MoP beta discussion - Page 10

clawberg · 08/19/12, 11:46 AM

Havoc12, Im trying to understand your formula, can you explain why you add M and T before subtracting it from 60?

Vintoran · 08/19/12, 5:15 PM

I have the same question: Would you mind sharing the starting point of the calculation? I cannot duplicate your resulting formula (wolframalpha - I'm lazy) and I'm very interested in factoring in haste. While your approach has certain problems like not depending on fight length and not being able to allow for different phases it's a lot better than anything I managed to think of.

I believe there are some minor issues with the PoH numbers you provided.
Solace currently regenerates 0.7% mana per 1.5 seconds, which is 0.47 mps, not 0.35.
However you need to cast Solace for 9.64 seconds to be able to cast one PoH (4.5% mana) more, not 5.

Havoc12 · 08/19/12, 5:58 PM

Originally Posted by Havoc12

A very rough way to calculate it is that solace needs to provide 4.95% of max mana per minute to equal the return from mindbender. This is because you get 0.35% per minute automatically since you can always use the 1s per minute needed to cast mindbender. Therefore you would need to cast solace for 14s every minute to equal mindbender. However that is not really a very good way to think of the comparison, because solace you also need to have enough time to use the mana you get from solace. Here is the detailed calculation I did for PoH
.......
In order to get an extra PoH in that time you need to cast solace for 5s and you need 2.5s for the extra PoH. Consequently if you fill your free time with solace you will get 1 extra PoH for every 7.5 seconds....mindbender provides roughly 1 extra PoH.

Its not enough to consider the time it takes to get the mana. You also have to have time to use it. If you filled all your free time with solace you would cast the exact same number of PoHs as you would without solace, but you would end the fight with extra mana instead of being empty. That is why the rough calculation above is highly inaccurate. You can't fill ALL your free time with solace you have to use some of that free time to cast extra heals. Remember that we are looking at the maximum number of spells you can cast per minute. That means if you don't cast any solace and you cast those n spells you will end up with zero mana.

Thus to find out how many extra spells solace can get you for a certain amount of free time you need to divide by the time it takes to gain the necessary mana PLUS the time it takes to actually CAST the extra spells. I mistakenly calculated the time taken to gain the necessary mana as M. This is because when I did the calculation solace returned 1% per second. Its actually 0.467% per second now, so the time taken to gain the necessary mana is now M/0.467

Thus the total time is M/0.467+T not M+T. Since we are looking at it per minute then the number of EXTRA spells gained by the use of solace is your free time (60sec - n*T)/(M/0.467+T)

I calculate the return of mindbender by dividing 4.95 (the retun per minute for mindbender) by the cost of the spell being cast which is PoH (I am using 5% but its actually been reduced now)

In order for solace to be better we require that (60-n*2.5)/7.5 > 1 => n < 21

Thus I calculated the general formula as by substituting in the time and mana cost (60-n*T)/(M+T) > 4.95/M

Unfortunately as you pointed out my numbers are off due to a series of errors. I used 1% per second as the return for solace when determining how much time it takes to get enough mana for a poh, but I used 0.35% as the return when calculating the return from mindbeder. I also used 5% not 4.5% for PoH.

I have now updated the original post with the correct numbers and a bit more detail on the calculation.

Haste is dead easy to factor in. You just modify T to T/(1+haste) and increase the return of solace to 0.467*(1+haste). The overall effect is just prolonging the time segment to 60*(1+haste) secs instead of 1min.

Mindbender currently does not benefit from haste. That might change if blizzard get their head out of their arse though.

Fight length can be factored in by changing the return of mindbender in the formula for a function that calculates the value of mindbender over shadowfiend for a particular fight duration.

The number of times mindbender is used depending on fight length can be approximated as int(FL + 0.5), where int converts the number in the bracket to an integer and FL is the fight length in minutes. I.e. for a fight that is 3minutes 20 seconds, you can use minbedner int(3.333 + 0.5) = int(3.8333) = 3 times: one at t=15s, two at t = 1min 15s, three at t = 2min 15s. The mindbender will come back at 3min15, but there is only 5s left in the fight, so its basically wasted.

The number of times the normal shadowfiend attacks can be approximated as 1+int[(FL-1.5)/3]. If for example the fight lasts 7.9 minutes, then the number of shadowfiend uses is 1+int[(7.9-1.5)/3], which equals 1+int(2.13) = 3. One at 1min15s, the 2nd at 4min15s and the final one at 7min 15s.

