When I first began testing, I copied an introduction, methods, results, and discussion type format similar to many typical research papers. Since the large majority of readers will not really care about sifting through this information, I decided to dedicate my methodology to a separate section. If I or any other reader needs to go back and see the process I used in order to try and replicate my results, this section will be the go-to reference. It will also contain my thought process while working through the data, so it might seem a little repetitive compared to the other blog posts. As a disclaimer, I try to involve statistics when they are helpful and necessary, but my overall understanding of statistics would be best described as novice (AP high-school to undergraduate) at best. My terminology is especially poor. A great deal of influence in my thought process and methods comes from the FFXI and FFXIV player, my friend, Kaeko Leta.
Table Of Contents
How To Read B.L.I.T.Z.B.A.L.L.
Red text will be reserved for chapter headings.
Green text will be reserved for section headings.
Orange text will be reserved for important text. This is to highlight messages/concepts I don’t want readers to just glance over.
Dark Orange will be reserved for very important text. I only use this a couple times per chapter. This is for concepts that every player should be aware of.
Yellow text will be reserved for all equation based text.
───Establishing The Number Generator───
The first step in assessing the damage mechanics with the new battle algorithms involves establishing which type of number generation system FFXIV:ARR uses. In FFXI they used a normal distribution (think bell curve) where the majority of damage followed a midpoint with slight variations of increased and decreased damage. In FFXIV 1.20, we established that there existed a random non-normal distribution. This means that the damage numbers are drawn at random with equal probability from a pre-determined set. The pre-determined set included an 8% variation from a midpoint value. For example, I am averaging 100 damage. The minimum damage I will do is 92. The maximum damage I will do is 108. The odds of doing damage for any number between 92 and 108 are equal. FFXIV:ARR follows the FFXIV 1.20 system. The only change I noticed is that they reduced the variation to +/- 5% instead of 8%. It is because of the equal probability and fixed set that allows my testing method to work. Instead of having to try to obtain 1000s of data points, I am able to “predict” when I have reached a minimum and maximum value. I would highly encourage readers to sift through this post by Kaeko Leta:
His overall methods are centered around Cure testing; however, the vast majority of his “Data Collection” and “Trial Size Discrepancy” sections are applicable to FFXIV:ARR. These are the core concepts behind my data testing.
───Collecting The Data Points───
For the vast majority of my testing I used Level 50 Water Sprites to test damage algorithms. I will make special note when I have used a different enemy. With that in mind, here is a typical example of a data collection spreadsheet:
You will notice several things:
1) I have not listed VIT, DEX, INT, MND, or PIE.
2) I have held DTR constant while I varied STR.
3) I have an “Avg” value for the Min and Max
4) The “%” column refers to the percent deviation from the average value from the min.
5) The % is almost never exactly 5%, but will vary a little over and a little under.
In regards to point 1, contrary to 1.20 and FFXI, there are now only 2 attributes that players can change with armor pieces that affect damage output. These two attributes are Strength (STR) and Determination (DTR). Vitality (VIT), Dexterity (DEX), Intelligence (INT), Mind (MND), and Piety (PIE) all do not influence non-ranged damage output. As a backwards method of confirming this, the high degree of linearity I have obtained in all my data without ever even attempting to hold these stats constant is strong evidence towards proving that they do not influence non-ranged damage output.
In regards to point 2, this one should appear obvious to most readers. In order to determine the influence of one variable, it is common procedure to hold all other possible variable constant. There are 3 actual variable that influence non-ranged damage output, Weapon Damage (WD), STR, and DTR. I kept my weapon (Mogaxe) constant and my DTR constant at 213, while I obtained 5 different STR data points.
In regards to point 3, I simply took the (Min + Max)/2. This number is to help calculate the “%” column.
In regards to point 4, I calculated the percent deviation from the midpoint for the min. This equation results in (Avg – Min)/Avg = % deviation.
In regards to point 5, this was alluded to earlier, but important enough to expand upon again. There is an inherent assumption (that evidence supports) that for damage output, the random number generator chooses values from a specified set. This set is:
Minimum Damage = Average Damage Output x 0.95
Maximum Damage = Average Damage Output x 1.05
This is the most important concept to grasp in understanding how my damage testing works. Now with this in mind, we can see that for many of my % columns I do not have a 5% value. The reason for this is due to rounding error. Take for example the % column value for the STR = 254 row, 5.52%. If you were to reduce the Max by 1 value and have a Min =77 and Max = 85, the % column would instead be 4.94%. The game’s rounding decides that this isn’t close enough to 5% and adds +1 to the Max placing it “closer” to its algorithm for 5%. To see how accurate this is, when you obtain damage output values in the 500+ range, you will start to see the % column approach a 4.95-5.05 range, due to rounding of higher numbers being much less significant. Once I get a reasonably close value to 5%, I will do about 50 more trials and if the number does not change, I will accept it as the true min/max.
With this in mind, I collected data points for varying WD, STR, and DTR while holding the respective other two attributes constant. I have so much raw data that it would be a pain to post. I will e-mail this data upon request to respectable endgame stat testers, not the general public. Sorry!
───Interpreting The Results───
Once I’ve attained about 3-5 data points, I will attempt to plot them to see if the returns of the variable I altered produces a linear or non-linear effect. Here is a typical graph I produce:
The X-axis in this graph represents my Strength values as I changed them from 270-350. The Y-axis represents the AVG damage output that I did on a L50 Water Sprite. As the chart title implies, this data was taken with a Weathered Spear on DRG, while holding DTR constant at 202. I then plotted a best-fit line or a Trendline in Excel. This line tries to provide a line that fits as close to as many data points as possible. The important information that I am looking for is seen with the equation located on the right-hand side of the chart. Let me break this down:
y = 0.1018x + 6.8066
R^2 = 0.9961
The 0.1018 value is the slope of the graph. When taking the units of the X and Y axis into account, the slope can be best identified as (+0.1018 difference in damage)/(1 addition of STR). By using some mathematical terminology we can change this into saying the slope is equal to (+.1018 DMG/dSTR). In general terms the slope will simply be (dDMG/dSTR). The “d” simply means the difference between two values. So there will be a difference of +.1018 Damage per 1 point of difference in Strength.
The 6.866 is the X-intercept. It refers to how much damage you will do at 0 STR. This value does not have a lot of meaning, but I will expand upon it later.
The R^2 value is of real importance. In simple terms, this is a statistical value that helps determine the “linearity” of a line. An R^2 value can range from 0<R^2<1.0. As R^2 approaches 1.0, the best-fit line for the data is best described as a linear line. As R^2 approaches 0, this indicates that the best-fit line is most likely not linear in nature. The vast majority of my data have R^2 values around 0.9900-1.0000, indicating a high probability that a linear equation best fits the data.
