Draewind

Works well with double one lined batch file. I forgot how to loop and too lazy to look it up, so I can just adjust these lines:

Which calls a batch file named ReportMulti.bat, which has the following command line:

 

Daidalos

One thing I have been wondering about is how the extremely high chain heal HEP is being arrived at. From a purely TC standpoint haste is slightly over 1 HEP. I've asked Stass about this, but hes been busy so was wondering if anyone else can explain what factors are making the haste HEP of 2 or so that I have seen many reporting. I haven't been motived enough to start digging through source code so I was hoping someone could enlighten me on this.

Restoration Shaman calculations:http://spreadsheets.google.com/ccc?k...IWv3sPmcYnEJUg

 

Bayushi-ko

Could it just be as simple as that in actual practice, the amount of overhealing seen is enough to inflate the value of haste to such levels? I don't believe the addon keeps track of raid health or incoming damage in any way, so if chain heal is producing even low to moderate amounts of overhealing, perhaps SP and crit will lose value to a greater degree than haste will? When looking to improve throughput without knowing whether there will even be any healing needed after a given cast and when said cast overhealed, increasing haste would theoretically be the best choice.

Purely speculative on my part of course, but its the tentative conclusion I've come to after watching my personal haste HEP more than double from prior to 3.2 (currently reporting a haste HEP of roughly 1.55, up from 0.8-0.9).

 

Daidalos

While I don't disagree with this idea in gerenal, how much less haste would overheal compared to spell power seems to be an unknown. So if you have 30% overheal, it seems dubious to assume that 50 more haste or whatever will yield 0% overheal when calculating its HEP. Just seems like you can't know theoretical overheal.

Restoration Shaman calculations:http://spreadsheets.google.com/ccc?k...IWv3sPmcYnEJUg

 

TeKniciaN

It may be that people getting high haste HEP values have low haste on their gear to begin with. I got high haste values, when I regeared towards haste, I got these type of values when I had ~700 haste from gear:

Shaman Healing Equivalency Points:
1 SP = 1
1 mp5 = 0.8874 (calculated)
1 mana = 0.0072
1 Haste rating = 0.7069
1 Crit rating = 0.7530
1 INT = 0.6803 (actual)
1 INT = 0.8888 (max theoretical)




Which is weighted towards spell power. so perhaps the high haste values reflect the fact that the person getting those values has low haste to begin with, indicating that there may be some kind of diminishing returns for stacking haste.

 

Draewind

I got these values with more haste on my gear. This is with 725 haste:

Shaman Healing Equivalency Points:
v2.62
1 SP = 1
1 mp5 = 0.9574 (calculated)
1 mana = 0.0083
1 Haste rating = 1.8395
1 Crit rating = 0.9186
1 INT = 0.7843 (actual)
1 INT = 0.9821 (max theoretical)

With under 400 haste:
v2.62
Shaman Healing Equivalency Points:
1 SP = 1
1 mp5 = 1.0414 (calculated)
1 mana = 0.0085
1 Haste rating = 1.7616
1 Crit rating = 0.8369
1 INT = 0.7205 (actual)
1 INT = 1.0015 (max theoretical)


The same combatlog using v2.47 and just under 400 haste:

Shaman Healing Equivalency Points:
1 SP = 1
1 mp5 = 0.7844 (calculated)
1 mana = 0.0064
1 Haste rating = 0.9304
1 Crit rating = 0.8768
1 INT = 0.6338 (actual)
1 INT = 0.8728 (max theoretical)

So, using the latest version of the program, the haste value does not seem to decrease with the addition of haste to gear. Using the old version of the program, there is a very large change in the haste value. Which value is most correct, and why? What were the changes in the program that brought about such a large change in the haste value. If my increase in haste, with decreases in Int and Crit, has theoretically increased my healing output, I am not seeing better numbers unless the other healers in my group have improved to a greater extent than I have.

 

lrdx

- insert huge facepalm picture here -

Maybe the newer version is newer, with bug fixes and tuned modeling?

RTFM. shaman_hep - Addons - Curse -> changes

What is a "better number"? The only "better number" of your performance is the smaller amount of dead ppl at the end of the fight, if you are healing. If you got more haste, you can heal more ppl. If you got larger heals (through SP/crit/int), you heal less ppl harder, with possible more overheal.

