN/DRK tanking has me a bit confused, are you advocating staying in enimity+ gear when they are flying (to maintain higher hate cap) or switching to resist gear?
I have my own personal item to ask about.. When playing BST I have noticed if I am sending a weaker pet at a stronger prey the damage will look something like this: (These are made up just to illustrate the damage difference)
Pet 15-15-20-15-20-15-20-15-20-15
Prey 30-35-30-40-30-40-30-40-30
Naturally you would assume @ anytime if I jump I would pull hate, but if I let them fight for 5-6 rounds, I can generally jump in and fight alongside for the remainder of the fight with out pulling hate. Even though my pet has never done more damage than the prey.
Would this be evidence that damage done is caculated differently then damage taken?
You definitely would not use enmity gear over resist gear on a wyrm while airborne. I think no matter what you believe for enmity, it's understood damage taken is a huge source of enmity loss. Resist gear was, in my opinion, one of the 3 big accomplishments in strategy developed and popularized on BG etc. that turned wyrm fights from huge multi-alliance fights to low man attempts (the other 2 being RDM/NIN, PLD/NIN, and NIN/DRK tanking and understanding and exploiting wyrm TP reset). Resist gear prevented massive hate loss in the air (and massive hate gain by mages on cures); new tank styles increased hate margins by a significant degree; and exploiting wyrm TP (and lack of tp regain) eliminated fear of any bad TP moves such as flail and horrid roar.
My idea is based purely on just what I see w/o real experimental testing. I couldn't tell you anything about your situation w/o really extrapolating and guessing. I am trying to set up some very basic testing to answer some questions. So far the results have been surprisingly odd and not what I expected.
My one question/situation ive seen happen many of times. A blm will sleep 2 a particular mob, and I will provoke it immedietly after it's been slept, yet when it awakes, it will go after the BLM even though no one has performed any further actions. What is the concept behind this?
Also, I was wondering about "idling" in enmity gear, and wether it makes any differnece, or is it just when you perform actions it's calculated in. Meaning if enmity actually does raise the hate cap, will removing it lower it immedietly? My understanding of it is you dont need to idle in your enmity gear, for it makes no difference, but I just want to be sure. I have not tried experiments with this.
The best way to do this would be to prove it indirectly but understanding how enmity gear works. If enmity gear increases/decreases enmity at the point of action, aka "instantanious", then it would not affect you when idling. On the other hand, if enmity gear affected hate decay or some other "continuous" process, then it would matter when idling.
This is really an extremely fundemental concept as far as enmity goes and I don't think anyone can really answer it at this point.
As far as your BLM sleep question, there really isn't much information you're providing me but I'll go ahead and guess it had to do with the relative distance between you 2 and the mob. It could be any number of missing observations though.
Coming back, so many comments. Here are some of mine (albeit probably too old for the discussion).
1) I believe there is a hate cap. However I think your description of the hate cap that I'm experiencing is a MUCH better description than the normal hate ceiling that gets commonly described/envisioned.
2) Enmity gear affects the hate level for sure. At the start, it may seem to not work, but it definitely does raise the cap for longer fights.
3) Your Thundaga anecdote is counter-intuitive. But when I think about it... I always scoffed at PLD's that Provoke + Flash immediately instead of spacing them out because it rarely worked for me in being able to hold the monsters attention longer. Perhaps another anecdote for the "Faster hate generation leads to faster hate loss" idea?
4) Sylphet's graphical model is probably best describes what I believe to be true. But seriously... with rukenshin the tank... there should never be that drop in tank hate! =p
Its known that loosing a shadow shed's some enmity. Generally a small amount, doesn't seem to effect exping on nin a lot (if any). Lets assume there is no cap on the amount of enmity you can have.
Maybe the reason that hate is so easily shared by nin/drk's, and the reason there "seems" to be a hate cap is that shadows at that enmity level are losing more enmity than expected. (big enmity, big enmity loss theory)
This would cause the mobs target to flip-flop around people on a similar enmity level. It also means there is some kinda of cap you could never get passed assuming the rate of shadows lost and the enmity generation rate never changes.
Aurik's experience with the cactuar seems, to me at least, to be somewhat valid under this idea if we assume that being hit for 0, and being missed is no enmity loss(Actually i think missing does, but cant explain it any other way). We also need to change the no hate cap assumption I made before and say that there is a hate cap, but normal tanking never reaches it because of the shadows and damage taken enmity loss.
He had a great deal of hate, and nothing he was doing was causing it to decrease. Other people tried to pull hate off him, but if we assume there is some kind of cap on hate, it could be reasonable to expect that the person that reaches that cap first to take priority over the other person.
BLM's enmity to my understanding decreases over time, otherwise kaeko's posts about how to nuke fafnir wouldn't work in practice. How to fit this in is confusing me. 2 types of enmity doesn't agree with me for some reason.
Hate Caps and enmity gear:
Some of the wrym fights i have been involved in have been rather comical when compared to how they are killed today. Looking back on them makes me think that enimty- or + gear does not effect the "cap" on hate.
My reasoning is that at some point every whm that was there (of the up to 6 week had outside curing) would have to logout. I didn't understand hate as well as i do know, at the time i just pilled on all the enmity- gear i could. But I still eventually would get drawn in. We used to nin/drk tank it with 3 of them, so roar was never a reason for drawing hate on tiamat.
What i noticed was that enmity- gear seemed to just delay the inevitable, not stop it. I would usually be the last to be drawn in, i also had the most -enmity gear on. The whms with the least on would be the first to go. So my observations are at this point:
The more enmity- gear you have, hate gained per action will be reduced (diminishing returns on the gear of course)
The more enmity- you have the faster you shed hate if you actually take damage.
Enmity -/+ gear doesn't seem to affect the hate cap.
Barefoot - levi
This idea of using + enmity gear to raise the hate cap is something I would like to be able to prove somehow in the future.
You didn't mention this in your post but I also believe that taking damage also lowers enmity level, and this is a tactic I use often. When I pull hate, rather than running away and hoping tanks pull it off me, I'd rather run toward tanks, or stand still and take the hits till the mob gets bored with me. Almost always I'll come away alive with an enmity level low enough to keep the mob completely off of me for the next several minutes while I rebuff and heal myself.
Good post!
I don't know if I would say that certain hate is really constant, but that is not contradicted by a hate cap, you can certainly have both, and it might just decay more slowly. But a lot of more complex situations have too much happening to make clear conclusions, people can reach hate caps, lose hate from getting hit, etc.
One much easier and clear test maybe would be having two people with the same job and equipment (say PLD with movement+ and same enmity) get on the hate list of a monster, then one cast a cure spell, the other use provoke while wearing some type of movement+ gear to kite a slow monster without getting hit.
If the two types of hate-generating moves decay at different rates (including possibly one not really decaying with time at all), then the monster should eventually change target away from provoke back to the person that casted cure.
Repeating but using a stronger cure spell to start with the monster running after the cure-spell guy, the monster shouldn't turn back toward the provoke guy at all.
A third experiment for confirmation would be to start with a number of cures that gives more hate than provoke and warcry used back to back, then trying again with provoke first and warcry only after several minutes of kiting, to see if the hate of cures was reduced enough to be surpassed by warcry.
I think those would be simpler more unequivocable ways to figure out how hate works, if certain moves decay at significantly different rates even when they generate the same hate (or not by themselves).