It seems that IF is incredibly bugged on Live too.
I already give you example when it works like +1.00 for Crit multiplier... That happened 28 january.

This log from 22jan
Expression Editor - 22-01 21:24 - SevereWar! - World of Logs

1st SS cast with IF was 207636
2nd cast w/o IF was 129200

It is not double. And it is not +1.00 like it was in some other cases...
Is *ucking capping at 100% crit!
I had ~25% crit there, so

207 636 / 2 = 103818
129 200 / 1.25 = 103360

So IF may work three ways right now:
1) SS = AvgHeal * (1+Mastery) * 1.3 * (1+Crit) * 2 [double]
2) SS = AvgHeal * (1+Mastery) * 1.3 * (1+Crit+1) [add 100%]
3) SS = AvgHeal * (1+Mastery) * 1.3 * (1+1) [capped at 100%]

And I dont know what reason behind that bs.
Maybe order which activated 1st. SS or IF.
But I clearly see all 3 possibilities in my logs.

P.S. BTW, need to exclude all procs here. Im gonna test it later hardly.

Perhaps it is not a mistake and it is designed to work like this.
ie : The total value of calculated shield is multiplied by 2 in case of IF.

Because it is necessary to keep only the aegis done by the critical portion of PoH.
The others aegis are also done if the crit PoH is replaced by a normal PoH so it is necessary to exclude them to analyse the impact of critical only.

Yes of course !
And without calculation , we can conclude that to have a 2.06 factor is not realistic because Crit heal produce ALWAYS more OverHeal than normal heal.
It is obvious.


So ok, i follow and agree with Havoc12 but it is necessary to conclude Crit v/s Haste on 5.2 !!!

Crit when there are continious damage on raid depending on boss capability ?
Haste when there are only high single damage (> 15sec) ?

I did it both on live and on PTR and everytime it doubles, changing my crit the order of activation and anything else makes no difference. The log values you see are probably due to procs. Just do it in game and check the tooltip. Trying to decouple stuff from WoL is difficult. You might also be hitting caps when you use IF, becuase it currently doubles the value of crit. Remember its only 220k cap just about. So if IF spirit shell was the 2nd cast you might have hit the cap.

I am pretty sure that IF currently exactly doubles your spirit shell hits, which is actually a very good boost to crit, because not only it does not reduce in value it actually gets double value from IF.

Give me the log and I can have a look at the sha fight.

The important thing to realise is that crit works well only in very specific situations, which are rarely met. Also it is difficult to uncouple.

Of course crit increases HPS, but the main problem is that its uncontrolled and the main contribution is from random aegis. Right now you have a way to keep that random aegis alive, come 5.2 you won't. If the fight has gaps in damage your aegis overheal will be high.

When all the limitations of crit are removed it is the best stat, but it is hard to find a fight without those limitations where you won't want to be holy instead of disc when 5.2 hits.

The limitation of haste is mana. If you don't have enough mana to support chain casting your best HPS spells, haste is rubbish (like it was at the start of MoP). In addition haste in 5.2 will be devalued because now penance and PoM are an important source of HPS boosts by themselves not via archangel.

So in 5.2 it will be pointless to stack mastery at all, so all of it needs to be reforged away or replaced. Spirit needs to be maxed and after that you need to stack haste as much as your regen allows and the rest is crit. Once you have hit 17% haste then its best to stack just crit.

If a fight has massive overheal then mastery might be more valuable than crit but I just can't see it happening.

This is another effective nerf for disc however. Our mastery is so devalued that we are barely getting anything from it. Where paladins and other classes get big boosts from their baseline mastery we get effectively nothing.

The important thing to realise is that our days of casting PoH are over. If there is a significant amount of raid wide damage holy is the only viable option. Disc is for fights where your main healing comes from atonement and spirit shell. PoH is too weak, too slow and has no smart targeting, so having it hit for so little and without guaranteed aegis makes it useless.

=================================================================================
In terms of gearing choices for disc, I would say if your mana can take it always take haste to 15%. If your strongly mana limited, then crit is the only option.

If you have high continuous raid damage (garalon style damage) then you need to be holy as disc is going to be a problem. I.e. if you feel tempted to PoH spam go holy.