My elemental shaman sheets: ESSE

 

TeKniciaN

this post contributes nothing to determining whether the inflated haste values are an accurate model for determining proper HEP values.

According to theorycraft posted by daidalos and others above, CH *should* benefit more from SP than haste, and LHW *should* benefit more from haste than SP.

The question is, why are we getting nearly 2:1 haste:SP values from shaman_hep for chain heal centric healing logs, despite the math that shows SP to be more valuable for CH.

 

Daidalos

I don't expect the HEP from the mod to match TC exactly (thats the entire point of the mod) but I would expect it to be +/- 30% or so anything over 50% deviation seems questionable. Since the haste HEP for CH is about 1 from TC I am dubious about the 2 HEP values I have been seeing. However, its possible I am just not taking something into account that Stass is.

Restoration Shaman calculations:http://spreadsheets.google.com/ccc?k...IWv3sPmcYnEJUg

 

Shamroq

For the last two posters, please refer back to the previous posts regarding haste vs spell power. This has been discussed repeatedly in both this thread (most of page 5) as well as the best practices thread and would be worth reviewing before posting these types of messages.

Healing is not dps - at the end of the day your goal isn't "do as much healing as I possibly can." A healer's goal is to keep the raid alive and running in the most effective way possible. I opt for haste to enhance my reaction times as much as possible - a fraction of a second may be what makes a difference. BUT - it becomes a bad thing if you don't balance haste with your actual group and the fights you're doing. I stack haste yes, but not to the point that someone with a group like Sixthy's does. Our raid groups are working through hard modes, but they're not 3 minute kills.

THIS is what Shaman_hep is designed to help you do. Review YOUR healing style in YOUR group and help you sorta that data out. I'm not technically oriented enough to tell you exactly how it works - but I'd imagine that if you're ending every fight with a lot of mana, shaman_hep will keep telling you that haste value is very high, regardless of how much you have. When you start ending fights with low mana, the mp5 value will increase accordingly.

 

Rapparee

TeK,
He told you to RTFM.

In the change log one should be able to read the following lines:

2.56:
* Major BUGFIX: Haste was undervalued for chain heal as it was not using the effective for the entire cast.
* Minor BUGFIX: Fixed the equation for the effect of Ancestral Awakening on Spell Power and Haste. The value of Spell Power and Haste should be slightly higher for spells effected by Ancestral Awakening and very slightly higher overall.


TeK also notes that Daidalos suggests Haste is overvalued by Stassart.

Daidalos' spreadsheet for chain heal only:
spellpower = Daidalos normalizes the other values to SP before calling them HEP.
haste = 1.00 to 1.13 HEP
crit = 0.36 to 0.37 HEP
Int = 0.26 HEP

Stassart's latest post's values for chain heal only:
1 SP = 0.841286
(0.2651 + 0.3023 + 0.1762 + 0.0977)
1 Haste rating = 1.49877635228794
(mp5 = -0.136734425289928)
1 crit rating = 0.439106811674114
(0.0686 + 0.1913 + 0.1086 + 0.0706)
1 int = 0.145147605782981
sp = 1
haste = 1.78 = 1.498.../0.841....
crit = 0.52 = .439.../0.841...
Int = 0.17 = 0.145.../0.841

Daidalos' numbers indicate haste is 2.92 times as effective as crit for healing throughput.
Stassart's numbers indicate haste is 3.41 times as effective as crit for healing throughput.

There is only a 20% difference in the ratios of Haste and Crit's effect on healing. Perhaps the bulk of the difference lies in how they determine spellpower's contribution to chain heal's throughput.

For instance:
Stassart's file suggests that adding one spellpower gives his chain heal 0.841 hp/s.
Daidalo's file suggest that adding 19.91 spellpower gives his chain heal 52.72 hp/s for one set of gear and that adding 22.5 sp, gives 53.88 hp/s... 52.72/19.91=2.65 hps/s, 53/22=2.39 hp/s.

I'm not sure why there is such a huge gap between these two gentlemen's mathematics on spellpower's contribution to the throughput contribution on chain heal of spellpower. However the difference is night and day.