Now comes the abstract thinking process in deriving a working damage equation. My thought process first wanted to determine if there was any intrinsic relationship between WD, STR, and DTR. In other words, did each variable have an independent effect, or did they influence each other? To test this, I obtained 3 special graphs:
I tested the rate of return for Weapon Damage (dDMG/dWD) for 2 different set STR and DTR values: 325 STR|202 DTR and 400 STR|202DTR. I also “calculated” a 200 STR|202 DTR value by using other data points. I then plotted the slopes of these lines and ended up with the following graph:
Now what exactly is this graph? It is the influence of increasing Strength while keeping DTR constant and seeing if the rate of return of Weapon Damage on Average Damage output changes. If this graph had a zero slope, this would indicate that STR has no influence on Weapon Damage and would thus imply that both are independent of each other. However, this is not the case. The higher your STR value, the more influence Weapon Damage will have on your Average Damage output. This is an important concept to grasp when trying to derive a damage equation. Instead of saying WD + STR = damage, the equation will most likely have a WD*STR term. As we will see later, this turns out to be the case.
So now in our minds, we have a mental picture that STR and WD influence each other. But now we need to ask ourselves: What if we had 0 weapon damage; wouldn’t changing STR still increase damage? The answer is of course yes, and leads us to a very important part of the damage equation. To get to this answer, I plotted the Y-intercept value for the 3 graphs I mentioned earlier:
These are the 32.074, 19.72, and 39.778 values for 325, 200, and 400 STR respectively. This produced the following graph:
This graph represents how Strength influences Average Damage at constant DTR, independent of Weapon Damage. So if I had 0 WD and 100 STR, I would still be doing damage.
Taking these two graphs into account, we can start formulating the backbone of the damage equation:
(0.0032 x STR + 0.4162). This is the equation to determine at a specific STR value, what the influence of WD is.
(0.1001 x STR – 0.3529). This is the equation to account for “base” damage independent of WD.
DMG = (0.0032 x STR + 0.4162) x WD + (0.1001 x STR – 0.3529).
Now we are not quite there yet. This equation will only be accurate for a constant DTR value of 202. I have not devised a method to incorporate DTR into a true working damage formula. Instead, I approximated DTR’s influence with a “correction factor.” While this will introduce a bit of error into the calculation, it will give a good approximation since the influence of DTR is, as currently tested, relatively small. However, there is a caveat to this. The influence of DTR appears to be different on Auto-Attack versus other Weaponskills. This is eerily similar to how the secondary stats influenced Auto-Attack in 1.0. Following a similar manner as my earlier testing by keeping WD and STR constant while varying DTR, I came up with the following correction factors:
Auto-Attack: + (DTR-202)x(0.11)
Weaponskill: + (DTR-202)x(0.035)
So in spectacular fashion, we can add either of these terms to the end of the previous formula to get:
AVG WS DMG = (Potency/100) x [(0.0032 x STR + 0.4162) x WD + (0.1001 x STR = 0.3529) + (DTR – 202) x 0.035]
It is important to note that this Determination correction factor is not 100% accurate. DTR could very well be dependent on WD and/or even STR. However, if you use primal/relic weapons with relatively high STR, I think these correction values will provide accurate enough estimations to be able to predict your average damage output.
Now if you are paying attention, you noticed I added in another term that I have yet not mentioned: (potency/100). All the testing I did with Weaponskills involved using 100 potency Weaponskills such as Vorpal Thrust (no combo). Through a similar process as the previous testing, I have established that potency is a linear modifier. That means, a 200 potency WS will do 2x the average damage of a 100 potency WS. So the (potency/100) term is to reflect this influence on damage output.
In a similar fashion, we must determine how the “potency” of Auto-Attack works. If you look at a weapon, you will notice the “Auto-Attack” stat. For example: With the Mogfork, the Weapon Damage is 41 and the Weapon’s Auto-Attack stat is 40.45. To calculate the Auto-Attack potency, simply divide WAA/WD or 40.45/41. This comes out to be 0.9866, and if multiplied by 100, comes out as 98.66 potency. This is the exact same “potency” stat as with weaponskills. For comparison, here is a DRG with a Mogfork doing Auto-Attack, True Thrust, and Vorpal Thrust:
Within a certain degree of error, the Auto-Attack damage appears to be about 98.66% the damage of Vorpal Thrust, a 100 potency WS.
So with this in mind, let us formally write the base Auto-Attack equation for non-ranged classes:
So far, all I have really said is that with a L50 DRG on a L50 Water Sprite, using Primal/Relic weapons, you can calculate your damage output with about 99% accuracy. This has strong internal validity. In order to see if this is applicable to other monsters, other level monsters, and other levels, I had to do a little extra testing. Luckily, I was able to do most of this in short manner.
I first tested on L40 Basilisks to see how the damage output changed. Surprisingly, I had the EXACT same min/max with the same STR/DTR/WD as on a L50 Water Sprite. I tested this again on L1 Star Marmots, and I had the same result. This seems to imply one of two things: Either all monsters have no defense, or all monsters have the same defense value. More importantly, it seems to imply that enemy level has no influence on damage output. This is in stark contrast to 1.0, where enemy level was the most important value in determining damage output.
As for player level influencing the equation, sadly the formula listed above does not accurately predict damage output. This was to be expected as the formula is more of a “working” damage formula rather than a “true” damage formula. The true formula most likely has exact constants/ratios while mine is more of a best-fit formula. As such, the level of the player potentially has one or more “hidden” attributes that influences damage output. That or some kind of tier system. Either way, it is not worth testing currently as all content will be done at L50.
───Introducing Critical Hit Rate───
There is yet one last damage output modifier that is tied to your attributes. That would be the Critical Hit Rate (CRT) modifier. Critical damage is an especially tedious attribute to test for since 1000s of hits are required to get any kind of meaningful data. To test CRT, I used a L50 PLD with Sword Oath. The current tooltip for Sword Oath is wrong. The ability actually adds an additional 25 potency auto-attack with your current auto-attack. So you are doing 2 auto-attacks at the same time, just with different potencies. Since I only cared about the crit rate and not the damage itself, this should be ok. Also, Sword Oath attacks have the capability to crit. As a result, either or both of the PLD auto-attacks can crit, so the data should not be adulterated in any manner. Furthermore, PLD weapons have really low auto-attack delays, so in essence, I was able to more than double my testing rate as compared to a class like DRG. Itis important to note that the Critical Hit Bonus appears to be a universal 150% modifier to your damage output, regardless of enemy level.
With this in mind, I set out to obtain between 1,500-2,000 hits on L50 Water Sprites to ensure that my data would be relevant on endgame level monsters. I obtained 4 values and plotted them:
I also took some crit data that a user named Doctor Mog made public and integrated it:
The fact that an independent source of crit data fit so well with my own data is strong evidence towards the accuracy of the formula:
Crit Chance % = 0.0693 x CRT – 18.486
While Doctor Mog’s data is most likely accurate, I did not want to alter the results and conclusions from my own data.To incorporate this into the final average damage formula, we will need to add the following term:
AVG DMG (including crits) = DMG Formula * (1 + 0.5 x (0.0693 x CRT – 18.486)/100)
The (1 + 0.5 x …) term is just to go along with the distributive property of mathematics to ensure you are doing at least base damage without crit. The 0.5 modifier is to represent the +50% universal bonus for crits. The /100 term is to convert from a percent into a decimal for formula calculation reasons.