Otherwise which stat is best depends entirely on the fight. If your overheal is low and your aegis loss is low, crit is strong. It takes 20% OH on crits and 15% overheal on aegis to make crit better.

However it is important to realise that at 35% crit haste becomes better than crit even with zero overheal.

Crit is devalued because you get a big amount of crit from your intellect and inner focus, while haste is not devalued at all because all haste buffs are multiplicative. That being said if the 2x IF factor on spirit shell remains, it might work out to favour crit instead.

It takes just 11% haste on gear to hit the first 15s breakpoint for PoH, because of the haste buff, so the first 11% points of haste on gear have a disproportionately higher value and that is why getting those first haste points (if you can afford it is very good).

Personally I am getting 12% haste from gear and the rest crit.

I think that if Holy/Sanctuary PoH is finally better than Disc PoH, that is probably a sign that something is working correctly.

There's also basically no question that mastery is the strongest stat for PoH in 5.1; my comments on crit have all been about 5.2.


On our debate regarding logs on the previous page:

I do understand the argument about how the value of crit is only the excess of normal heals--and like I said it makes sense to me (won't fully reiterate it again). I also understand that if you're trying to measure the value of crit, you could make a case for only including the effective value beyond 50% (or 48.5%) of the raw crit heal.

However, it equally valid (and somewhat simpler and more empirical) to ask, "how much more are crit heals healing for, on average, than non-crit heals." If the answer is 2x (or 2.06x) as much as non-crit heals, then that's a fact you have to deal with. No matter how much we talk about what we might think the value of crit should look like, the (admittedly currently not every detailed) factual information is what it is.

Maybe these two things aren't easy to reconcile with each other. But if it's in fact the case that a crit heal, on average, effective heals for twice as much as a non-crit; you can't just explain that away by saying we should conceptualize the value of crit in some other way.

Let h be the normal full value of a heal. Not counting Divine Aegis, the added value of a crit is zero (i.e. the crit ratio is 1) when the healed health deficit d is between 0 and h. When d is between h and 2.06h, the added value of a crit is d-h, and the crit ratio is d/h. When d is above 2.06h, the added value is 1.06, and the ratio is 2.06.

The thing is, for any probability density function on d, the expected value of a crit is the average observed crit. The expected value of a noncrit is the average observed noncrit. While it's absolutely true that a big crit is only better than a noncrit by the difference between the crit and a full noncrit heal, that's OK! Havoc, the example you gave is a little misleading, because you used different distributions on d between the crits and the noncrits. Let's try another one, where we have the same distribution.

1/3 of the time, the target has a hp deficit of h/2. 1/3 of the time, the target has a hp deficit of h. 1/3 of the time, the target has a hp deficit of 3h/2. Obviously, 2/3 of the time, a crit will be useless. 1/3 of the time, it will be 50% more valuable than a noncrit. The expected value of a normal heal will be 1/3*h/2 + 1/3*h + 1/3*h = 5/6*h, and the expected value of a crit will be 1/3*h/2 + 1/3*h + 1/3*1.5h = h. This gives an empirical crit ratio of 1.2*h. This means that per spell you did 1.2x as much healing with a crit as you did with a noncrit. This is a bit different than saying that the average spell would be boosted 1.2x by critting. In fact, in this example, on average critting makes a spell heal only .5/3 = 16.6% more.

I think this difference (how much do crits help your overall healing done, vs. how much would each individual spell be helped if it were a crit) may be the point of confusion? To me, the former is a more sensible way of thinking of things. "How much does crit help my total healing done?" is a much more interesting question to me than "How much would critting improve the strength of my average heal?" If I'm healing someone with a 10k HP deficit, I obviously don't care whether I crit or not.


Things I didn't take into account:

Intelligent use of Inner Focus will make Crit look better as a stat, while actually devaluing it. Unintelligent use (macroing it to PoH, say) will not have the former effect but will still have the latter. Of course, if IF keeps doubling Spirit Shell heals, that's going to be the way to maximize its returns.

Intellect procs raise both SP and Crit. This will inflate the apparent utility of Crit, but the effect is small.

Meta gem means crits are actually 2.06x rather than 2x. Another small effect.

