Enmity Testing by Kaeko

Credit: Kaeko
Source: Kaeko's Enmity Tests

Overview In the post 'Modeling Enmity', I went over a possible "ball-park" method of looking at enmity in this game which was simple, easy to understand, and plausible for a good number of situations seen in the game. While easier to understand, and I think a better model to base things on than the JP Wiki version of things, it is still not what actually happens in this game. At the time, I felt it was not possible to fully understand what was going on regarding enmity; however, I think this may actually be possible now. In this post, I want to lay out how testing, and proving, how enmity mechanics work in this game can be done. Something like this is not immediately intuitive, so I will lay each concept out in an experimental test based on rejecting or accepting a "null hypothesis". My hope is that these tests and findings will build upon each other, eventually leading to a full comprehension of enmity mechanics. The first series of testing involves ideas that are incredibly basic - many of them are probably "no-brainers"; however, I felt that it was important to establish them, as they are the building blocks of harder to reach concepts. Some of the initial findings were quite surprising and worth reading about I think - and the key idea is that these are all proven in easily reproducable, controlled tests. Nothing about this post is anectodal. This post is going to involve a lot of scientific methodology and does not allow many opportunities for screenshots that are actually meaningful to the discussion. Anything you see regarding pictures is probably going to be a chart or some screenshot I have taken from the testing that really adds little or no content (aka it's filler).

Questions to be Answered

These initial tests provide the foundation of all future testing, making these by far the most crucial. That being said, most of the ideas that are 'proven' in these early tests are no-brainer ideas - they still have to be proven by an experimental test. These early concepts I've chosen to focus on include...

     -  Does being attacked for 0 damage cause an enmity loss?
     -  Does evading an attack cause an enmity loss?
     -  Does evading an attack through Utsusemi Shadow loss cause an enmity loss?
     -  Does the distance at which the spell is cast from affect enmity gained by an action?
     -  Does the act of pulling grant an enmity increase to the puller?

While extremely basic, I think most people could not give you a really good answer to any of these other than the question on Utsusemi, which SE stated is supposed to be true. The tests in these posts will answer these questions, although the later ones not fully.

These basic tests were chosen first for a few reasons. It is very difficult to prevent hate loss if getting hit for 0 or evading causes a hate loss - the mob you are testing on will have to get attacks off unless you constantly bind or use worms to test. Distance to cast was also important for similar reasons. Really these tests form the basis on what can and cannot be done for the more complicated testing to be done later on; so while they may seem boring or mundane, they are incredibly important in validating the methods used in future tests.

Test 1

Does evading an attack or being attacked for 0 damage cause an enmity loss?

     H0:  Evading or taking 0 damage from an attack causes no enmity loss
     H1:  Evading or taking 0 damage from an attack does cause an enmity loss

Testing Setup:

- 3 level 75 members with Dispel (RDM/NIN x2, BLM/RDM) - 2 characters must have 0 enmity merits (+/-), be stripped of all gear, and have access to "distance plug-in" - A puller character (3rd character) does not require specific enmity matches - Use a "level 0" monster - we chose Bumblebee in West Sarutabaruta

- All characters must have either Stoneskin or Phalanx buffed prior to starting to ensure 0 damage - No Utsusemi shadows can be active

- "Puller" pulls mob with Silence (a move causing no damage, but is known to grant less hate than Dispel) - The 2 test characters move to the SAME DISTANCE from the monster, we chose 0.7 distance - Both test characters, attempt to cast Dispel as close timing to each other as possible - Wait for monster to attack and choose targets - duration chosen was 2 minute wait

     H0:  The target that initially took hate after both Dispels were cast will maintain hate throughout 
     H1:  The target that initially took hate after both Dispels were cast will NOT maintain hate throughout

Before discussing the result, it should be noted that second character to cast dispel on the target always took hate. Why this occurred is not discussed in this test, this is a nice observation for future testing. In this test, distance from mob is controlled, as is time for the most part.

At this point, we have not shown that enmity decay over time exists or does not exist, so we must account for both possibilities. It turns out that this tests works for both situations. Obviously, if no natural enmity decay over time occurs, there is no issue. If it does occur though, the fact that both casts happened near instantaneously would limit the decay difference. Also, once both Dispels have been cast, assuming any decay works equally on both players, it would not, on its own, cause a shift in target, which was what was tested.

The results showed that the target never switched in the 2 minute period, after numerous misses and damage for 0 attacks.

Conclusion:

Evading or taking 0 damage from an attack causes no enmity loss

img451/7631/img20071029010735st2.png

Test 2

Does evading through Utsusemi Shadow loss cause an enmity loss?