To bring it all together: For a fight lasting FL minutes, mindbender can be used int(FL+0.5) times, while shadowfiend can be used int(FL/3+0.5) times

From that a complete formula can be made that approximates the relative return of mindbender compared to shadowfiend and that can be inserted in the formula instead of 4.95 (actually its 4.833).

However this is kinda pointless as I explained. You need to average things out rather than include the discontinuity in mindbender return relative to shadowfiend with increasing fight time. If the return of mindbender is fixed in this way then the relationship does not depend on fight time.

The existence of phases does not change the relationship between solace and mindbender at all. It does not matter if you chain cast for 3 minutes and do absolutely nothing for 1 minute. Its the average number of casts per minute that matters as long as at the end of the fight all your mana is gone.

Vintoran · 08/20/12, 6:55 AM

Thank you for the clarification and the number fixing. The wolframalpha solution I linked earlier was actually correct as well, just developed further (updated version) and I didn't realize that.

Why fight length is important: Firstly in many fights where mana matters it's actually pretty common to have the fight length down to seconds (down to 10 seconds or so I'd say). Examples from this tier are Ultraxion, Spine, Madness and Zon'ozz (maybe not so much). Secondly the whole idea relies on knowing how many casts per minute you can cast (N). Since our mana should be nearly empty at the end of a fight, we can spend a bigger portion of our original pool per minute than we regenerate. This can be quite substantially for short fights, e.g. in a 6 minute fight you can spend 16.67% mana more per minute. That's more than MB regenerates and thus a substantial portion of our "regeneration".

My goal is to create a function that takes haste, fight length, spirit and some sort of spell selection (basically your M and T) as input and determines whether MB or solace is better. Obviously certain fights will have mechanics that swing the scale in one or the other direction, but I think a general understanding can be achieved.
Particularly the scaling of spirit and haste are interesting, since they work directly against each other. Someone in this thread proved earlier that spirit is at least at the beginning strictly superior to haste for mana regeneration via Solace, but at a certain point of high spirit and not high enough haste MB should be better than Solace.
The biggest problem I see at the moment is that M and T are most likely not equal for someone using MB instead of Solace. Unless the MB priest stands around doing nothing a lot he'd be using smaller heals more.

I don't have a lvl90 Priest at the moment, could someone give me a rough ballpark figure on how much haste we can expect this expansion? From what I heard for spirit I'd say 5k to 20k makes sense. Hopefully I'll get around throwing things together this week. Since the theoretical part was already provided by Havoc12, it's mostly a matter of displaying output usefully.

Havoc12 · 08/21/12, 5:04 AM

If you are holy with >8k spirit then solace is probably not worth it now with the reduced mana costs.

Right now in scaled gear for LFR I have 30% mastery, 5% crit and about 4% haste. There is a haste and crit buff available however, so I would expect 10% haste to be a reachable value in raid gear. Spirit and mastery/crit are more important as long as mana is really tight, so atm I personally reforge haste to mastery. Also it looks like the first accessible epic gear has mastery on it, not haste.

In inner will you can now practically chain cast PoH/CoH/PoM/renew if you have divine insight.

With this algorithm:

If Lightwell expires then lightwell
If CoH available then CoH
else if PoM available or DI procs then PoM
else PoH/renew in a 2:1 ratio
Sanctuary and cascade when needed
Mindbender on CD from 30s into the fight.

I find I can keep going for over 8 minutes in 25man lfr on feng without breaks in casting, pulling over 50k HPS. I cast at least 2 instants per 5 seconds and save 2.5-3k mana equating to an additional 2.5-3k mp5

For single target healing staying in inner fire and keeping renews up with single target heals is also completely sustainable.

Havoc12 · 08/21/12, 12:13 PM

Originally Posted by Havoc12

Hmmm how did you arrive at this conclusion. Before the recent changes solace gave you 1.5% of your mana with a 1.5s cast i.e. 1% mana for every second spent casting solace. Solace is now 2% mana for a 2s cast so still 1% mana per second of casting.

At 90 you will have 306k mana so 1% of that is 3k. To gain 10k mana you need 3.3 seconds of casting. So you would need to cast solace 3.3 seconds out of 5 with zero haste to get 10k mana back. Since you have 21% haste that would increase the amount of mana gained per second of casting to 1.21% or 3.6k mana per second of casting equating to roughly 2.75 seconds every 5 spent casting solace. You would basically need to cast solace 55% of the time just to break even with spirit.

There is absolutely no way you could use solace and haste to replace spirit for mana regen.

Hmmm I was wrong there 10k spirit only gives 5.6k mp5. So the original poster was right about there being only a 9% loss in throughput

Polopretress · 08/27/12, 7:01 PM

Sorry, my reply is probably late....