With everything discussed in mind, we can finally write the damage formulas accounting for every player attribute. These formulas are for overall damage, as in if you did 1000 of the same attack, you’re average damage done would be predicted by this formula. There will be 2 formulas, one for AA and one for WS. Also, depending on which class/job you play, there will be about a +/- 2 modifier. The formula will produce numbers that are consistently 1-2 points off. I am not sure why, but since the error is a consistent systematic error and not a random error, this should not be much of a concern.
AVG DRG AA = (WAA/WD) x [((0.0032 x STR + 0.4162) x WD) + (0.1 x STR – 0.3529) + ((DTR – 202) x 0.11))] x (1 + (0.5 x (0.0693 x CRT – 18.486)/100))
AVG DRG WS = (Potency/100) x [((0.0032 x STR + 0.4162) x WD) + (0.1 x STR – 0.3529) + ((DTR – 202) x 0.035))] x (1 + (0.5 x (0.0693 x CRT – 18.486)/100))
There are primarily 3 attributes that reduce true damage. These are Physical Defense, Magical Defense, and Elemental Defense. The first step in testing these values is trying to ascertain the min/max of damage taken and if enemies follow a similar damage algorithm to player damage. As will be seen in the following sections, the +/- 5% from the average still applies to enemy damage.
I started the Magical Defense (MDEF) testing by using L50 Water Sprites whose primary attack is a magic spell called Water Attack. I started out naked on a L50 PLD with 0 PDEF, 0 MDEF, and 261 VIT. I also made sure not to equip any Water Resistance gear. I then first tested to ensure that VIT did not influence damage reduction. This was tested by job switching to GLA with 241 VIT. Changing VIT by 20 points did not alter damage taken at all. After assuming that VIT did not influence damage reduction, I tried on various gear pieces without regards to any primary attribute such as STR, DEX, VIT, INT, MND, or PIE. Here is the table and 2 graphs I obtained from the L50 Water Sprites:
The table should be self-explanatory. The red values are just the difference in baseline values. Since you can get 0 PDEF and MDEF, these red values are the same as the black values. For VIT, since 261 is the base VIT on L50 PLD, the red values become more meaningful. This is more an artifact from my 1.0 testing, but I already have these images uploaded so I haven’t had time to change them.
With that out of the way, look at the two bottom graphs. I plotted the actual Average Damage taken versus MDEF. This results in a very linear line. Originally I thought this would be just like the player damage formula, but I decided to go a step further. The second graph is a representation of the percent reduction in damage per point of MDEF. So we are looking at 1 MDEF reducing damage taken by 0.0445%, with 100% damage taken at baseline 0 MDEF. The importance of this will become apparent in the next paragraph.
To assess the influence of enemy level, I tested with a similar methodology on L28 Water Sprites:
Notice several things about this image. First, the actual Average Damage v MDEF graph has a different slope as compared to the L50 data. However, notice the Percent Magic Damage reduction graph’s slope. It is almost identical compared to the L50 chart. This seems to imply that MDEF provides a universal percentage reduction in damage that is independent of monster level. Note: In version 1.0, MDEF was Magic Evasion and offered partial resistances. This is no longer the case in 2.0; you either evade completely, or you get hit and take linearly reduced damage with MDEF.
With the process of MDEF testing in mind, I went about testing Physical Defense (PDEF) using L40 Basilisks as my testing monster. I recorded their auto-attack damage and how VIT and PDEF influenced damage taken. Once again, VIT appears to not influence damage mitigation like it did in version 1.0:
Again, similar to the results of the MDEF testing, I obtained a 0.00044 percent reduction in physical damage taken. Kaeko Leta, a linkshell mate and known tester, has also done some PDEF testing independently and obtained very similar results:
By taking the data from both MDEF and PDEF as a whole, it strongly suggests that both follow the same linear percent reduction of 0.044% damage taken per point of PDEF or MDEF. Both also strongly suggest that enemy level is independent of player damage mitigation. This means that this data will potentially be very pertinent to endgame bosses as it should be easy to calculate how much you can potentially reduce endgame attack damage. The formula is simply:
Percent Physical Damage Taken = 1 – (0.044 x PDEF)
Percent Magical Damage Taken = 1 – (0.044 x MDEF)
Kaeko made a good table for easy viewing:
Removed this section for now since I really don’t have meaninful data and some people are misinterpreting it. I will update it later.
───Effective HP Part 1───
Ultimately the knowledge we have gained to this point has been in search of answering one question: Which stats should I invest in to increase my ability to survive? In an attempt to provide a meaningful answer to this question, we must define a new term. This term is Effective Hit Points (EHP). We may define this as the amount of true damage that a player can sustain, where true damage is the damage a player sustains with zero damage mitigation. Let us perform an academic exercise to help clarify:
Miko the PLD has 2500 HP and 0% damage mitigation. Bahamut hits Miko for 2500 damage and Miko dies.
Rhys the PLD has 2500 HP and 50% damage mitigation. Bahamut hits Rhys for 1250 damage. Rhys has 1250 life. Bahamut hits Rhys for another 1250 damage. Rhys dies.
Both Miko and Rhys had the same HP; however, Rhys took twice as many hits from Bahamut. Alternatively, Rhys sustained twice as much true damage as Miko and therefore had twice as many Effective Hit Points as Miko. In mathematical terms we can write this as follows:
EHP = HP x (100/(100 – % damage reduction))
With the knowledge that each additional Vitality point added is the equivalent of +15 HP, we can alter the formula to be:
EHP = (Base HP + 15 x dVIT) x (100/(100 – % damage reduction))
A L50 PLD will have roughly 245 VIT and 2400 HP baseline. A L50 WAR will have roughly 245 VIT and 2465 HP at baseline. With this in mind we can make generalized formulas of:
PLD EHP = (2400 HP + 15 x (VIT – 245)) x (100/(100 – % damage reduction)
WAR EHP = (2465 HP + 15 x (VIT – 245)) x (100/(100 – % damage reduction)
Block and Parry
Block and Parry are two special damage mitigation actions that proc a % damage reduction as seen in your battle log. Even though these are considered damage mitigation properties, I decided block and parry were important enough to have their own chapter. Per the Beta Phase 3 instruction manual:
Block: Damage mitigation is influenced by Strength (STR) and Shield Block (SB). Rate of block influenced by Dexterity (DEX) and Block Rate (BR).