     H0:  Evading an attack through Utsusemi causes no enmity loss
     H1:  Evading an attack through Utsusemi does cause an enmity loss

Testing Setup:

- 3 level 75 members with Dispel (RDM/NIN x2, BLM/RDM) - 2 characters must have 0 enmity merits (+/-), be stripped of all gear, and have access to "distance plug-in" - A puller character (3rd character) does not require specific enmity matches - Use a "level 0" monster - we chose Bumblebee in West Sarutabaruta

- All characters must have either Stoneskin or Phalanx buffed prior to starting to ensure 0 damage - Utsusemi shadows must be cast and active on both testing chacters prior to beginning

- "Puller" pulls mob with Silence (a move causing no damage, but is known to grant less hate than Dispel) - The 2 test characters move to the SAME DISTANCE from the monster, we chose 0.7 distance - Both test characters, attempt to cast Dispel as close timing to each other as possible - Wait for monster to attack and choose targets - duration chosen was 2 minute wait

     H0:  The target that initially took hate after both Dispels were cast will NOT lose hate after Shadow Loss
     H1:  The target that initially took hate after both Dispels were cast will lose hate after Shadow Loss

This test was conducted in the EXACT same way as test 1, with the only exception being Utsusemi was pre-buffed in this test. The result showed that no hate was lost for misses and attacks for 0; however, the instant a shadow was lost, the target immediately switched! This shows that unlike evading and attacks for 0, shadow loss must cause some type of hate loss. This is a good thing since SE claimed this to be true after one of it's updates.

Conclusion:

Evading an attack through use of Utsusemi Shadow Loss causes an enmity loss

Test 3

Does the distance at which the spell is cast from affect enmity gained by an action?

     H0:  Distance at which the ability is used does not effect enmity gained
     H1:  Distance at which the ability is used does effect enmity gained

Testing Setup:

- 3 level 75 members with Dispel (RDM/NIN x2, BLM/RDM) - 2 characters must have 0 enmity merits (+/-), be stripped of all gear, and have access to "distance plug-in" - A puller character (3rd character) does not require specific enmity matches - Use a "level 0" monster - we chose Bumblebee in West Sarutabaruta

- All characters must have either Stoneskin or Phalanx buffed prior to starting to ensure 0 damage - Utsusemi shadows must be off when this experiment takes place

- "Puller" pulls mob with Silence (a move causing no damage, but is known to grant less hate than Dispel) - 1 test character moves to 0.0 distance from the monster, while the other moves to 20.0 distance - Both test characters cast Dispel with close timing - Run this same experiment twice - one where the 0.0 distance target uses Dispel first, then second - Wait for monster to attack and choose targets - duration chosen was 2 minute wait

     H0:  The target that cast Dispel second will get and maintain hate regardless of distance cast in all cases
     H1:  The target that cast Dispel first will get hate at some point in time in ANY of the experiments

This test was conducted in the EXACT same way as test 1, with the only exception being the distance at which the spell was cast between the target and the player. The result showed that the second player to cast Dispel always took hate and maintained it throughout - this was true whether the second player to cast was 0.0 distance away or 20.0 distance away - signifying that the distance you perform the action from does not affect the hate gained or possibly decayed in that 0'-20' range

This is a very important test to nail down since most tests we perform won't necessarily require someone with Distance Plug-in. It is nice to control the distance used still anyways though. It has been pointed out to me that most BSTs know that there is some fringe distance at which you can be away from a mob and not take hate from actions - this is true but whatever that fringe distance is, it is greater than 20'. This test was more to make future tests easier rather than tackle the issue of where this fringe distance occurs.

Conclusion:

Distances between 0'-20' at which an action is used relative to the mob does not affect the hate gained

img141/9711/img20071028021405bd7.png

Test 4

Does the act of pulling generate an extra amount of enmity for the puller?

     H0:  Pulling does not add an extra amount of enmity to the puller
     H1:  Pulling adds an extra amount of enmity to the puller

Testing Setup:

- 2 level 75 members with Dispel (RDM/NIN, BLM/RDM) - 2 characters must have 0 enmity merits (+/-), be stripped of all gear, and have access to "distance plug-in" - Use a "level 0" monster - we chose Bumblebee in West Sarutabaruta

- All characters must have either Stoneskin or Phalanx buffed prior to starting to ensure 0 damage - Utsusemi shadows must be off when this experiment takes place

- 2 characters cast Dispel at close times, but far apart enough to be discernable who cast first - Distance from pulling is not taken into account due to the results of test 3 - Wait for monster to attack and choose targets - duration chosen was 2 minute wait

     H0:  The 2nd player to cast Dispel will get and maintain hate 
     H1:  The puller will get and maintain hate even though he cast first

This test was conducted in the EXACT same way as test 1, with the only exception being there was no initial 3rd charcter to pull. The distance at which the initial casts were made were not factored in due to test 3 showing that distance an ability is used from relative to a mob does not affect the hate gained by an action. This test questions whether a 3rd player acting as a 'puller' is even needed - if pulling doesn't grant extra hate and distance doesn't matter, the puller can be excluded from future tests.

