Actually there is no softcap for disc priest, atm I have over 700 haste and after casting PWS GCD is reduced below 1.0 sec. Even with 700+ haste and using engineering glove enchant my FHs are 0.8s and still slightly slower than the GCD.
I believe that at hard cap, 1180 haste (36% from gear), it will be at 0.75 sec. It seems even thought the tooltip says spellhaste the mechanics are the same as PI or Heroism/Bloodlust.
Actually there is no softcap for disc priest, atm I have over 700 haste and after casting PWS GCD is reduced below 1.0 sec. Even with 700+ haste and using engineering glove enchant my FHs are 0.8s and still slightly slower than the GCD.
I believe that at hard cap, 1180 haste (36% from gear), it will be at 0.75 sec. It seems even thought the tooltip says spellhaste the mechanics are the same as PI or Heroism/Bloodlust.
There is a limit on the GCD and that is 1 second.
Whatever you are seeing, either you're looking at it wrong or there is latency issues.
I'm not sure but I think I see something wrong with this. You can't apply the 10% haste to the amount of spells you cast but you have to apply it to the GCD and casting times then look at how many spells you can cast.
[...]
If you have 0 haste your GCD is 1.5. Let's say you cast a cycle of PoM, CoH, Renew, Renew, Flash Heal. It takes about 8 seconds to spam those 5 spells. (With your cycle, it should only take about 6 seconds since the last spell is instant, even with some lag)
That's a minor point, but even for an instant, you need to take the GCD into account. You will be locked for that time, and even if your last SOL heal occurs a T=6s, the next sequence will starts at T=7.5s (or T=8s with some lag).
If you gain 10% haste, your GCD is now 1.36. In that same 8 second window, you can cast 5.9 spells, we'll say 6 spells though since it really takes slightly less than a full 8 seconds to cast the original 5 spell cycle. Adding 10% haste allowed a whole extra spell to be cast in the same window of time.
In fact, we're back into granularity questions, and cycle changes. These are tricky point with haste, and there are no real proper way to handle them in a general case.
You could well argue that you would use your cycle with 10% haste.
You could also argue that with 10% haste, my previous cycle requires 6.81s, and including lag, it will be more that 7s, and you will keep the same cycle. Then in an arbitrary 8s interval, you'll have in mean really 2.2 HC-friendly casts.
I think that if you want to treat it correctly, but in a general way, the following approach would be the correct one :
1/ Determine which frequence you use PoM / COH (can be 7s at best). That leads to a cycle length T, which should not really depend on your haste.
2/ Determine how many cast you can squeeze in that period : you get a real value N = T * (1+haste) / 1.5
3/ Consider that 2 cast at least are POM and COH, so you get at most T*(1+haste)/1.5 - 2 HC-friendly cast.
4/ Take into account your SOL casts : either with a fixed number A of cast, or with a proportion B of free casts.
5/ Your final number of HC-friendly cast is then either K=T*(1+haste)/1.5 - (2+A), or K = (1-B)*[T*(1+haste)/1.5 - 2]
6/ In 8s, you get in mean K*8.0/T HC-friendly cast.
If you want it rigorously, you need to get rid of real cast number, and consider the probability Pn of having cast exactly N HC-friendly spells in the last 8s, where the current time is taken uniformly in the cycle.
For example, assuming a 7.5s with COH - POM - SOL - Renew - Renew (and 0% haste)
my mean number of HC-spells in 8s is 2 / 7.5*8 = 2.13.
But if I consider t in [0-4.5] or in [5-6] or in [6.5-7.5], I have exactly 2 Renew cast in the last 8s before T. That's a probability p2 = 6.5/7.5= 0.866
If t is in [4.5 - 5] or [6 - 6.5], I have 3 Renew in the last 8s. Probability p3 is then p3 = 1/7.5 = 0.13333.
Note that we have exactly 2 * p2 + 3 * p3 = 2.1333, as it shall be.
Now, you can have precise HC uptime, with is
But that's the kind of thing that are not really sexy to compute, and change in weird ways with haste.
Due to the convexity of x -> a^x function, the approximation HC ~ 1 - (1-Crit)^Mean is not exact, but it is not too far.
To investigate the effect of both haste and crit on Holy Concentration uptime, I decided to run a monte-carlo analysis. I assumed the rotation outlined above (ProM-CoH-Renew-Renew-FH) and calculated the uptime for a simulated five minute fight with zero haste. I ran thirty thousand simulation runs for each crit value and plotted the median and 90% confidence interval. The results are plotted below.
I then repeated the analysis with 25% haste. The mean uptime value was nearly identical -- within the width of the line on one plot so not shown overlapping. The only noticeable effect was that the confidence interval narrowed slightly. This means that on the fights where you get very lucky or very unlucky with crit procs, haste will move HC uptime closer to the mean.
