Rawr Healadin Model - Page 9

Endoscient · 03/30/09, 4:56 PM

Originally Posted by galzohar

Will we be able to choose glyph of beacon (which will save mana based on uptime % and fight duration), mostly in order to compare it to the other not-very-useful glyphs?

Yes, and I added a nice chart to easily compare glyphs.

Note Holy Shock is shown as 0 since it will increase your Holy Shock usage, which will decrease your overall fight healing. I am going to trying to find a good way to model Holy Shock into burst healing rating, because that is what it is used for.

galzohar · 03/30/09, 5:16 PM

Isn't burst healing calculating just max HL HPS? After all that's what I'll have casted on the tank if I didn't mess up, holy shock is only "I messed up and he'll die if he doesn't get a heal NOW" or "he messed up and will die if he doesn't get a heal NOW" or "I have to move and need an instant" kind of spell. Lowering its cooldown only does anything if you spend a huge amount of time moving around. Maybe set HS to a "every X seconds" rather than a % so it doesn't scale with the glyph though.

I'd love it if we could set what rawr treats as "burst healing", personally I would set it to how well it scales non-crit HL HPS.

Just to make sure, does rawr compare burst to efficiency by relative gains compared to current stat? That is, if I set burst to "50%" will it count 1% burst as equal to 1% efficiency?

Endoscient · 03/30/09, 5:32 PM

Good point about Holy Shock, in most situations the Glyph won't change how often you use it for Burst Healing as opposed to I am moving and have nothing else to do. I am just going to leave Burst Healing as it is now.

I don't see the merit of discounting crits in Burst Healing, what is your argument for it?

Yes that is how Fight/Burst scaling works, this is the exact code.

calc.FightPoints = calc.AvgHPS * (1f - calcOpts.BurstScale);
calc.BurstPoints = calc.HLHPS * calcOpts.BurstScale;
calc.OverallPoints = calc.FightPoints + calc.BurstPoints;

Kandiru · 03/31/09, 11:42 AM

Most of the time a crit HL will be overhealing so increasing spellpower will only really benefit you by boosting the non-crit HL rather than the crit HL.

Very fight dependant though.

Endoscient · 03/31/09, 1:40 PM

Originally Posted by Kandiru

Most of the time a crit HL will be overhealing so increasing spellpower will only really benefit you by boosting the non-crit HL rather than the crit HL.

Very fight dependant though.

Most of the time a non crit Holy Light will overheal too. The times that you care about your max HL HPS a crit will probably not overheal.

galzohar · 03/31/09, 2:35 PM

I discount HL crit healing for burst for the same reasons I discount tank avoidance for burst. A situation where burst is needed is a situation where by definition all RNG fails - which includes both avoidance and crits. Even timed (non-RNG) burst is one you need to be able to heal through every time, and 1/2 the times your HL will not crit (or if it's 2 HLs, 1/4 of the times, etc), therefore whether or not your raid lives through the fight or not depends on how much HPS is done by your non-crit HL. Overhealing of crits is irrelevent here not just because you assume you need your max healing possible, but mostly because you're simply assuming that worst-case-scenario will not crit.

Assuming I'm reading the code right, the way you're factoring burst and fight points can bias either one depending on a bunch of factors. I'd rather have it normalized to relative increases, so 1% increase in burst points gets valued the same as a 1% increase in fight points. Fight HPS and HL HPS are both things you want to maximize, but 1 fight HPS is hard to compare to 1 HL HPS, while 1% increase in fight HPS is not as hard to compare to 1% increase in burst (though still not easy to compare, as you can see in the repeating SoL VS SoW and 1% healing VS 1% crit discussions, but I think it's at least somewhat easier than comaring 1 fight HPS to 1 burst HPS).

