Holy Paladin - BIB


#1

I have read this post and still have a question.

My question is:
I’ve done a lot of research about what holy paladins are using for legendaries. Inevitably, the best holy paladins are using Velen’s Future Sight and Ilterendi, Crown Jewel of Silvermoon. I own both of those legendaries but the BIB selector is inevitably selecting Marad’s Dying Breath and Prydaz. I’m finding these legendaries to be hovering around the 3-6 range on BIS lists and yet Mr. Robot says they are best for me. I’m wondering if it’s time to break up with Mr. Robot and go and try to best gear myself on my own. Please tell me there’s a way to save our relationship!!


#2

Picking legendary items is a very complicated problem.

First, for a healer, what does it even mean for a legendary item (or any item really) to be “best”? Also, what makes a holy paladin one of the best holy paladins? And, how are those people deciding what legendary items to use?

What we have done is created the first healing simulator for WoW that can actually simulate enough of the variables to give actionable advice. We run simulations to find what the most effective items would be. This is still fairly new (the current version was just finished prior to 7.2.5) so we only have one healing script, based on a krosus-like fight. I have more scripts in the making.

So, if you simulate the different legendary combinations for your character, you can get a pretty good idea of how effective they would be:
Velen + Jewel, 930.5k HPS
Prydaz + Maraad, 958.0k HPS

Note for the velen + jewel combo, I assumed that you have a drape of shame you could use instead of maraad’s, which is the best back item besides the legendary. I replaced your neck with an optimal ilvl 900 neck item.

So, as we can see here, the healing output of each set is theoretically very similar. The legendary items preferred by the “best” paladins will typically vary a lot based on trends more than any real data. It is also heavily influenced by the fact that mythic raiding is almost a different game for healers compared to heroic or normal raiding.

Every holy paladin in mythic raids uses beacon of faith as their level 100 talent, as an example. They stand in melee range and heal the melee/tanks exclusively. They also heavily overheal on purpose in mythic raids, as well as compete with other very aggressive healers, making Velen’s a popular choice. If you are doing heroic progression and struggling to keep your team topped off, Velen’s loses a lot of power.

But really, this all comes back to data vs feelings. Mr. Robot uses completely data-driven theorycraft. That is our style. That is what we bring to the table that is unique. Other people use the items they “feel” are best and giving them “good parses” - even though it is literally impossible to do any sort of in-game test to verify which of two legendary combos of fairly similar value is actually “best”. Actual game conditions vary too much, and you’d have to do the fight over a thousand times with almost nothing changing. That is why we make simulators! Even the best mythic guilds only do fights a couple hundred times at most.

You can always lock in your preferred legendary items and have Mr. Robot optimize around them as well. Whether an item is theoretically optimal or not might not matter as much to you as the play style, which is totally fine. Healing isn’t the same as DPS - topping healing meters means very little in reality, since healing meters are so easy to game.

Creating an optimizer that can always pick the legendary items that simulate to the highest HPS is the problem that we are currently spending most of our time on. The amount of data required to reliably accomplish this is massive, so we are trying to find ways to find the answer with a “reasonable” amount of data. Reasonable to us means a few million simulations. The current default gearing strategies will get you in the right ballpark - say within 5% or so. You could get slightly better results by making a custom gearing strategy. And then you can always manually tweak it from there if you have strong feelings about particular items or want to squeeze out that last couple theoretical points of output.


#3

This is very interesting to me, thanks for writing it up!

Together with a couple of other Holy Paladins I’ve been responsible for a fair portion of the recommended gearing strategy for Holy Paladins, most notably in the Paladin Discord (Hammer of Wrath). I maintain a big spreadsheet that attempts to calculate stat weights and do trinket recommendations, and I built WoWAnalyzer.com (and its Holy Paladin implementation) to assess actual performance of talents, traits, mastery and items from logs to provide data to base recommendations on. Long story short; I know a thing or two about Holy Paladins and the recommended gearing strategies.

A best legendary does competitive HPS compared to alternatives, is reliable, not much harder to use than alternatives and helps save lives. A top parsing holy paladin is often one that underheals a fight. It’s still very important to play very well to get an absolute top rank (which a lot of people consider to be the best), and hitting the right buttons at the right time and often enough is way more important than the legendaries you pick. I’m pretty sure you can get just as easily get a top rank with any combination of the the top 5 or 6 legendaries.