The results showed that pulling a mob does actually generate some bonus hate! If you do this test, you'll realize it's really the same thing as test 1 except 1 person performs the action of pulling the mob. You'll find that the player that pulled it will maintain hate throughout - even though based on test 1, player 2 really should have had hate by virtue of casting 2nd. This shows that the action of pulling grants a hate bonus. Note that this test does not tell you how much of a bonus, only that one exists.

Conclusion:

The act of pulling grants an extra enmity bonus to the puller

Is there Hate Decay or Not?

This is one of the big questions out there - however, I have failed to try to answer this at this point in time because these background tests are required as controls before such a test can be done correctly. The problem with not asserting a hate decay exists or not is that these 4 early test results have to reach the same conclusion in both a model where hate decay exists and does not exist! The way they are constructed, these 4 tests actually do reach the same conclusion whether or not decay over time occurs.

The reason this works comes down to the mechanics of test 1 - all other tests stem from changing 1 factor from the controlled test 1. In this test, the second player to cast Dispel always took hate - this is true whether the casts were 0.1 seconds apart or 3 minutes apart. We always tried to cast as close to each other as possible unless we were trying to prove a point - the easiest way we found was to just time our casts at the same time - since we had a BLM/RDM and RDM/NIN, the difference in fast cast allowed the RDM to always finish first. You can really come up with 2 main arguments why the second caster always got hate from the same spell -

     1)  Hate decays, therefore the the player to cast first has decayed more hate than player 2 at all times

                   img147/1130/chart2eu3.png

           OR THE ALTNERATIVE...

     2)  No decay, so both players had a 'tie' in hate - in which case the more recent action gets priority

                   img132/4361/chart1dx1.png

This is what happens in our test 1 results - the hate mechanics after the spell is initially cast will not matter as long as the mechanics work the same for both players. The important thing is not where exactly on the curve 1 person is, only which player is higher than the other. The difficulty of enmity testing is you cannot measure the exact value of enmity, but only observe who has the most hate.

If, in test 1, we saw the opposite effect - i.e. the other player (the one that first cast the spell) eventually did get hate back from the second caster, we see that this alternative possible observation is covered in both models with and without hate decay; note that these graphs below do not model what the results and observations that test 1 showed, but shows that test 1 was ready to account for both scenarios (which it must to be a good test). These graphs do accurately describe the 2 possible models for test 2 - the test showing Utsusemi loss does have a hate decrease.

     Model 1 (Enmity Decay Exists) - Loss of hate due to Utsusemi Loss causes player 1 to take back hate.

                   img456/3345/chart4am3.png

           OR THE ALTNERATIVE...

     Model 2 (No Enmity Decay) - Loss of hate due to Utsusemi Loss causes player 1 to take back hate.

                   img237/6408/chart3ox1.png

So again, both possible common models - hate decay exists, or does not exist, are fully covered in the way these tests are performed. I've only drawn out charts for tests 1 and 2; however, tests 3 and 4 really follow the same methodolgy and I think you can work it out in your head pretty easily that the same would apply.

Conclusions

To recap the conclusions found in these tests -

     1)  Getting attacked for 0 damage or evading an attack causes no enmity loss.
     2)  Dodging an attack using Utsusemi Shadows does cause an enmity loss.
     3)  Distance between 0'-20' relative to the mob at which an action is used does not affect enmity gained.
     4)  The act of pulling (first action) causes an enmity gain for the puller.

I found most of these results to be pretty much what I thought they would be except the 4th, which surprised me a bit. Again, I don't there there is really anything novel in the testing gone over in these posts - they focus on extremely basic concepts. However, these concepts have to be nailed down and PROVEN before we can really build an experimental test that proves anything more conceptually difficult.

There is really a long way to go from here in order to really comprehend how enmity really works. I can tell you right now that some of the more complicated testing has resulted in very profound and new information. This post and the tests in it really just provide the foundation for any future testing (it establishes what factors can be controlled and/or how to control them).

Supposedly, all 4 of these tests should work no matter what base spell you choose to use. We chose Dispel; however, anything should work - Blind, Provoke, Flash, etc. I would avoid anything that messes with mob action though like Gravity, Bind, Sleep, or Stun since we are basing observations on what the mob does. Again, the key I want to stress about these tests is that they work in both a hate decay model and a non-hate decay model - this is and will become even more key later on.

Critique on the testing method is welcomed as they must be very sound in order for anyone to build upon their results later on. The results of these tests will be called upon many times later on - in fact, I have already done so within just these 4 basic tests. I will admit this post is very dry; however, for those who found this boring, I want to assure you that the results coming in the near future will pose very meaningful information on how people actually play this game, which is really the point of doing this.