Our regen options are interesting now. Atm the formula for fiend and mindbender appears to be

number of attacks = 1+(duration)/1.5.

This is not modified by haste as borrowed time and borrowed time+PI do not increase the number of attacks. This may be a bug that will be fixed in the final version

i only see one change. 1 tic added ?
Are you sure , you get 11 tics ?

and for the number of tic , they have never been modified by borrowed time or PI or haste. (except hero when friend is casted before ==> do not remember the spell in English ...)

Havoc12 · 09/02/12, 10:39 AM

Originally Posted by Polopretress

Sorry, my reply is probably late....

i only see one change. 1 tic added ?
Are you sure , you get 11 tics ?

and for the number of tic , they have never been modified by borrowed time or PI or haste. (except hero when friend is casted before ==> do not remember the spell in English ...)

Yes the fiend attacks immediately when you summon it and then every 1.5s, so in 15 seconds it will attack 1+10 times (1.5, 3, 4.5,....., 15)

As far as I could tell for the most of wrath the entire cataclysm and the early days of beta fiend has been scaling with your haste including all procs. This is something I have tested a number of times.

Polopretress · 09/03/12, 6:03 AM

Originally Posted by Havoc12

As far as I could tell for the most of wrath the entire cataclysm and the early days of beta fiend has been scaling with your haste including all procs. This is something I have tested a number of times.

Thanks for the answer.
I am surprised that pet was scaling with haste because it seems to me that number of tics was always 10 in Cata 4.x (for example, same number of tics with and without Borrowed time)

Anyway, as it is not scaling anymore with present live, it is fixed.

One additionnal question :
Did you check if the tics are increased (as they were in 4.x) when the pet is launched before the cast of Heroism (or equivalent raid cd) ?

According to this formula, i do not see any advantage to choose Mindbender in the template.
If i take into account that
- 30sec. is necessary to launch the pet after the pull, and
- Efficiency at the end is good if pet is launched at least 30 sec before the down

Formula for Total mana return is :
Shadowfiend : =ENT(1+(Fight duration-1)/3)*3%*ManaBase*(1+12/1,5)
Mindbender : =ENT(1+(Fight duration-1)/1)*1.3%*ManaBase*(1+15/1,5)

With MP5 value = Total mana return /12 / fight duration

Depending on fight duration, the mana return is :

4 min
Shadowfiend = 1125 MP5
Mindbender = 1192 MP5

5 min
Shadowfiend = 900 MP5
Mindbender = 1192 MP5

6 min
Shadowfiend = 750 MP5
Mindbender = 1192 MP5

7 min
Shadowfiend = 964 MP5
Mindbender = 1192 MP5

The difference is too small to choose Mindbender.
In addition, if the Mindbender is not launched immediatly after its cool down , its performance decrease again and can be lower that Shadowfiend.

Havoc12 · 09/03/12, 4:55 PM

This is a graph showing the ratio of mindbender to shadowfiend uses over the whole fight plotted versus fight time in minutes

As you can see the average is about 3:1 ratio and the value oscillates less and less as the fight gets longer.

I plotted int(T+0.5)/int(T/3+0.5) vs T in excel.

I am not sure why your values don't quite match up. I cut out the times at which the breakpoints occur and I also added the -fold increase in mana return for bender vs shadowfiend using the formula int(T+0.5)/int(T/3+0.5)*(11*1.3)/(3*9) = int(T+0.5)/int(T/3+0.5)*14.3/27 =

T	-fold increase	use ratio
1.5	1.059259259	2
2.5	1.588888889	3
3.5	2.118518519	4
4.5	1.324074074	2.5
5.5	1.588888889	3
6.5	1.853703704	3.5
7.5	1.412345679	2.666666667
8.5	1.588888889	3

T = fight time in minutes
-fold increase = -fold increase in mana return
Use ratio = (Number of mindbender uses)/(Number of Shadowfiend uses)

Again I am assuming that MB should be used at least 30s into the fight and fiend 1.5 minutes into the fight. In fact you can use mindbender after casting like 3 spells, if there is enough damage right off the bat.

Shadowfiend on the other hand gives back 30% of your mana, so you cant really use it right at the start.

I see now this is not correct
Shadowfiend : =ENT(1+(Fight duration-1)/3)*3%*ManaBase*(1+12/1,5)
Mindbender : =ENT(1+(Fight duration-1)/1)*1.3%*ManaBase*(1+15/1,5)

The correct formulas are
int(T+0.5) and int(T/3+0.5), assuming you are using MB 30s into the fight and fiend 1.5min into the fight. The int() function retains only the integer part and discards the real part. So 3.9 becomes 3 just as 4.333566 becomes 4.