Divine Aegis, in many fights, makes crits much much better.

Yes--and a few things I'd add are:

1) When you're healing someone with a very small HP deficit, then it doesn't matter at all what any of your stats are. The result is identical in every case. You may as well remove them from your mental analysis since, just as a matter of logic, it's only worth considering the cases where different choices actually produce different results. (Again, usual caveat--DA complicates things because the outcomes of even overheals can be different. But since crit is the best stat for improving DA in 5.2, this works in our favor anyway).

2) And the important point that shouldn't go unnoticed: your spell usage is not constant with stats. People point out that a crit on what would have been an overheal anyway is meaningless, but there's a flip side. A crit that's fully effective not only logs twice the effective heal, but also obviates the need for one subsequent cast. It's similar to the argument for people who can't see the value of Int--when you're casting the overheal on someone with a small deficit, it doesn't matter, but having that means you're less likely to even have to make that cast at all.

Jegges and hamlet, you are making a very big assumption. Lets say you have zero % critical. Take a group of 100 heals from a pool of 1000. Then do it again. What are the changes the the average heal will be the same. The answer pretty unlikely. So the group of heals which are crits, if you turn them to normal heals and you average them you will find that their average is NOT the same as the rest of the average heals. That is why averaging creates an artifact that depends on the actual variance of the crits and non crits.

Now lets look at this

Cannot be estimated from the logs.
What you get from the logs is average crit + error and average heal + error. So subtracting them adds an extra factor and dividing them magnifies the error.

In other words you are assuming that the density functions are the same for normal heals and crits but it is extremely unlikely to be the case. If it was you would never ever see crit have 100% value, as lonas as there is at least one heal with overheal different from either 0% or 100%. You have explained so yourself. So your values arise from different distribution functions from crits and heals, which are caused by random variation from the most part but they are actually pushed further by the use of cooldowns on low overheal phases.

So the two values are not identical. The value of each crit can be calculated directly and accurately from the log. The average heal can't. Attempting to do that and especially averaging the crit to heal ratio over different spells with different sample sizes and potentially different distributions is meaningless. You just can't accurately estimate the value of crits in this way.

What you are observing is just an artifact. It is easy to understand how and why. The very assumption required to make it valid (identical probability function d) would make it mathematically impossible for crit to have a 2:1 value. Either our understanding of maths is wrong or your assumption is wrong. Which one do you think is more likely.

The different probablity functions D means you can't work with the average you have to do it on a heal by heal basis or not at all.

1) Patently untrue, haste is not affected by overheal as long as its not 100% overheal. I have proven that in previous posts. Crit is disproportionally affected by overheal compared to any other stat.

2) That is actually what makes crits so bad. You can't know which target will get the crit so when you are targeting a target with a small deficit there is the same chance you will get a crit as when you are targeting a target with a large heal and targeting a person multiple times might mean you get the crit in the end rather than before. So crit does not have 100% value. Same with int, it never has 100% value. No one is forgetting that crit has *some* value, the question is how much is *some*. It is never 100% however and that is for sure.

If you cast 1 spell with 80% overheal without haste you will cast 1.1 spells with 80% overheal with 10% haste. For every part of the fight whether a high overheal or a low overheal part haste increases your number of casts in the same amount of time so it increases your HPS by a % rather than adding extra healing. In other words if you cast 10 spells at 99% overheal and 10 spells at 10% overheal you will cast 11 spells at 99% and 11 spells at 10% so it is still exactly 10% more HPS.

It does not matter if you are overhealing by 99% haste is unaffected. If you are overhealing by 100% then (100+x)%*0 = 0. So you are doing zero HPS in that phase no matter how much haste you have.

Unless you are overhealing by 100% haste is unaffected. The only thing that affects the value of haste really is mana. High HPS cds can potentially devalue haste, but not always.

Potentially you might think that it is better to wait until the damage becomes enough for you to not overheal. The obviously haste has zero value for you. You can however achieve better HPS by chain casting.