The problem with the above is that the item evaluation will also have to change - I assume currently it just calculates your new stats when adding the new item, in which case you'll need to add a calculation to relative benefit (X% increase) for items on top of the actual HPS gained. I currently do these calculations manually. I also use the % increase in value as a way to give item score in pawn a meaningfull value - an item with 100 more points, the way I have it set up, means a 1% increase in total healing power (adding up the % burst increase and % efficiency increase values).

Endoscient · 03/31/09, 3:16 PM

The tank always has a probability of dying to burst while you are spamming HL, which is what burst healing rating is trying to minimize. You need to demonstrate to me that crits do not reduce that chance for me to remove them for burst healing rating.

The only time it would be useful to compute HL hps without crits is if you are trying to achieve a minimum hps to tackle a specific exactly known hurdle, for example Flame Breaths on Sarth+3. The value you want to know then is the absolute non crit HL hps you need, like you need be able to guarantee 20k healing in 1.5 seconds after breath. Not the increase of how much HL hps you need to gain from your head slot, which is what the burst healing rating on charts is displaying.

The point of the scale is for you to define the relative value of Burst Healing to Overall Fight Healing. 1 average hps increase is just as hard to compare to 1 burst hps increase as 1% average to 1% burst. At all points they can be converted easily back and forth between each other.

I'll explain how Rawr calculations work in general. The module can be passed in a set of gear, buffs, enchants, etc which it then computes how much rating that setup is worth. The program creates charts by creating a base rating (for example removing the item from the shoulder slot) and then computes the comparison rating for various items (the different shoulder items you are testing). The value that is assigned on the chart is value of the comparison rating minus the base. The module calculation function has no notion of what you are upgrading from. So there is no point arguing about relative increases anyway, since they are not possible to implement with how Rawr is designed.

galzohar · 03/31/09, 4:08 PM

Just like you said, it's for having a minimum HPS to handle a bad scenario. HL crit increases the chance for the tank to not die, but it does nothing to reduce the variety of situations that have any chance to kill the tank. That is, if your non-crit HPS is higher, the boss will need to hit harder to have any chance to kill the tank, but if your crit chance is higher the boss can hit just as hard as he did before and have a chance to kill the tank. It's really the exact same discussion as stamina VS avoidance - avoidance will help you by not requiring as much total healing done, but it won't save you from the maximum possible burst of the boss, as the maximum, by definition, happens when avoidance (and critting) fails.

Then again the "average" burst value (factoring crits) isn't completely pointless, as on something like a longer "burst phase" (say, at least 20-30s as opposed to the 1.5-5s I was refering to above) the actual healing value is much less likely to have a bad roll, yet the phase is short enough for you to not care about mana for its duration.

The reason I think relative benefits are much easier to look at, is that when I see "HL HPS increased by 50" I immidiately think "how much do I care? Is it 500+50=550, or is it 9000+50=9050, which is very very different?" Same goes for X HPS increase over the whole fight. But according to what you say this probably needs a slight overhaul to the general rawr application (to have it also calculate relative increases compared to current values).

Endoscient · 03/31/09, 4:28 PM

Yes there are some situations where sp burst is better then crit burst, but there are also situations that the reverse is true. For sp burst to be better it assumes you can achieve a level of sp the can reduce tank death chance to near 0, while that is true on some fights I don't think it is the case for difficult fights, otherwise it wouldn't be that difficult since crits do happen very often.

Its easy to see in the calculation pane what your average and hl hps are, so you can get a sense of how much this slot/buff is contributing to your overall healing. When comparing two items/buffs/etc you care about the value of the two ratings compared to each other.

galzohar · 03/31/09, 8:41 PM

Basically there's always an amount of damage where with less healing you just can't heal it when dice roll against you. More spell power and more haste increase that amount, while crit does not. Taking crit into account is only relevant to that longer-term kind of burst (e.g. phase-based) rather than RNG burst.

Endoscient · 03/31/09, 8:55 PM

Originally Posted by galzohar

Basically there's always an amount of damage where with less healing you just can't heal it when dice roll against you. More spell power and more haste increase that amount, while crit does not. Taking crit into account is only relevant to that longer-term kind of burst (e.g. phase-based) rather than RNG burst.