[quote=“Swol, post:2, topic:1528”]What we have done is created the first healing simulator for WoW that can actually simulate enough of the variables to give actionable advice. We run simulations to find what the most effective items would be. This is still fairly new (the current version was just finished prior to 7.2.5) so we only have one healing script, based on a krosus-like fight. I have more scripts in the making.

So, if you simulate the different legendary combinations for your character, you can get a pretty good idea of how effective they would be:
Velen + Jewel, 930.5k HPS
Prydaz + Maraad, 958.0k HPS

Note for the velen + jewel combo, I assumed that you have a drape of shame you could use instead of maraad’s, which is the best back item besides the legendary. I replaced your neck with an optimal ilvl 900 neck item.

So, as we can see here, the healing output of each set is theoretically very similar.[/quote]
I believe a Krosus-like fight is a pretty unique fight that you don’t see a lot, but it’s definitely a nice start. Warcraft Logs statistics seem to give some different results, but I guess that’s not “personalized”; https://www.warcraftlogs.com/statistics/11/#class=Paladin&spec=Holy&combatantinfo=Legendaries&metric=hps&region=1

Are you considering the damage taken from Light of the Martyr in your simulation? From my quick calculations I reckon they’d end up at basically the same HPS (±2k) if you adjust the healing done by that as is done by all acceptable loggers and models. Also I’m seeing that you’re using the old 4 set with very low critical strike chance (for that set), when you’d run a more acceptable critical strike chance you’d find that you’re proccing the 4 set bonus more often which would make the opportunity cost of casting a Maraad’s boosted Light of the Martyr worse. It’s also unclear how well this sim was at using Velen’s and Ilterendi, if you don’t do anything special they’ll perform slightly worse. Is there any form of mana insight? Mana is very important.

I’m left wondering what the point of showing these results was? It doesn’t show us that it’s right and that OP should continue to rely upon Mr. Robot, it just shows the results he was apparently already aware of, with questions whether it’s actually accurate still remaining.

This is true for a large part, but keep in mind that a large part of these trends are started by people that have done their research and did base it on real data. Still our recommendations are based around more than just data showing their HPS throughput. One such additional factor is “utility”; legendaries more likely to save lives where it matters (e.g. Maraad’s Dying Breath) get valued a little bit higher than legendaries that provide passive but often insignificant healing (e.g. Obsidian Stone Spaulders). This is a heavily discussed subject, so we try to only make small adjustments. There are a few other factors we take into account. This is also true for our other recommendations, like trinkets (for example Deceiver’s Grand Design can be responsible for a lot of HPS but it’s generally valued weaker).

We don’t believe that healers should be 100% focused on HPS but instead on killing bosses by keeping people alive. Losing a small amount of HPS but gaining life saving capabilities (such as with more spot healing, damage reductions, etc) is often a positive exchange towards killing bosses. The lost HPS is often made up for with reduced overhealing from yourself or from other healers, less damage taken or elsewhere.

Directly healing tanks is considered inefficient and only done if needed to keep a tank alive. While healing melee while standing near melee is preferred, it can not be done exclusively as this would simply cause too much overhealing. We try to spot heal whoever is lowest, and while our Mastery is important we do not hesitate to heal low ranged simply for the fact that keeping everyone alive is far more important than maximum mastery effectiveness.

No, we do not. I have never heard of anyone overhealing on purpose, and quality healing teams are working together, not aggressively competitively against each other. This means actively trying to avoid sniping other healers in your team. One small exception is that the general recommendation during the Velen’s buff is to use the highest efficiency spells even if you expect it will overheal since Velen’s will work nicely with that, but your goal is still to avoid overhealing as you’d still lose 50% healing when you do.

[quote=“Swol, post:2, topic:1528”]But really, this all comes back to data vs feelings. Mr. Robot uses completely data-driven theorycraft. That is our style. That is what we bring to the table that is unique. Other people use the items they “feel” are best and giving them “good parses” - even though it is literally impossible to do any sort of in-game test to verify which of two legendary combos of fairly similar value is actually “best”. Actual game conditions vary too much, and you’d have to do the fight over a thousand times with almost nothing changing. That is why we make simulators! Even the best mythic guilds only do fights a couple hundred times at most.