 

Daidalos

I didn't want to bog down this in tons of math, but so far no one has been able to point out anything useful, so I"ll lay out my math and perhaps we can get to the bottom of this (this is not a "what stat you like best" debate. This is purely mathematical / model question). I went through some work earlier today to try and make it easier to understand my calculations but HEP is somewhat complex by its nature (which also means bugs are both likely to occur and hard to spot).

Lets stick to one thing at a time and take a look a haste HEP values. Once we got that sorted out we can examine crit.
I'll start out with the standard CH equation.
[base amount + spell power*1.88*(2.5/3.5)] * (purification) *(imp ch) *(set bonuses)

For these figures lets ignore set bonuses as well as relics since we should all be using a spell power relic now anyways (we'll assume the spell power is already factored in).

Raid buffed Stats:
3000 spell power
50% healing spell crit (includes talents)
700 haste rating + WoA + Swift Ret/ Moonkin

average hit (without crit) amount for first bounce of CH
[1130 + 3000*1.88*(2.5/3.5)]*1.1*1.2 = 6809

with crit factored in
(1-critrate)* avg ch hit + (critrate)*1.5* avg ch hit
(1-.5)*6809 + (.5)*1.5*6809 = 8511.6428

next calc cast time.
(ch cast time)/(WoA*imp moonkin)/(1.0+(haste rating/ haste per 1 percent)/100.0)
2.5/ (1.05*1.03) / (1+ (700/32.79)/100) = 1.9049

Lets assume a 4 hit CH (it doesn't matter since haste HEP value will not change since as overall amount healed increases the effective spell coef increases by the same amount)
8511.64285714286 * (1+.6+.36+.216) = 18521.3348

HPS = total amount healed / cast time
18521.3348 / 1.9049 = 9722.9958

Now we calc the new cast time if we added additional haste to get our HEP values.
adding 50 more haste
2.5/ (1.05*1.03) / (1+ (750/32.79)/100) = 1.8812

new hps value = total amount healed / new cast time
18521.3348 / 1.8812 = 9845.4894

Increase in HPS = new HPS - old HPS
9845.4894 - 9722.9958 = 122.4936

Now we figure out how much spell power would result in the same HPS.

First we take the increase in HPS and convert it to amount healed.
Total increase in amount healed
difference in hps * old cast time
122.4936 * 1.9049 = 233.3380

That is the total increase in healing over those 4 hits.
To convert this back to spell power we need the effective spell coef with crit for a 4 hit ch
hit 1 ch coef without crit
((1.88*(2.5/3.5))*1.1*1.2) = 1.77257142857143

adding in crit
(1-critchance)*1.77257142857143+(critchance)*1.5*1.77257142857143
(1-.5)*1.77257142857143+(.5)*1.5*1.77257142857143 = 2.21571428571429

adding in the 3 other jumps to the coef
2.21571428571429 * 2.176 = 4.8213942857143

so finally have the effective coef now we take the total increase in healing and divide by the effective coef to find spell power
total increase in healing / effective coef
233.3380 / 4.8213942857143 = 48.3963737816208
this is how much spell power would have done the same hps as the 50 haste we used
so
to calc HEP (finally!)
equiv spell power / haste rating
48.3963737816208 / 50 = 0.9679 HEP

If anyone has any corrections or criticisms feel free but please keep this technical (at this point this is purely a math modeling question not what is best for situation X )

Restoration Shaman calculations:http://spreadsheets.google.com/ccc?k...IWv3sPmcYnEJUg

 

TeKniciaN

Daidalos's math does not take overhealing into consideration, but stassart's apparently does, perhaps this is the root of the problem.

Perhaps this mechanic is what is causing stassarts model to have such a large discrepancy. My question is if this is an accurate model? Shouldn't haste be penalized by overhealing as well? Faster heals also overheal. Most heals we cast will overheal to some degree, if we are not devaluing haste proportional to spell power this could cause SP to be devalued, and cause the inflated haste values.

Perhaps we should penalize haste by average overheal % in order to put it on even footing with the overheal penalties for SP?

Maybe stassart could shed some light on the math he is using for each, or perhaps we can determine that by looking at his program. I will try to find it but I'm not the best at these kind of things.



Also, I miscredited Daidalos for showing why haste is better than SP for LHW, it was raparee, I apologize.