I can say with a fair degree of confidence that haste does not affect HC uptime in any meaningful way. I confirmed that the same is true for inspiration using a discipline rotation. I suspect that this true for any similar ability.
----
Edit: I replaced the word "mean" with "median," since that is what is actually plotted. I changed "95% confidence bounds" to "90%." The three lines corespond to the 5th, 50th, and 95th percentile of uptimes from the simulation runs. Thus the uptime you see in a given fight should fall between the top- and bottom-most lines 90% of the time.
One of the things that I think has to be considered in considering the value of haste is the impact of lag/latency/reaction time, especially for holy priests. For most people and most servers, I would be surprised if there was less than an average 200 ms combination of latency/server side lag/reaction time. That means that with 0% haste, you would not be able to fit in 3 renews between CoH cooldowns, since the time lost makes the effective GCD more like 1.7 seconds. Assuming 200 ms, you would need around 20% raid buffed haste (requires about 400 haste rating from gear + standard raid buffs) to be able to fit in a 3rd instant cast between CoH cooldowns. To get a 4th cast in, you would likely need to be close to the soft haste cap and close to 1100 haste from gear.
I question the value of stacking haste past about 400-500 when you are using a instant cast heavy renew-CoH-PoM playstayle. Obviously, if you are using a lot of FH and PoH, it becomes more valuable, but for the instant cast playstyle, it seems like there is little benefit to stacking haste that doesn't get you an extra spell cast. Once you can hit the threshold to be able to get that 4th instant cast in, obviously it becomes much more valuable. However, until I have access to almost all of the 264+ haste/spirit pieces (and only 2 T10 items have haste on them unfortunately), I don't think it makes sense for me to gem pure haste or pass on spirit/crit items that are a SP upgrade.
Another factor is, if you are in a 25 man raid environment where replenishment uptime isn't always optimal (i.e. often 0-1 replenishers in a 25 man raid), would that not decrease the value of haste since it will lead to mana issues? Right now, it seems that only ret pallies are the only remaining 100% reliable replenishers. End game hunters typically do not run SV anymore, apparently destro locks take a huge DPS hit from speccing for replenishment, and there seems to be a trend for shadow priests with T10 4 pc to ignore replenishment uptime, since maintaining it reduces DPS.
Since no one else bothered to do it, I parsed the ICC10 log file mentioned above for rates of overhealing. That is, fraction of spells cast that result in overhealing.
Looks like our disco buddy (Baptistine) had over healing rates around 50%, depending on the spell. But they ranged wildly from 8% for Greater Heal (how in the world?) to 93% for Holy Nova. 38% for CoH in Holy mode, 64% for Penance.
I don't know if I trust these numbers, as this is my first time digging around in the combat logs like this. GH should have a high Overheal%, but maybe it was only used on a fight with a particular mechanic. The other holy priest (Khunanya) had >90% overhealing rate on his GH's though.
For penance, you'd expect with 64% OH% that the second tick must have OH'd and the third tick would be useless, but the snipe rate for that is only 2%, i.e. only 2% of his penance dots did nothing, rather than the 33% expected. So perhaps there was a lot of damage in between ticks; I'm not sure.
Perl implementation started from the above snippet. I broke this down per/healer and per/spell for completeness.
Toon: Who cast it
Spell: Spell cast
Count: Number of times cast
RawHeals: Real healing accomplished (i.e. excluding overheals)
TotalHeals: Total healing, including overheals
CritRate: % of casts resulting in crits
CritFrac: % of real healing resulting from crits
SnipedRate: % of casts resulting in 0 real healing (i.e. too slow or wrong target)
ORate: % of casts resulting in overhealing
OHealRate: % of non-crits resulting in overhealing
OCritRate: % of crits resulting in overhealing
OFrac: % of total that is overhealing
OHealFrac: % of non-crit subtotal that is overhealing
OCritFrac: % of crit subtotal that is overhealing
Now that heals have been nerfed in pvp wouldn't haste result in a higher HPS in PVP?
That's one of the reasons why good Priests get the haste offset for their PVP gear. Hydra posted a full gear guide and stuff on http://hydramist.net if you're interested in looking into it.
Good priests don't get crit because so many of their defensive spells can't crit. And their offensive spells that can crit only crit at 1.5x damage before resilience.
Priest mana regen is peeling from team mates and drinking.
So spellpower and haste are left as a viable stacking stat. Some priests go spell power, some go haste. If you're part of a good team you'll play less of a healing role and more of an offensive/utility role. Haste is probably better for this.
Stacking haste has very little to do with the healing nerf in arena, as haste stacking happened long before that change was implemented.