Ninahagen · 09/04/12, 5:36 PM

I'm fine with 1min30 for Shadowfiel and 30 sec for Mindbender. (I don't really care).
One should also consider the fact that the last cast might be partially useless.

Wishing to find %mana per 5 sec:
FLOOR( (T - shadowfiend_penalty) / 180 ) * 3% * 9 / T * 5
FLOOR( (T - mindbender_penalty) / 60 ) * 1.3%* 11 / T * 5
and plotting it for [180 : 600] (3 minutes to 10 minutes), I find average values of

0.7370% / 5sec for shadowfiend and 1.1917% / 5 sec for mindbender.

The difference being roughly 0.45% mana / 5sec.

So opting for the mindbender forces you to use 2 GCD more each 3 minutes, and returns 0.45% mana per 5 sec.

Solace is 0.7% per cast, and getting the same return mana-wise than mindbender implies having to cast one Solace every 7.77 seconds, aka ~23 solaces / 3 minutes, aka 23 GCD / 3 minutes more compared with a no-talent situation, and 21 GCD compared with mindbender (time-wise).
This is one GCD free each 8.5 seconds, one GCD every 5 or 6 GCD (or 7 depending on haste), just for solace to be on par with Mind Bender.

Polopretress · 09/05/12, 10:31 AM

Some of your comments makes me confusing.
Need to clarify or understand.
I agree also that modification/adjustement of the penalty for shawdow fiend and mindbender can be done according to what is really done IG.

Nina , You write :
FLOOR( (T - shadowfiend_penalty) / 180 ) * 3% * 9 / T * 5

Assuming that FLOOR( (T - shadowfiend_penalty) / 180 ) is the number of time we can launch the Pet, imagine that shadowfiend_penalty is 60 sec (because i take the assumption that shadow is lauched after 30sec and must be launched at least 30sec before the end of fight to be efficient (of course these parameters can modified to 90sec + 30sec or others values)

That means that on a fight of 4 min, FLOOR( (T - shadowfiend_penalty) / 180 ) = 1

The real result is that you are able to launch the fiend twice (at t=30sec and t=3min30) like the formula i gave :
INT(1+(T-shadowfiend_penalty)/180)

*******
Havoc, You write :
The correct formulas are
int(T+0.5) and int(T/3+0.5), assuming you are using MB 30s into the fight and fiend 1.5min into the fight.

The formula that you write is exactly the same i gave but it is already applicable for 30sec for MB and 1,5min for Fiend.
I prefer to use the complete formula :
INT(1+(Fight duration-Penalty)/CoolDown)

In your assumption:
penality is 30sec=0.5 for MB (CD=1)
INT(1+(T-0.5)/1) = INT (T + 0.5) ==> same result.

penality is 90sec=1.5 for Fiend (CD=3)
INT(1+(T-1.5)/3) = INT (T/3 + 1 - 1.5/3)= INT (T/3 + 0.5) ==> same result.

so
INT(1+(T-Penalty)/CD) is applicable with any value of penalty.

On the penalty itself :
- Agree to use 1min30 for fiend assuming that it is 1min after the pull and 30sec before the down in total.(even if it is always probably possible to launch it 30sec after the pull and maybe just before to use the first Spirit Shell)

- I am not confident with total of 30 sec. for MB since it is not efficient if the last MB is launch 15sec or less before the down.
I would prefer 60sec (inclusing 30sec after the pull and 30sec at the end)
An other possibility is 15sec after the pull and 30sec at the end

Let me know your feeling to be as close as possible than IG

I wanted also to warn that even with these assumptions, MB is probably to much valorized because it appears to me that launching MB exactly at the CD every minute is probably unrealistic on each boss.
First, it is very easy to loose some seconds to launch it when Cd is up (maybe 5sec each time of delay)
Second, depending of the event of some boss, it is impossible to launch the pet when the CD become up. (ex : transition phase on morchoch, phase with no add/boss on blackhron, change of platform on deathwing etc...)

There are also examples where it will be mandatory to delay the launch if you want win all tics (example : not launch the pet before the dark phase on morchoch, not launch the pet before a transition phase on Hagara, difficult to launch on an add during the dark phase on zonn ozz etc...)
A lot of possibility to delay the launch of pet....

If it is the same for simple fiend, any delay on MB of few second have more impact on the number you will be able to launch the pet. (you can loose very easely 1 launch with MB if total delay is 15/20sec for example.
So the optimization of MB is more difficult than the simple fiend.
WoL analyze will give us a good information about this efficiency but i think that it is not stupid to remove 1 to the total of MB calculation.