 

Daidalos

Correct. Since Stass takes overhealing into account I expect some deviation based on your actual overhealing, however values of 2 HEP seems way beyond any normal variance due to overhealing. Also I've had HEP calculations for all spells (including LHW) for a long time :-)

I was hoping someone else that was familiar with Stass's mod would know since I'd rather not wade through source code, but it seems that no one else knows so I might take a look if I have time at work.

Restoration Shaman calculations:http://spreadsheets.google.com/ccc?k...IWv3sPmcYnEJUg

 

stassart

I am going to try to shed some light on how shaman_hep calculates haste HEP value and SP HEP value. I think the difference in SP value is likely to be the difference in how overhealing is handled. Hopefully this may help Daidalos and I determine why our haste values differ (or anyone else).

The main code that assigns a value to Haste starts on line 10,444 in version 2.62.

It calculates the effective hps per cast of each spell including any Ancestral Awakening effective healing attributed to that spell using the average combat cast time of that spell.

1 haste rating for that spell = (effective hps per cast)*0.01*(1/32.79)*(percent of casts that aren't haste capped)

(0.01)*1/32.79 = 0.0003 / haste rating
(Without this it is the hps increase with 100% haste, this take 1% of that hps and the haste rating/1%)

Then that number is multiplied by the percent of that spell comprises overall effective healing for its overall effect on haste HEP.

Then the increased mana cost is subtracted:
mana cost/haste rating = (mana cost per cast/ average cast time) * (0.01)*(1/32.79)

And again for the overall impact on haste this is multiplied by the percent of that spell comprises overall healing.

Later in the code the haste hps/haste rating is then normalized to hps/SP to provide a ratio of haste rating to SP.

The main code for SP starts on line 10181 in version 2.62.

It is very spell specific as talents and set bonuses effect the modifiers of spells. So first the coefficient * modifier of that spell is determined.

It then takes the (percentage of noncrits) * (percentage of noncrits that didn't overheal) * coefficient * modifier
Plus (percentage of crits) * (percentage of crits that didn't overheal) * (crit modifier) * coefficient * modifier
Plus (percentage of ancestral heals that didn't overheal) * (crit modifier) * coefficient * modifier * (ancestral modifier)

That is then converted to increased hps.
Then it is modified by the percentage of effective healing that this spell comprises overall healing (including ancestral).

And then later is normalized to hps/SP.

Edit: I tried to do this more as description than mathematical equations, but if it is would be more useful to do it as more formal math I can do so although I'm an engineer not a mathematician so it might not be perfect. Although anyone wanting the more formal math would probably be able to look at the actual source with the line numbers given above.

shaman_hep Healing Equivalency Point combatlog parser

 

TeKniciaN

thanks stassart.

Being less mathematically inclined than other participating in this discussion I will approach this from a conceptual standpoint more than hard math.

SP overhealing model:
It seems to me that the way stassart is handling overheals for SP is extremely accurate, it appears that it even takes into consideration overheal crits that proc AA hits that actually heal.

Haste overhealing model:
It seems to me that the haste overhealing model is much more of an approximation than the SP overhealing model. I'm not sure that increased haste will increase HPS in the linear fashion of 1% haste increases EHPS by 1% However I cannot think of a better way to calculate haste HPS increase as accurately as the SP. I would assume that this is overvaluing haste to some degree since only while chain-casting will haste provide that linear increase in EHPS, and even then that is provided damage is consistent enough to not cause more overhealing on subsequent casts.


My question is, would it be accurate if we rated SP overhealing in the exact same manner as haste, at least in relation to each other? (even though the haste overhealing model is less precise than the SP one) Simply because there is no more precise way of considering overheals effect on haste throughput?

 

Daidalos

The logic seems fine at first glance. I'll try to work through an example when I get some time but got a pretty busy raid schedule this week. I'll edit this post when I get enough time to work out the math.

Restoration Shaman calculations:http://spreadsheets.google.com/ccc?k...IWv3sPmcYnEJUg

 

Jessamy

This issue was already discussed in the Best Practices thread, starting here. The consensus opinion seemed to fall on the side of favoring Stassart's model.

 

TeKniciaN

The discussion there involved haste, so it was related, however the closest thing I found was a post you made describing how haste does not suffer from diminishing returns for throughput like crit, somewhat akin to the way people think there is armor DR, but there isnt in terms of effective health.

what were talking about here is directly related to haste, but we are trying to determine whether the HEP values of 2:1 are actually correct, despite math that shows extremely different numbers.