There is also an amount of damage where haste/sp won't be enough to keep the tank alive and you need to win a dice roll. You can make a theoretical situation that favor each of the stats for burst. Also the times I care about burst healing are normally for longer phased based, there have been pretty few situations where max HL spam isn't enough hps and if it isn't its normally an execution issue not gear issue.

galzohar · 03/31/09, 11:10 PM

Originally Posted by Endoscient

There is also an amount of damage where haste/sp won't be enough to keep the tank alive and you need to win a dice roll.

If that amount of damage is possible, you're going to wipe at least some of the time if not most/all of the time. Fights are generally long enough for non-insanely-low-chance-RNGs to pretty much happen at least once every time. If you can't handle them you will need more healers in the fight, better overall stamina on tank and spellpower/haste on yourself, or just pure luck. Removing +healing for crit (in a way that keeps HPS the same) is like a tank dropping stamina for avoidance in a way that will keep his "effective HP" the same. It's not a good idea where short burst is an issue unless the only way in hell you'll ever handle it is by getting good rolls on the RNG. I've yet to see a fight where 100% successful RNG was required to succeed (granted, by "successful" I mean every time at least 1 of the good stuff roll, but it's still going to eventually not happen), but I've seen fights where bad gear/skill caused "not rolling bad RNG" to become a requirement.

Endoscient · 04/01/09, 12:05 AM

Originally Posted by galzohar

If that amount of damage is possible, you're going to wipe at least some of the time if not most/all of the time. Fights are generally long enough for non-insanely-low-chance-RNGs to pretty much happen at least once every time. If you can't handle them you will need more healers in the fight, better overall stamina on tank and spellpower/haste on yourself, or just pure luck. Removing +healing for crit (in a way that keeps HPS the same) is like a tank dropping stamina for avoidance in a way that will keep his "effective HP" the same. It's not a good idea where short burst is an issue unless the only way in hell you'll ever handle it is by getting good rolls on the RNG. I've yet to see a fight where 100% successful RNG was required to succeed (granted, by "successful" I mean every time at least 1 of the good stuff roll, but it's still going to eventually not happen), but I've seen fights where bad gear/skill caused "not rolling bad RNG" to become a requirement.

Sure if the average case requires you to get a crit you are screwed or probably doing the fight wrong. It is reasonable that if other things went wrong (say a cleanse resist on Steelbreaker) that a crit can save the tank on a difficult fight.

You still haven't answered my comment before. The tank always has a probability of dying to burst while you are spamming HL, which is what burst healing rating is trying to minimize. You need to demonstrate to me that crits do not reduce that chance for me to remove them for burst healing rating.

galzohar · 04/01/09, 1:33 PM

We're talking about different things. One is reducing the chance a tank will die to burst, and the other increasing the amount of burst a tank/player can take and still have 0 chance of dying. The first scales with crit, the second does not. If the first was the primary issue you'd see tanks stacking avoidance over stam when given the choice.

Endoscient · 04/01/09, 1:55 PM

Originally Posted by galzohar

We're talking about different things. One is reducing the chance a tank will die to burst, and the other increasing the amount of burst a tank/player can take and still have 0 chance of dying. The first scales with crit, the second does not. If the first was the primary issue you'd see tanks stacking avoidance over stam when given the choice.

In anything but contrived scenarios the amount of the sp/haste you are gaining is not enough to guarantee that a tank won't die. Does it matter that we are going from (totally made up numbers) 8,000 to 9,000 guaranteed non-crit hps or when the tank would need in the worst case 10,000 hps to live? You are still going to need some luck for the tank to live, whether it be on your end or the tanks.

Are you saying the goal of maximizing your burst healing isn't to reduce the chance for your tank to die to burst. Obviously making that chance 0 by getting enough sp/haste is the best way to do it if possible, but it isn't realistic in challenging content.