You can always lock in your preferred legendary items and have Mr. Robot optimize around them as well. Whether an item is theoretically optimal or not might not matter as much to you as the play style, which is totally fine. Healing isn’t the same as DPS - topping healing meters means very little in reality, since healing meters are so easy to game.

Creating an optimizer that can always pick the legendary items that simulate to the highest HPS is the problem that we are currently spending most of our time on. The amount of data required to reliably accomplish this is massive, so we are trying to find ways to find the answer with a “reasonable” amount of data. Reasonable to us means a few million simulations. The current default gearing strategies will get you in the right ballpark - say within 5% or so. You could get slightly better results by making a custom gearing strategy. And then you can always manually tweak it from there if you have strong feelings about particular items or want to squeeze out that last couple theoretical points of output.[/quote]
I’m excited about this. Having another source with a considerably different approach to base our theorycrafting on can be great. If the simulation is accurate enough this could even beat our log analysis through WoWAnalyzer, and give insights into some of the harder/impossible to accurately measure effects such as Divine Purpose. Of course we still wouldn’t base our recommendations on HPS alone, but we’ve always taken it into account significantly so it (presumably) being more accurate is going to be helpful for us too.

At the same time this sounds scary. Making recommendations purely on HPS throughput will misdirect people towards selfish and unreliable choices that in the end make their team weaker.

I have to admit if I had the time I’d be building a simulator to make stat weights to replace my spreadsheet model. When executed well it’s definitely a better way to do it, but when we’re unable to review the implementation it’s hard to assess if it’s done well and worth recommending to players wondering about their stat weights, trinkets, other items and legendaries.


#4

Thanks for stopping by and posting.

I can clear up a few things regarding the simulator that you asked about:

1.) Yes, the self-damage done by Light of the Martyr is taken into account in the simulations. I could add something in to make it show up on the report probably, to help visualize it.

2.) The implementation of everything is totally open source and can be viewed by everyone. http://www.askmrrobot.com/wow/theory/?spec=PaladinHoly&version=live - This is the page you would go to in order to view the implementation of all the Holy Paladin abilities. You can also view the boss scripts here. If anything is confusing to you, don’t hesitate to ask. Our simulator was designed from the ground up to be an open project for the community, making it easier for theorycrafters who are not programmers to be involved. The only part of our simulator that is not open source is the “engine” that runs the simulations and does all the back-end work like advancing time, doing rotation checks, etc. Class-specific information is all on the wiki.

3.) Those results I simulated for this person were just a couple examples with the gear he posted. The point was just that the two legendary combos are of similar theoretical value, so, the suggestions made by the website are not unreasonable - and could be used successfully in-game.

4.) I’m going to make an update to the default display of the simulation reports, because they make it seem like HPS is the only factor. In reality, the simulation tracks all the allies in the raid and also reports how many of them die. I want to move this information to be as prominent as the HPS. The rotations are set up to prioritize keeping as many allies alive as possible. The rotations that keep people alive also tend to do more HPS, but I’m sure it would be possible to get a higher HPS number if we weren’t worried about ally deaths. The simulations also include ally healers that are scripted with a set of AI rules - also completely editable in the boss script.

This can all be controlled by the rotation: http://www.askmrrobot.com/wow/theory/rotation/amrpaladinholydefaultraidhealingkrosus?spec=PaladinHoly&version=live

I wrote a little tutorial on the healing rotations that goes into some of the big additions we made that are necessary for healing simulation. Healing Rotations(APLs)

5.) The rotation does not directly heal tanks unless it is necessary to do so. The rotation also does not purposely overheal - it will save mana when there aren’t allies with sufficiently low health to heal. If you want to view mana usage, set the output level to “log” and then expand the log section on the report. Uncheck all event types except “energize” and you will be able to get an idea of mana usage. We may add a graph for this at some point when we have time, but the data is there.

I am a healing theorycraft junkie - I’ve played healers since vanilla and have created spreadsheet models in the past for all healing specs (except shamans). In legion when we made our simulator, I was determined to make simulation work for healers, since it is such a better tool to use. I really think that this latest version of the simulator finally does what we need for healing TC. I’d be happy to work with you and other TC’ers on setting up simulations.

Another user of the site, @twitchyftw, has been heavily involved in the healing simulations (mostly resto druid) and recently created a healing script for Sassz’ine. We’re looking to get that into the default script choices soon as another option for folks as well.