There are only 3 classes/jobs that are currently able to equip shields and block. They are GLA/PLD, THM/BLM, and CON/WHM. For the block testing in this section, all testing was done on a L50 PLD unless otherwise specified. In order to test blocking and its modifiers, I selected a physical damage dealing mob since I do not believe you could block magical damage in 1.20, and suspected it to be the case in 2.0. I have yet to find a monster that uses magical attacks that I am able to block in 2.0. With this in mind, I chose L40 Basilisks and recorded only their auto-attack and not their TP moves. I also made sure that I faced the monster at all times since I believe direction influenced blocking in 1.20. Here is my raw data:
Much of this table will be familiar to viewers who read through the damage mechanics methodology. There are a few new terms I will address:
Block Rate Ratio: This refers to how many unblocked hits I took as compared to blocked hits. A value of 171/14 for example means that I took 171 hits, 14 of which were blocked. This was all done by hand, so there might be a higher margin for error than most of my testing. Also, I did not counted parried hits into my total.
Block Rate %: This refers to the actual percent from the ratio listed to the left of the Block Rate % in the table above. For example, 7/124 = 0.0564 or 5.64%.
Block Min/Max: These values refer to my new Min and Max damage taken while blocking.
% Reduction: These values refer to the actual (%) value that appears in your battle log when you successfully block. They appear to always be whole numbers.
First let me go over the weaknesses of this data. I only collected about 100-170 data points per stat/shield tested. Also, I did not test over that wide a range of stat values, simply large bulk changes. This data collection was more-so to get an overall idea behind the mechanics of blocking. However, even with this rudimentary data, we can already make several observations:
1) STR appears to have very little influence on damage mitigation from blocking.
2) The base block rate is most likely somewhere around 5%, with you expecting about a 14-15% rate of block with 188 DEX with high Block Rate shields. With a Relic shield and high DEX, we might expect to see this value approach 18-20%.
3) Equipping a shield does not offer you an intrinsic damage reduction. You will still have the same Min/Max damage received unless you proc a block.
4) The % damage reduction from blocking appears to be very small for low Shield Block values.The floor for damage reduction is probably 4 or 5%.
This was about the extent of the testing I did on blocking. However, a player by the name of Khagen Dragnir shared with me his data, and from it I was able to make much more concrete observations. You can find his spreadsheet here:
When using other players testing, I can only make assumptions on his part to the fidelity of what he tested on and what he held constant. However, with quick referencing from the tanks in my LS, I have confirmed that much of his data appears to be accurate. First, let us look at his table comparing various values of STR and the percentage damage mitigation from blocking:
Looking at the left-hand column, we can begin to have a clearer picture on how STR influences block damage mitigation. Unlike many of the other influences STR has shown in my data testing, STR seems to have a tiered scaling effect with regards to blocking. This means that for specific STR ranges (X<STR<Y), you will experience the same benefit in damage mitigation for blocking. Taking this a step further, you can track the columns to the right and notice that this trend holds true for varying Shield Block values. So regardless of what shield you have equipped, you can expect the same tiered increase in damage mitigation from changing your STR attribute. These target values appear to be 270, 337, and 404 STR. They follow a +67 point trend that I have not been able to confirm at higher values due to lack of gear. There is one last great piece of information we can gain from Khagen’s block data. By comparing the percent damage mitigation versus Shield Block, we can make a graph plotting these values. We get:
This graph represents how changing Shield Block will increase your damage mitigation from blocking with STR held constant at 337. The equation at the top of the graph can be described as follows:
0.000583: This is the slope or the point-per-point return in % damage mitigation from adding 1 Shield Block.
X: This is the variable where you would plug in your Shield Block value.
0.063779: This is the Y-intercept of the graph or what you would expect the floor for blocking to be if you could equip a shield with a Shield Block value of 0. This was predicted to be (4 or 5% + 2%) or between 6% and 7% . As mentioned earlier, 4% or 5% appears to be the blocking floor. The 2% comes from the +2% bonus from being in the 337 STR tier which all values in this graph were at.
The high degree of linearity presented in this graph is not surprising. Virtually every other attribute I have tested expressed linearity in its influence. The importance of this graph is that now we can predict, independently of STR, what the Shield Block value will produce in terms of percentage damage mitigation. If you would like to see the culmination of Shield Block and STR, head to the Conclusions section below and see the final table listed.
Parry appears to be a very similar damage mitigation action compared to blocking. There appears to be a flat percentage damage reduction, the reduction is rounded to the nearest whole number, and parrying does not seem to proc on magic attacks. There are only two noteworthy differences:
1) Every class can parry
2) Damage mitigation from parrying is only influenced by STR. All other stats (PAR and DEX) only influence the rate of parrying.
Once again turning to Khagen Dragnir’s data, we can notice several trends. First, I would like the readers to take a look at this graph:
There are several things we can take away from this graph:
1) 10% seems to be the floor damage reduction from parrying.
2) STR seems to follow a tiered system similar to blocking.
3) The STR tier seems to follow a 40/41/40 trend.
4) The tiered system seems to be similar for all tank jobs/classes and most likely for every class/job.
I confirmed his STR values and the percent reductions for up to 405 STR. The STR tier thresholds for parrying are 243, 283, 324, 364, and 405. Following the 40/41 pattern, the next predicted thresholds would be at 445 and 486. With points 1, 2 and 3 in mind, we can somewhat predict that parrying will realistically never cross the 16% damage reduction threshold. To do so would require an STR value of 486≤STR<526. While we have yet to see the new endgame gear, this seems like a somewhat safe assumption for now. As for testing the rate of parrying, I did not approach this subject at all during my beta testing. If there is a player who recorded the influence of DEX and PAR on parrying, I would be very grateful to see your data.
───Chicken Or The Egg───
Now certain questions arise: Which action occurs first, blocking or parrying? And can you parry and block attacks at the same time? In regards to the second question, while I have not thoroughly ruled out the possibility, in all the testing I have done and all the Beta experience my tanking friends have given me, none of us have ever noticed a simultaneous proc of parry and block. Furthermore, one of the more tech savvy members of my LS tells me that the battle log does not allow for a simultaneous proc of block and parry. It is my current opinion that until I have seen evidence to the contrary, blocking and parrying are mutually exclusive events. In regards to the first question, I do not have a concrete answer at this time. That being said, I recently thought of a method to test this, but I simply did not have enough free time during the Beta to test this. My thought process is as follows:
First get a baseline for the rate of parry without a shield. Then, get the highest DEX and Block Rate shield you can have, use Bulwark, and only collect data points while Bulwark is active. Theoretically, you should have around an 80% rate of blocking. Compare the frequency of parrying from having no shield equipped to the frequency of parrying about an 80% rate of blocking. If the frequencies are similar, then you can essentially deduce that the rate of parrying is not affected by the rate of blocking and is thus the action proc that the game checks for first. If the rate of parrying is significantly lowered, then you can deduce that blocking is checked for first. With Bulwark being on a 3 minute cooldown and a 15 second duration, this test would be very time intensive. Alternatively you could do this test without using Bulwark; however, a significantly larger amount of data points would be needed to deduce the difference. Yet another option you could do is to test with Parry on GLD using the LNC skill Keen Flurry. This ability only has a 90s cooldown.