The only thing you achieve by chain casting with 80-90% overheal is dropping your HPM abominably low. Instead of doing X*haste casts with 80-90% overheal, you could just wait out the damage and do half of those casts and do exactly the same healing, with less overheal. You just described a scenario where more haste is useless because there is nothing to heal. With 10% haste you just ended your cast 0.1 second quicker and you'll just have to wait 0.1s longer before it's worth casting another heal.
If you decide to chain cast (because you want to get in more DPS for example), you're only increasing your healing and lowering someone else's healing because we have the unique capability of doing something meaningful where other healers are meaningless. I admit it is something we can consider, but I doubt it will increase the value of haste by so much compared to crit that it's worth it, considering any gain from haste will increase your mana throughput.

Also Havoc, are you considering Divine Aegis when computing effective healing and overhealing of criticals? For Discipline, there is no such thing as H = Hn + Hc where Hn = Hc. It's more in the lines of Hc = 2Hn due to aegis.
You say it's impossible to have a 100% increase on crits compared to normal heals due to there always being some overheal. But a disc priest's crit with 0% overheal starts off with being a 200%(actually even more) increase compared to a normal heal, so if you have exactly 50% overheal on the Hc part of a crit heal, you already increased that heal by twofold compared to when it wouldn't have crit.
Basically you could state that as long as you have less than 50% Overheal on the Hc part of a crit heal, it has increased your heal's value by more than 100%.

Your haste does not change how much damage the raid is receiving.
If each party member is taking 40k damage per second (with 2 parties) and your PoH is healing them for 100k per second thats 20% over heal. Bump up your hps by 10% and you get...
27% overheal.

Gloryrider, lowering the healing of other healers is never a concern for maximising output. If there is too little damage for the other healers to heal, one of them goes DPS. The objective is to always chain cast and never sit around doing nothing. Whatever you decide to do if you chain cast instead of waiting around when you could cast results in a drop in HPS. If there is little incoming damage you can smite and even if its mostly overheal its still HPS and DPS. The value of haste is the same. Waiting for people to take damage is pointless. If that is how you want to heal that your business and in that case for you crit is better than haste, simply because you are unable to exploit the power of haste fully. However doing that means choosing to not push your HPS to its limit. That means you and I have different aims and hence there is little point in a discussion.

For discipline it is a very good idea to chain cast PoH during periods of low damage, because you are building a big stack that will be absorbed when the high damage spike hits, which is very useful for your raid. Alternatively if your raid needs the DPS than its a good idea to smite during periods of low damage, but then you are still healing with atonement and hence producing HPS.

Haste is only limited by mana when you are chain cast as much as you possibly can.

I am afraid you didn't think this through. If overheal is high the incoming damage is low so aegis is being wasted. Aegis only matters when you can be certain it will be absorbed i.e. when you can keep it alive through PoH spam. I considered aegis and its "overheal" in all theoretical calculations. High overheal is bad for crit, it means nothing for haste assuming you don't decide to sit around and wait for people to take damage before you cast and for mastery it is beneficial if you can guarantee absorption or terrible if you can't.

You can't lump aegis in the crit and assume you are going to come up with a meaningful result.
===============================================================================

Heroic sha kill and wipes: Healing done - 28-01 21:05 - SevereWar! - World of Logs

Top heal atonement, but the overheal here is strongly affected by the presence of pets

Spell details for Искупление вины - 28-01 21:05 - SevereWar! - World of Logs

It turns out that 24% of the healing done by crits before overheal went to pets and it was complete overheal. Pets take very little damage so aegis on them expires, which is probably why your aegis wastage rate is so high.

Looking at the actual healing done by atonement we see that 12764320 was the effective healing done on players with an average overheal of 40%, however look at the actual logs: Expression Editor - 28-01 21:05 - SevereWar! - World of Logs

Taking out all the contribution of crit to atonement we find that 5468501 is the contribution of crit (without aegis) to effective healing. That means 16.3% of total Atonement healing done to players is due to crits.

Lets look at the rest of your spells

Cascade: 657744.5 (12.6% of total cascade healing)
PoM: 662078.5 (16.7% of total PoM healing)
PoH: 193324.5 (20.22% of total PoH healing)

Based on the observed crit rates we expect the return from crit heals to be:

0.205278592 atonement (efficiency: 0.7936896)
0.193717277 cascade (efficiency: 0.649105849)
0.241758242 PoM (efficiency: 0.692371901)
0.205882353 PoH (efficiency: 0.981933882)

Your aegis wastage rate is really high. Even though this is mitigated by loss of aegis from adds this is what you can expect when you don't keep aegis alive. This is a preview of what will happen come 5.2.