Polopretress · 09/05/12, 11:19 AM

But under this comparaizon between MB, Solace, the real question is : how much we win in term of MP5 and what are the consequences ?

We speak only about a gain of 300/400M5 (@LV85) depending on the duration of the fight.

For solace, do we agree to modify so deeply our gameplay (10 to 12 sec per minute) for a so minor gain of mana ?
(on disc, SS already take time per minute without the possibilty to use Archangel which is not active under SS)
Answer is "NO" for me.

For MB, the risk to miss a launch is very hight and to win a miss will reduce the efficiency of MB almost to the level of normal fiend...
Answer is "NO" for me, i cannot take this risk for 400 MP5.

When FDCL is beautiful and just waiting to be taken ...and probably my final choice for Disc spec.

Ninahagen · 09/05/12, 1:04 PM

yeah
FLOOR(1 + (T - shadowfiend_penalty) / 180 ) * 3% * 9 / T * 5

That's the formula I used though. Just forgot to write the +1 here.

I don't think it's relevant to give personal opinion about one particular talent being better overall, because we can switch whenever we want, and thus we should. Point is, which numerical conditions dictate shifting talents.

Mind bender and Solace give mana, Solace being better if you have some particular spare time. You may use this mana however you want.
The other talent is restricted to monotarget healing. Since you "like" that one, have you estimated why?

MB is Mind Blast.

I do not share your fears about not being able to "optimize" mindbender. I don't really care about the best estimation of arbitrary penalty in the formulas. It does not change anything much, and to each his own. If you want to add finer parameters, do so and see the results instead of speculating.

FLOOR(1 + (T - time_penalty) / (cooldown + delay_penalty) ) * 3% * 9 / T * 5
For a delayed Mindbender to be "as good as" a perfect (never delayed) shadowfiend in your apocalyptic scenario, you have to delay it 30+ seconds each time. That's gross.

Havoc12 · 09/06/12, 1:17 PM

Originally Posted by Ninahagen

I'm fine with 1min30 for Shadowfiel and 30 sec for Mindbender. (I don't really care).
One should also consider the fact that the last cast might be partially useless.

Wishing to find %mana per 5 sec:
FLOOR( (T - shadowfiend_penalty) / 180 ) * 3% * 9 / T * 5
FLOOR( (T - mindbender_penalty) / 60 ) * 1.3%* 11 / T * 5
and plotting it for [180 : 600] (3 minutes to 10 minutes), I find average values of

0.7370% / 5sec for shadowfiend and 1.1917% / 5 sec for mindbender.

The difference being roughly 0.45% mana / 5sec.

So opting for the mindbender forces you to use 2 GCD more each 3 minutes, and returns 0.45% mana per 5 sec.

Solace is 0.7% per cast, and getting the same return mana-wise than mindbender implies having to cast one Solace every 7.77 seconds, aka ~23 solaces / 3 minutes, aka 23 GCD / 3 minutes more compared with a no-talent situation, and 21 GCD compared with mindbender (time-wise).
This is one GCD free each 8.5 seconds, one GCD every 5 or 6 GCD (or 7 depending on haste), just for solace to be on par with Mind Bender.

I think that is the wrong way to compare solace with mindbender. Please refer to my previous post. To summarise. Its not enough to have enough free time to cast solace you have to have the free time to also use up the extra mana.

Lets say for example that with the mana that you have you are able to chain cast all the way through the encounter and that the time you spent not casting because you are oom is exactly 1 GCD every 8.5 seconds. If you fill all those GCDs with solace you will end the fight having cast the exact same number of casts as you did without solace, only you will now end the fight with a surplus of mana.

In my post above I derived a relationship showing exactly how much free time you need to have before mindbender and solace become equal.

What we are neglecting of course is using HoH to boost the returns of both solace and mindbender/fiend. Since fiend is a 3 minute CD you can boost it more than you can mindbender.

Disc has absolutely not time for solace. If there is any downtime in healing you need to be building aegis. So disc has to stack spirit until they can chain cast pretty much. Otherwise its really difficult to compete.

I just did an LFR with 4monks. The raid spikes to 20%, 4 monks press their uplift button raid back to 100% in one GCD. I took a screenie of the monks hitting 73k HPS with 40-50% overheal(!!). Yep monks are definately balanced. Shamans are also really bad. Totem+healing rain on 25 ppl is well over 60k HPS. Basically a shaman can cast nothing but totem and healing rain on a stacked group taking constant and outheal you massively.