With this set aside, from a purely logical perspective, it would make sense for the block to be checked for first. Since the main purpose of a shield is for Paladin/Gladiator and the damage reduction from an endgame shield is much better than the damage reduction from parrying, it would make sense for block to be checked for first. This is of course assuming that blocking and parrying are still mutually exclusive.
Now, readers at this point might be asking why I am spending so much time dedicated towards answering these questions. Consider if Parry is checked for first. Assume I have 100% chance of parrying 15% damage and 100% chance of blocking 25% damage. Since the action of parrying will always override the chance of blocking, my damage reduction would actually be decreased even though I maxed out the rates for both stats! Depending on how the numbers work out, there could be a similar situation in which adding the PAR attribute to your equipment could actually lead to a detrimental increase in damage taken. The potential for this would be especially high during Bulwark use when a parry might override a near guaranteed block. Once again, this is probably a small possibility, and if block takes precedence as the first action the game checks for then the point is fairly moot. However, until proven otherwise, it is still a worthwhile academic exercise to consider.
Taken all together, we can start formulating an idea of how to best represent damage mitigation from both Parry and Block. Instead of forming an equation as seen in previous examples, I believe the best way to represent the damage mitigation is in the form of a table due to the tiered nature of the STR attribute. To do so, I listed all the threshold values of STR and tabulated Shield Block in increments of 50:
The first thing readers will probably notice is the “0″ column for Shield Block Value. The numbers are to reflect what I believe is the true floor value. This column could very well be 1% point lower, but in truth, there is very little merit to debating this point since a 0 Shield Block attribute does not exist in the game. The other columns should be fairly self-explanatory. Simply match up your current STR number and your Shield Block Value number and that should be within a +/- 1 % of what you can expect to block. The reason I am allowing for this +/- 1% error is that the FFXIV rounding function seems to be very unpredictable in its function from my experience. So if you have 420 STR and a Shield Block of around 275, you can expect to have a damage mitigation somewhere around the 23-24% range. Once again the power of this table is not in having 100% accuracy but in the power of it to offer a good approximation of the damage mitigation potential of increasing the Shield Block attribute. The values in this table were obtained from the formula mentioned above in the “Block” section of this chapter. Another very important feature to notice is how close the 404 and 405 STR tier thresholds for blocking and parrying are. For tanks, 405 STR might be the ideal number to aim for in terms of damage mitigation depending on what other potential stats you are giving up in the process. If you can not reach the 405 STR threshold, then in terms of damage mitigation, there is no reason to have your STR attribute above 364.
Finally, I would just like to reaffirm that I did not do any pertinent amount of DEX, Block Rate, or Parry data. If any player has data on these, please leave a comment here or message me on reddit, reddit.com/u/valkky.
Post Beta phase 3 I was made aware of this data testing a lodestone user by the name of Hulan made public:
This chart is a little disorganized, but I still think we can take something away from the Parry %. Just by increasing Parry by 81 points, Hulan added about 6.1% to his parry rate. That’s about 0.076% added per point of Parry. We’re talking about 20-21% parry rate with endgame gear aimed towards maxing Parry. Without more accurate data, I’m afraid that’s all we can really take from this chart.
Healing Mechanics And Formula
Now that we have sufficiently covered the basics in damage dealing and damage mitigation, it is time to move on to the third leg of the gameplay tripod: Healing. As with many stats in Beta Phase 3, the Beta instruction manual offers us a bit of insight into how the core mechanics of Healing work. Per the manual there are 2 attributes that influence Healing: Mind (MND) and Determination (DTR). However, as will become quickly apparent to any player who plays on WHM, there is a third, more important attribute that influences Healing: your weapon’s Magic Damage (MD). Let us now broach the subject of how to go about understanding these mechanisms at play.
───Three Is Greater Than Five───
In the Damage Mechanics And Formula section, we approached a methodology to establish the random number generator used in damage algorithms. Similar to those principles, Healing also follows a non-normal random distribution. However, there is a difference. The key difference with Healing is that the +/- deviation from the average is only 3% as compared to 5%. So we can write:
Minimum Heal = Average Heal x 0.97
Maximum Heal = Average Heal x 1.03
Now. the natural question to ask is why have a different, smaller degree of deviation for Healing? The answer may be seen with the interaction of end-game bosses and tanks. Tanks are going to potentially be taking enormous percent HP deductions from end-game bosses in short time periods. Therefore, Healing abilities are naturally numerically higher than damage abilities. To prevent frustration, it is important to have a consistent and dependable amount of HP healed; otherwise, players could have the illusion that they are failing due to unfair mechanisms. This is the same principle why crit healing builds on WHMs are undesirable for many players.
───Pro-curing The Data───
Heal testing is unique in the aspect that you do not need any other player or NPC in order to obtain data. Even at full HP, the heal value still displays the same value as when at low HP. With this in mind, I used self-Cure I’s to form the basis for much of my data. Additionally, I was worried that WHM and CON might have some inherent healing boosting property so I did all of my heal testing on L50 THM. I followed a similar process to my damage mechanics testing in that I looked for the +/- 3% variation, and when reasonably close, I assumed I had the correct Min/Max for Cure I at a specific stat-line. I accounted for three variables influencing Healing and made sure to hold two of the values constant while I increased/decreased the third variable. These variables were MND, DTR, and MD. I did not record STR, DEX, VIT, INT, PIE, or any other stat. The highly linear results you will see in the subsequent sections are strong evidence towards proving that there are only three variables that influence Healing as listed above. Here are the results for one specific stat-line:
If you do not understand the terminology used, reference the damage mitigation section here (LINK). The “X” listed under INT just means that I did not record the number and it is unknown.
When graphed, we obtain the following picture:
I also included a L50 WHM with comparable stats/weapon for comparison. Once again, note the highly linear nature of the influence of MND on Cure I as seen with nearly every other stat mechanism. To reiterate what this means: For every 1 point of MND that I add to a L50 THM equipped with a 61 MD weapon at 202 DTR, you will increase your HP/Cure I by 0.9278 points. There appeared to be no diminishing or exponential returns and no threshold/tier values to aim for.