The effectiveness of PoH and atonement is quite high. This is because for some reason be it luck or the mechanics of the fight you have a favourable distribution with a high proportion of either 100% or 0% overheal on atonement and PoH

Refer to this post to see what that means: Disc Priest - Mists of Pandaria

Overall effectiveness is ~75% for the heal part. I personally think it is on the high side, because the distribution is fortuitous.

Spirit shell has 33% overheal and it looks like a high proportion of partial absorbs. Expression Editor - 28-01 21:05 - SevereWar! - World of Logs

This is bad for both mastery and crit, but Mastery is more affected since it has a much higher value.

Assuming 0.7 for aegis (the maximum possible, given the overheal and distribution) and 0.6 for aegis we are looking an estimated overall benefit for crit of (1+0.25*(0.75+0.6+0.6*1.4))*0.75 +(1+0.25)*(1+0.25*1.3)*0.7*0.18 = 1.3693125

So 1.5% increase in HPS for each point of crit. Personally I think this is an overestimate because you got an unusual distribution. I think getting is more down to luck on that particular attempt but the mechanics of the fight are also helpful. Crit does work like that: some times you can be spectacularly above the expected average, but on the flip side you can also be spectacularly below.

7% is actually quite a lot and you are dividing it so the error is variable1/variable2*sqrt((error1/variable1)^2 + (error2/variable2)^2), so its magnified. Then you are using a poison(10) so you are pretty much looking at 10% fractions. For means for your calculation 10% overheal and 19% overheal are the treated as the same.....

More importantly even though d is for health deficits, you are using that as a proxy for overheal estimate. To get the overheal estimate you don't want just d you also want the size of the heal. Then again the distribution varies with time, so you have a different sample size for the different distributions at different parts of the fight. You are not dealing with a 1-dimensional distribution but a multi-dimensional one.

If you think 100 samples out of 2000 heals is enough you are sadly mistaken.

You can clearly see from the data posted that if you half the crit heals and take the average you have a significant deviation from the average heal.

If you are still confident about your calculation post your logs and we can calculate the benefit for each individual heal. This is the foolproof way of doing it, because crits don't magically prefer lower deficit targets.

As for a normal distribution being a "plausible" model for raid HP deficits. If it was mate then crits would always have more overheal and they would depreciate twice as fast with overheal as normal heals. When they don't its exactly because they dont follow a normal distribution.

All in all your model is wrong, what you are calculating is an artifact of using averages when its inappropriate to do so.

Crit heals simple cant have 100% value. The distributions required are just too rare. It is mathematically impossible for crits to have more than 100% value.

I really don't understand how you can fail to see where your own arguments are leading you. A 2:1 ratio of crits to heals means that crits have the same overheal as normal heals. If d is the same for crits as it is for heals, then it is simply not possible to have crits with the same overheal as heals. Your own data tells you that the distributions of crits and normal heals are different. There is a difference in d due to sampling errors, which in this system are considerable and more importantly there are key differences in the size of the heals due to CDs.

consider 10, heals for 20k each on a deficit of 18k and then the same 10 heals with a 25% boost on a deficit of 100k.

you can expect 1 crit in each. The average for non crit heals is 18/2+25/2 = 21.5 the average for crits is 18/2+50/2 = 34

Without the 25% boost you get 18/2+20/2 = 19 for non crits and and 18/2+40/2 = 29 for crits

34/21.5 = 1.58

29/19 = 1.53

Do you see how the exact same d with non-varying heal size and 100% accurate crit rate still gives different crit/non-crit ratios?

As if that was not enough you are averaging the crit/non-crit ratios from different heals.

I think the discussion of whether the crit/non-crit ratio is a valid metric for crit efficiency is over. If anyone still believes that it is valid, then you are welcome to your opinion. I maintain that the facts prove you are spectacularly wrong and we end the discussion there.