Modeling Enmity
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EDIT 11/09/07: This entire post is now outdated and COMPLETELY INCORRECT.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.html
____________________________________________________________________________________________________
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 goes along with the idea that the idea of Cumulative Hate is flawed. Because the hate cannot decay according the the theory, the longer you are actively performing actions in the fight, the more "hate" you have. If someone joins in the fight much later on, supposedly he can never catch up unless he really turns it up a notch. We know this is not true in practice...
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/viewtopic.php?f=2&t=23324
http://www.bluegartrls.com/forum/viewtopic.php?f=2&t=23599
Obvious pick this time... FFXI WIki gets caught cheating donors.
The idea of BG Wiki sounds exciting!
____________________________________________________________________________________________________
____________________________________________________________________________________________________