With the process of how we developed the damage formula in mind, we can use a similar process in trying to develop a Healing formula. This involves first addressing the question: Are MND, DTR, and MD independent or are they intrinsically related? We saw back with the damage formula that at least STR and WD were partially dependently related. In order to ascertain this for Healing, I obtained five different average Cure I heals for constant MND and DTR while varying MD. I then repeated this for thee different MND values keeping DTR constant.This produced the following 3 graphs:
To my surprise, I had R^2 values of exactly 1 for all three graphs. This strongly suggests that for a set MND/DTR, you will experience linear returns in the amount HP healed per Cure I from increasing you weapon’s Magic Damage. Also note how the slope of the three graphs changes fairly substantially in what appears to be a proportional relationship to MND. This strongly suggests that MND and MD are not independent variables but have some kind of dependent relationship. This is important when trying to develop a working Healing formula. This indicates that there is most likely some kind of non-additive/subtractive relationship such as (MD*MND). Now, let us plot the three slopes from the graphs above: 3.0000, 3.3382, and 3.6478 versus their respective MND values: 204, 234, and 260
If this graph had a zero slope, this would indicate that MND has no influence on Magic Damage’s rate of return on HP healed from Cure I and would thus imply that both are independent of each other. However, this is not the case. The higher your MND value, the more influence Magic Damage will have on your Healing output.The high R^2 value strongly suggests that MND produces a linear influence on the rate of return of MD on HP gained from Cure I. Furthermore, this implies the existence of a linear (MND x MD) term in the Healing formula. Thus we can take the first steps towards formulating our working healing formula:
HP Healed = (0.01156 x MND + 0.6390) x MD + …
Now, we still have to deal with the Y-intercept values from the three graphs I mentioned earlier:
These are 59.5, 66.224, and 71.09 for 204, 234, and 260 MND respectively. These are essentially the “base” healing amounts if you had 0 Magic Damage. As you might have guessed, these values will increase as MND increases. This implies that there is a distinct term in the Healing formula that involves a (Unknown + (MND x Rate)) term. To help clarify what I mean, let us graph these Y-intercept values:
What is this graph telling us? Independent of your weapons Magic Damage, Mind will produce a linear increase in HP healed per Cure I. So no matter what the value of your Magic Damage is, adding 1 point of MND will add at least a 0.2229 HP per Cure I. While this doesn’t appear that relevant at first, it allows us to make an important observation. As mentioned previously, this implies we can add the following term to our Healing formula:
+ (0.2229 x MND + 14.042)
HP Healed = (0.01156 x MND + 0.6390) x WD + (0.2229 x MND + 14.042)
So is this our working Healing formula? The answer is both yes and no. For a constant DTR value of 202, this formula will provide accurate predictions for HP healed from Cure I on an L50 THM. To produce a more useful formula, we must try to account for the third variable DTR. Now once again, I am not a great mathematician or abstract thinker. The best solution I could come up with in short order was the use of a “correction factor.” To do this, I followed a similar process as earlier, plotting 5 data points at a set MND and DTR while varying Magic Damage. This produced the following two graphs:
This is where things get somewhat dicey. The slope value did increase as I increased DTR from 202 to 230. However, this was only about a 1% change and could very well be attributed to rounding error. Therefore, based on this graph, I am not able to claim with 100% certainty whether DTR and MD have an independent or dependent relationship. However, we are still able to make use of this data in the hope of approximating DTR’s influence on Healing. Let us instead make a new graph representing the net increase in HP healed from Cure I by increasing DTR by 28 (202->230) versus the Magic Damage. We get the following graph:
At first, this may seem like a very specific and abstract graph that offers little help in perfecting the Healing formula. However let us try to understand what the equation in this graph represents.
y1 = 0.0368x + 1.7009
For every 1 point of Magic Damage added, you can expect your HP healed when increasing DTR by 28 to increase by 0.0368 points. Let us factor out the 28 DTR. To do so, simply divide every term by 28. This results in:
y2 = 0.001314x + 0.06071
Now why did the y term’s subscript change? This is simply to reflect the influence of the 28 division. Writing y1/28 is just an artifact and holds no real meaning. It is the slope and Y-intercept that hold the real importance. If you would like a better explanation, e-mail me and I will try to clarify what I did. Moving on, let us now incorporate this term into our working Healing formula:
+ (0.001314 x MD + 0.06071) x (DTR-202)
Where did (DTR-202) come from, and why is it not simply (… x DTR)? Remember when I went about calculating this correction factor, I was calculating/graphing the 28 point increase from 202 DTR, not from 0 DTR. Also notice one very important thing: DTR and MD have a dependent relationship! This means that both DTR and MND influence the rate of return of MD when calculating HP healed. Now with this final term in mind let us write the working THM Healing formula:
L50 THM AVG HP Healed from Cure I =
(0.01156 x MND + 0.6390) x MD + (0.2229 x MND + 14.042) + (0.001314 x MD + 0.06071) x (DTR – 202)
If we do some rearranging we can write:
(MD) x (0.01156 x MND + 0.001314 x (DTR-202) + 0.6390) + (0.2229 x MND) + (0.06071 x (DTR – 202)) + 14.042
Further arranging yields:
(MD) x (0.01156 x MND + (0.001314 x DTR) – (0.001314 x 202) + 0.6390) + (0.2229 x MND) + (0.06071 x DTR) – (0.06071 – 202) + 14.042
Simplifying results in:
(MD) x (0.01156 x MND + 0.001314 x DTR + 0.3736) + (0.2229 x MND) + (0.06071 x DTR) + 1.7786
Looking at the equation in this form makes a lot of sense based on the data we have studied. Let us break down each section again just to reiterate and explain the thought process in their derivation.
(MD) x (0.01156 x MND + 0.001314 x DTR + 0.3736)
This term represents the dependent relationship between Magic Damage, Mind, and Determination. As the weapon’s Magic Damage value increases, the influence of MND and DTR increases. In a similar fashion, as the MND and DTR increases, the influence of Magic Damage on the healing equation also increases. This makes sense as we saw the rate of return of Magic Damage have an increasing slope as we increased MND or DTR.
(0.2229 x MND)
This term represents the independent influence of MND on Healing if your weapon’s Magic Damage is 0.
(0.0607 x DTR)
This term represents the independent influence of DTR on Healing if you weapon’s Magic Damage is 0.
This appears to be an artifact of the process of developing the formula due to the Y-intercept values being used in the derivation process. There is perhaps a better explanation for this term. As we have seen with Magic Damage, there seems to be a dependent relationship between the variables. This 1.7786 term could very well be representative of a MND x DTR term and the other constants in the formula might need to be adjusted slightly. As we will see in the upcoming White Mage (WHM) reconciliation section, there is evidence towards this being the case.
As the astute reader is aware, we have not yet approached the subject of healing potency. Let us make a brief mention here that to modify the Cure I healing formula, we must add a strict external linear modifier to the equation. This will term will also need to correct for the base 300 potency from Cure I used in the formula’s derivation. In layman’s terms, we shall add a (potency/300) term:
───WHM Reconciliation Facilitation───
Let us preface this section by first warning readers that the majority of the discussion in this section will be based on data that I did not collect myself. I collected most of my data on THM and regrettably did not choose to pursue testing on WHM. This turned out to be a big oversight, and I have had to rely on external data sources to come to the conclusions that I will present in this section. The bulk of this data reconciliation process will be based on WHM data made publicly available by the Blue Garter (BG) LS and the Daeva of War (DW) LS.
With our recently derived THM healing formula in mind, it is only natural to see if this formula is accurate for the WHM job. Based on DW’s public WHM data, I found that my formula seemed to skew the predicted Average Cure I by about 9-11 points from the actual Average Cure I value. If we take a primitive approach and try to correct for this discrepancy by adding a flat (+10) to the THM equation, we get surprisingly accurate results. However, once MND and DTR begin to rise to much higher values, it becomes clear that the formula’s accuracy deteriorates. This implies one of three things:
1) The WHM formula has different values/constants than the THM formula.