 

Mandydeth

Your post was a little ambiguous, but I don't believe anyone claimed that 'haste has a 2:1 HEP ratio in comparison to SP'. Just that people were getting higher values for haste on a consistent basis, and some were getting values as high as 2:1. Furthermore the math that shows "extremely different numbers" didn't evaluate overhealing and its effect on SP effectively. Overall I find Stassart's model to be more accurate, and after regemming from full SP to full haste I can vouch for it 'feeling' correct as well.

 

MatsT

It's more likely that haste will decrease your overhealing than that it will increase it. Only in the situation where the overall raid healing already is "enough" will it increase overhealing and in that situation it doesn't really matter what you do.

 

Daidalos

The question that I have is not "What stat is better?", "What stat yields the most effective hps in practice?", or "Which model (I don't even consider my sheet a model. Its TC) more accurately reflects my healing style?". The question is "How do haste HPS HEP values of ~2.0 arise from analysis of your combat log and why is it 2x times the TC amount?". I think everyone agrees that haste is one of the most effective stats due to faster reaction times and less tendancy to overheal. Again this is not what being debated.

Restoration Shaman calculations:http://spreadsheets.google.com/ccc?k...IWv3sPmcYnEJUg

 

Draewind

Overhealing is always high when there are druids and such with powerful HoTs on targets. Overhealing is not a problem, except for the special fights like Gen. Vezax, or anytime mana becomes an issue for the Shaman. When the raid is taking massive damage we shine, and our overhealing evaporates. Is there any chance that shaman_hep could give a set of values which ignores overhealing and gives raw data like a Healing Total Points which ignores the effective healing and concentrates on the total values? Much of the difference could be addressed by the individual to lower overhealing and mana conservation with technique and use of the HEP values to focus on efficiency, keeping an eye on the total values to verify the total output is not being lowered drastically. In difficult situations, it is the total healing output that will be a very important number, and haste will become a very interesting value as it applies to casts that were unsuccessful (hit) because the heal target died while casting.

I may be asking that which has been asked many times before, but I am old and slow. If the values are already there, then I simply don't know how to read the output as well I should. I get lost trying to read many of the threads as they tend to meander and not make sense to me many times. I do know that my healing is going through the roof about the time "WIPE" is called, and often in the past, my healing has made the difference in raid survival.

Competing to get a heal in there before someone else gets it will not benefit the raid, and is not the strength of the Shaman. It will certainly decrease my overheal values, but it will simply raise the overhealing of others. I like to think of my position as the reservoir or accumulator (hydraulics) of healing. My healing is always contributing but at moderate levels (with large overhealing) until the raid comes under pressure. It is then that I become a powerful healing tool to the raid. The Shaman acts as a firewall against spiked and sustained high damage levels ... "We keep them up long enough for the situation to improve and the other healers to respond efficiently ."

I want to gear to become the most powerful healer I can be in times of trouble. Everything else will take care of itself. I imagine that shaman_hep is designed just for this. Being old and slow, I need a little help figuring the gearing and technique to keep myself set up and ready to start spamming the Chain that keeps us up.

Are the "For healing output:" values made using Total healing values (including overhealing), with the "Shaman Healing Equivalency Points:" values calculated using the Efficient healing (corrected for overhealing) and mana regen values?

 

stassart

No, the "For healing output" just does not factor in mana in its score. The overall values factor in mana.

Being able to provide HEP values for total healing output including overhealing would be interesting and I would guess that it would get rather different results. I will add this to the TODO list although it may be a while.

shaman_hep Healing Equivalency Point combatlog parser

 

Draewind

I know the feeling. Prior to the patch, I went with max Intellect and Crit. My healing numbers soared, and my mana pool ran to 3000 when raid buffed. I seldom went oom, and had 60% crit (from reports). The change back to haste HURT. I am raiding with 725 to 750 haste, and 4pc T8.5 bonus, but Chain Heal still takes about 1.8 secs if I remember correctly (and I could be wrong because that is the cast time of nearly every spell). Now that I have changed, my +healing is still about the same at about 2800 when buffed. It must have been difficult to take such a large hit in SP.