2) There is a term in the equation that I am not accounting for.
3) The THM formula is wrong.
Initially in the process of reconciliation, I opted for option 1. I tried dozens if not hundreds of minor alterations in numerical values listed in the THM formula to try and narrow the gap. Upon closer observation it becomes noticeable that DTR seems to have more of an influence with WHM than it seems to have with THM. The THM formula appears to underestimate the HP healed as DTR increases. To account for this discrepancy, naturally we would consider increasing the DTR constants/modifiers in the THM equation. However, in doing so, the HP at low DTR values becomes overestimated. At this point there are two options to correct for this. We can either start changing the MND modifier values or we can consider an alternative term that involves DTR as a modifier. Let us proceed with former option first:
L50 WHM HP Healed = (potency/300) x [(MD x (0.01123 x MND + 0.0021 x DTR + 0.28) + 0.242 x MND + 0.1 x DTR
This adjusted formula increases the influence of DTR from Magic Damage while also eliminating the external constant we had earlier. From comparing other player's WHM data, this formula appears to have 99.5%+ accuracy for almost every data range.
Let us pause here and proceed with the latter reconciliation method mentioned earlier. We will now attempt to instead introduce a second DTR term to reconcile the formula.
The new DTR term needs to be distinct from the simple constant x DTR term. Otherwise we could simply add this to the existing 0.0607 x DTR term that already exists. Additionally, we are aware of the existence of MD x DTR and MD x MND terms. With this in mind, let us declare the existence of a MND x DTR term as a possible solution to reconciling the WHM formula.
With the notion of a (+ DTR x MND) term as a distinct possibility, we first must establish the plausibility of its existence. There are three pieces of evidence that suggest such a term's existence. First consider the abstract 1.7786 term from the THM formula. It is HIGHLY unlikely that this term exists as a constant in the true healing equation. The second piece of evidence comes from the fact that the WHM predicted values are off by a consistent 9-11 points. This suggests the existence of another additive term in the formula. The third piece of evidence comes from when the graph the dHP/dMND graphs for two different set DTR values via DW's WHM data:
The slope has a marginal increase from 0.9462 to 0.9552. It is hard to discern whether this is a true increase or simply the variation due to rounding error. With so many data points graphed, however, it seems plausible that this is a true difference rather than just error. Just for comparison, on THM at 202 DTR, the slope is 0.9278:
While concrete evidence is not available, proceeding with the notion that a MND x DTR term exists leads to the question of how do we determine the influence of such a term. As a rough means of approximation, I divided the 11.7768 (1.7768 and the +10 correction factor) term by 204 and 202 to represent baseline MND and DTR respectively. This came out to be 0.000286. Thus the initial equation I worked with for WHM became:
(Pot/300) x ((MD) x (0.01156 x MND + 0.001314 x DTR + 0.3736) + (0.2229 x MND) + (0.06071 x DTR) + (0.000286 x MND x DTR)
This formula, however, is not accurate. It will overestimate HP healed for MND and DTR values over 204 and 202 respectively. As a note, all reconciliation to the WHM Healing formula made here comes from tinkering with the constants rather than from hard mathematical derivations. So do not be alarmed if numbers appear to arbitrarily come from no where. With the goal in mind of decreasing MND's influence while increasing DTR's influence to account for the data I reviewed from BG and DW's public forums, this led me to the following formula:
Average HP Healed On L50 WHM = (Potency/300) x [((0.0114xMND + 0.00145xDTR + 0.3736) x Magic Damage) + (0.21xMND) + (0.11xDTR) + (0.00011316xMNDxDTR)]
Now I would like to reiterate that the final term I added is mostly theoretical and even at this moment I am leaning towards it not being a real term in the true Healing formula. With that said, this formula appears to predict with over 99% accuracy the correct Average HP healed values:
How robust is this formula? The range for MND in the above chart varied from 272 to 390. The range for DTR in the above chart varied from 202 to 280. The range for Magic Damage in the above chart varied from 7 to 65. In every combination listed, the accuracy of the predicted value compared to the actual value exceeded 99%. With no apparent trends in diminishing accuracy as seen earlier with the THM formula extrapolated onto WHM, this new formula appears to be able to accurately predict the Average HP healed over a much broader stat range. Whether the terms in this formula are true representations of the actual healing formula does not really matter as long as this formula produces consistently accurate results.
First let us reiterate: These formulas are definitely not the true in-game formulas. They are simply working calculators that produce 99.0%+ accurate results. The power of these formulas comes from the ability to determine which stats produce the greatest point-per-point returns.
L50 THM AVG HP Healed = (Potency/300) x [(MD) x (0.01156 x MND + 0.001314 x DTR + 0.3736) + (0.2229 x MND) + (0.06071 x DTR) + 1.7786)]
L50 WHM AVG HP Healed = (Potency/300) x [(MD) x (0.0114 x MND + 0.00145 x DTR + 0.3736) + (0.21 x MND) + (0.11 x DTR) + (0.00011316 x MND x DTR)]
L50 WHM AVG HP Healed = (potency/300) x [(MD x (0.01123 x MND + 0.0021 x DTR + 0.28) + 0.242 x MND + 0.1 x DTR
Enmity may be best thought of as the consequences of your actions. Every action in the game either adds at least 1 Enmity or takes away 1 Enmity. Throughout this chapter, there are several terms for Enmity that I will use: hate, threat, and aggro. Readers will find that I often interchange these terms freely. The purpose of this chapter is to guide readers through a mental process of understanding how to conceptualize Enmity and use this knowledge to increase their Enmity management.
In 1.0, there were two types of Enmity: Healing Enmity and Damage Enmity.To set the groundwork for determining the Enmity generation for all jobs, we must determine first whether damage actions and healing actions produce the same Enmity in 2.0. From a conceptual standpoint, players would expect Damage Enmity to produce more hate than Healing Enmity. This is due to the principle that bosses do substantially more damage to a party than the tanks and damage dealers do to the boss individually. In order to prevent healers from constantly stealing hate, a non 1:1 ratio is usually utilized. In 1.0, the Healing Enmity to Damage Enmity ratio was 1:0.625.
To begin with, let us establish a credible means of testing Healing Enmity versus Damage Enmity. The testing process that my LS used is as follows:
Lvl 5 Orobon.
Only engaged 1 enemy while testing.
PLD engages with Provoke to get 1 Enmity.
WHM casts Cure I.
WHM heals PLD for 120 HP.
WHM has hate.
THM auto-attacks with weathered sceptor.
THM does roughly 15-20 auto-attacks at 3-4 damage/hit for a total of 61 damage.
THM has hate.
Notice a few things. First, three characters were used in this test. This was to eliminate any possibility of a “claim” bonus to Enmity generation. Second, only one enemy was engaged while testing. In 1.0, the number of monsters engaged influenced the Enmity generation from healing. To eliminate this confounder, we tested strictly with one monster engaged. Now let us turn to the results themselves. The L50 THM auto-attacking did exactly 61 damage to steal hate. Readers may be wondering why the THM had to do 61 damage to steal hate rather than 60 damage. This is due to a tie-breaking rule. If two players do an equivalent amount of Enmity generation, the player who used an action first will be at the top of the hate list. At 60 damage, the THM and WHM had equal Enmity. At 61 damage, the THM broke the tie and stole hate on the monster. With repeated testing using the method, we can somewhat confidently surmise that:
DAMAGE ENMITY = HEALING ENMITY X 2
1 point of damage = 2 points of HP healed
As a means of proof-of-concept testing, my LS performed another means of confirming the 2:1 ratio based on a methodology proposed by another user on the Beta forums:
L50 Water Sprite
Only engaged 1 enemy while testing
WAR engaged with Auto-Attack for 120 damage.
WHM casts Cure I for 309 HP gained on WAR.
WHM steals hate.
This simple proof-of-concept test shows that at least on L50 enemies, Healing Enmity is somewhere between 1-2.6 the equivalency of Damage Enmity. With repeated testing process, we narrowed it down to around 2.2:1 without any degree of attempted accuracy. As seen previously, this is most likely the 2:1 ratio. We chose to use all L50 characters and enemies to eliminate the possibility of dLVL confounding Enmity generation. Furthermore, we saw that with multiple enemies engaged, the hate generated from healing is at least partially divided by the total number of mobs. We did not perform precise measurements, but it is our best guess that the division is a strict division by ‘n’ where n equals the number of monsters engaged.
There is actually a third type enmity generation: Buffing Enmity. Every action that produces a buffing icon appears to generate Enmity. This Enmity value appears to be a flat +70 Damage Enmity in terms of equivalency. This would include actions like Regen, Stoneskin, Shield Oath, Sword Oath, Defiance, and many others. While not all actions were tested, every buffing action that was tested produced the same Enmity generation regardless of class or job. Additionally, in a manner similar to Healing Enmity, the +70 value is divided if multiple enemies are engaged. With this in mind, it is my current belief that all buffing actions produce a universal, flat +70 Damage Enmity generation.
There are several Weaponskills and abilities that have a special property assigned to them. These properties are listed as either increased Enmity or reduced Enmity. For example, Savage Blade is listed as Additional Effect: Increases Enmity while Quelling Strike is listed as “Reduces Enmity generated from each attack.” The astute reader will naturally wonder what kind of bonus these actions produce, i.e. is the bonus a flat addition/subtraction or some kind of multiplier. Following a similar testing methodology as used in determining Healing Enmity, we can answer this question. It appears that every action that is listed as modifying Enmity has a multiplicative effect. Here are all the actions and their respective multipliers:
Furthermore, the Enmity multipliers for Shield Oath and Defiance stack with the other multipliers. For example, if a player used Shield Oath + Rage of Halone and did 100 points of damage, the total Enmity generated would be the equivalent of doing 100 x 2 x 5 or 1000 points of damage. Additionally, Shield Oath and Defiance multiply the Enmity generation of every other action including buffs and healing. Elusive Jump deserves to have special mention; this ability appears to reduce the entire accumulated Enmity by 50%. We did not establish if there was a cap to this reduction; however, we tested as high as 2000 Damage Enmity being eliminated.
Flash is a Gladiator ability that is for gaining hate on a group of enemies. It is an AoE spell that is on the spell cast timer and uses MP instead of TP. Flash is different from other Enmity generating abilities in that the action itself does no damage. However, Flash appears to be influenced by Weapon Damage, Strength, and Determination. These modifiers were determined by comparing fully geared and naked Gladiators. While no strict numbers were determined, the Enmity bar was used for a rough approximation. Similar to the damage formula, Weapon Damage appears to have the biggest influence followed by Strength and then Determination. Just for example, a L50 Warrior with a Moogle Axe (WD=41), 353 STR, and 238 DTR will do roughly 495 points of Damage Enmity. Additionally, damage buffs do not appear to increase the hate generation from Flash. Now this is where my testing ended. I assumed that the Enmity generation would follow a similar formula as the damage formula. However, a Reddit user by the name of Kestiel made an interesting post on how Flash may follow a Normal Distribution. Here is his post:
Now there are several important things to notice. Kestiel did precise measurements obtaining values within 1-2 points of the Enmity generated. Also, Kestiel obtained virtually the exact same value for Enmity generated with Flash for a specific WD, STR, and DTR. This is in stark contrast to what I expected. I assumed that Flash would follow the +/- 5% variation as seen with damage; however, this appears to not be the case. Perhaps the most interesting part of his testing comes from the fact that increasing Strength did not correlate with a linear increase in Enmity generation. Instead, the Enmity generation appeared to follow a normal distribution in which a certain Strength value had the most “bang for your buck” in terms of Enmity generation. His data and testing methods appear to be very solid. However, in a mindset similar to Occam’s Razor, I am still uncertain whether this distribution is the true trend or whether there is some other modifier that is not being accounted for. The developers have appeared to have a simplified mindset in terms of stats and algorithms as compared to 1.0. And every other mechanism I have seen in the game has followed some form of linear trend. For the developers to choose to follow this kind of non-linear distribution for Flash appears to be very nonsensical. Regardless of this discussion, in terms of gameplay application, Flash is still less powerful than the Rage of Halone and Butcher’s Block combos for Enmity Generation. So determining the actual mechanism of Flash Enmity is not of immediate concern for me at this time.
For quick referencing, here are tables for Paladin and Warrior to help visualize the Enmity generating capacities for both jobs. Since no other jobs have a substantial number of Enmity multiplying abilities, I do not plan on making tables for other jobs.
AA: Auto-Attack, refers to the actual auto-attack damage being done.
AVG DMG: Average Damage, refers to the average of the min and max of damage done.
BR: Block Rate, refers to the number on shields labeled “Block Rate.”
CRT: Critical Hit Rate, refers to the player attribute found in the character info.
DEX: Dexterity, refers to the player attribute found in the character info.
DTR: Determination, refers to the player attribute found in the character info.
ER: Elemental Resistance, refers to the 6 player elemental resistance attributes collectively found in the character info.
INT: Intelligence, refers to the player attribute found in the character info.
MDEF: Magic Defense, refers to the player attribute found in the character info.
MND: Mind, refers to the player attribute found in the character info.
PAR: Parry, refers to the player attribute found in the character info.
PDEF: Physical Defense, refers to the “Defense” attribute found in the character info.
PIE: Piety, refers to the player attribute found in the character info.
SB: Shield Block, refers to the number on a shield labeled “Block.”
WAA: Weapon Auto-Attack, refers to the number on a weapon labeled “Auto-Attack.”
WD: Weapon Damage, refers to the Physical Damage value of a weapon.
WS: Weaponskill, refers to the actions labeled as “Weaponskills” in game.