I think I should clarify here what my spreadsheet does and why I think it is a useful model with regards to t10 2P. Obviously a better model would be to run a million seconds of combat with a real random function for DS reset, but I wanted something that would run really fast and still give a strong approximation. The option of just running 300 seconds of combat with a /random function is a poor one to my mind since you would get different results each time you entered that information and I want the sheet to return absolutely fixed values for any configuration of inputs.

What it does is simply assign 10 possible states to a variable I will call DSRefresh here. DSRefresh starts at 1 and 1 is added to it each time an event that could proc it comes up. If DSRefresh is at 1, 2, 4 or 6 it resets the cooldown of DS and the sheet continues calculating your rotation from there. Once DSRefresh hits 11 it gets reset to 1.

Of course this is going to give exactly a 40% rate of proccing over time, but it also models most proc event sequences quite reasonably.

Chance of a 1 proc 'dry spell'

Real: .6
Sheet: .6

2 proc 'dry spell'

Real: .36
Sheet: .3

3 proc 'dry spell'

Real: .216
Sheet: .2

4 proc 'dry spell'

Real: .1296
Sheet: .1

5 proc 'dry spell'

Real: .0778
Sheet: 0

Obviously not a perfect match, but it isn't too far off reality. I could change the proc events to 1,2,4,5 which would add .1 to all the Sheet values above (barring the 1 proc event) and would change the sheet from slightly overestimating the effect of t10 to slightly underestimating it. Because we can fill up dry spells in the proc rates with so many other spells though the actual dps difference is extremely marginal between these models, probably in the neighborhood of a few tenths of a %.

Here is some information based on the assumption (guesswork for now) of CS, DS, Judge and AA proccing t10 2P:

Highest dps rotation VS. Undead:

We shift to a new rotation because we are so much more GCD locked than before. Higher raw damage becomes much more important with t10 2P.

Judgement
DS
HoW
CS
Exorcism
Consecrate

Ability Uses:

DS 101
CS 33
Judg 33
Exo 10
Cons 5
HoW 6

Downtime per 300 seconds: 2 seconds.

We activate Exorcism and Consecrate so little that even if we did hit a bad string of luck on the DS refresh there are almost always going to be spells to fill that gap up, and those spells do virtually the same damage as DS per cast anyhow.

As a note, the Consecrate glyph in this setup adds 24 dps and the Exorcism glyph adds 42 dps. The DS Glyph would be adding 245 healing/sec in a single target encounter, up to 1k healing/sec in full 4 target AOE situation. The total healing from the base DS amount and the glyph combined in a full AOE situation would sit at 2617 healing/sec.

Now these admittedly completely outrageous numbers do come from the assumption that everything procs the DS refresh. How realistic that is we have no idea at the moment: it may well be that the DS numbers are substantially lower. However, it does illustrate just how powerful our passive healing could end up being.

There's nothing necessarily wrong with doing it like you are now. A lot of people (even hardcore programmers, mathematicians and statisticians) tend to have a wrong perception of "random".
The only problem (but this is an issue with any PRNG) is running into a harmonic state between the PRNG period and the modelled engine. You have a perfectly valid PRNG with a period of 10. It's a short period, so it's making harmonic state more likely.
This is the main reason why there is a need for "better" PRNG's, not really ones that return "better random numbers", because the numbers themselves are usually fine already even with simple formula's, but PRNG's with longer periods because detecting harmonic states can be nigh impossible on complex systems.

The "different results each time" isn't necessarily true either. A good PRNG will always return the same sequence of random numbers (yes, this sounds countradictory) given the same initial seeding. A lot of programs tend to 'randomly' seed the PRNG via some means (usually using system time) so it ends up giving different sequences each time. But you can force a seeding to a specific initial state to get reliable and comparable results between simulation tests if that's what you need.

If you aren't running into a harmonic state, then your way of doing it will yield the same overall results as the worlds best RNG given a long enough running time.

Those numbers look off though. Both Jeh and me made a FCFS simulator independantly using an actual RNG for melee swings and while the "all melee abilities proc T10" did come up in my case as proccing 102-104 times (with various rotations), that doesn't mean you get to use all of those as DS. My best case scenario's with DS up front in the rotation only ended up giving 75 as most DS used over a 5min fight. I'm sure Jeh can confirm similar numbers (or tell me I'm wrong and you're closer to the truth).

If the spreadsheet is coming up with 101 DS uses over a 5min fight, I'm inclined to think you either have a bug somewhere now or you do in fact have a harmonic state issue with the way you're doing it now.

I'm currently more or less expecting that only melee white will actually proc it though.

Good points by Redcape and Neraya. I was intending to re-seed the RNG every run (so the 3 runs would close to guarantee different results which could be averaged). However, based on the above I think I'll adopt this approach:
Have the seed an editable field with a default value.
Every run uses the seed on the sheet.

This allows you to compare apples and apples. Run once with priority xyz, again with priority yxz and have same sequence of attacks proc).

You can then reset the seed and re-run any series to see a change from initial run.

Bonus - timesaver as my spreadsheet will also calculate only once, not three times.

Since DiSac and AM are being nerfed for Ret (AM is bad for Ret since a lot of times you need to run Ret Aura to give the 3% damage bonus), I would like your numbers to be true (could drop Exo/Cons glyph for a small dps loss in order to get lots of passive healing power).

All item in the game with the same text as 2 piece work off all melee hits. That doesn't mean much though, since there are exceptions. What may end up happening is the 2 piece is procing off all melee, and if it turns out too good it will be nerfed/changed later.

Sanctified Retribution applies to all auras.

Some testing has shown odd bugs re: the buff dropping if you're not using ret aura.

If you run out of another Paladin's Aura while you do not have Ret Aura, 100% of the time the 3% damage/haste buff turns off.

On that topic, I'd recommend rather than "Latency" it be called "Delay." Human reaction time starts at about 100ms and averages about 215ms - this is for a task you're prepared and watching. Reference: Human Benchmark.

Average latency is probably close to 100ms, as you listed.

What's important is the time from the ability coming off cooldown on your client (time zero) and the server receiving the new ability usage.

Using an average reaction time of 200ms, and latency of 100ms you get:
Time 0: Ability off cooldown, can press.
Time 0.2: You press. GCD begins on your system.
Time 0.25: Server receives and acknowledges press (latency is round-trip, half latency is time required to reach server). GCD should be at 1.45 seconds remaining.
Time 0.3: Your client syncs the GCD to the server - GCD is "still" at 1.45 seconds remaining, not 1.4.

Only something such as a G15 keyboard/Autohotkey could reduce this delay. I say reduce, because something bound to spam the button every 0.1 seconds while held (note: while held, no rule breaking automation - holding the 1 key effectively spam presses 1 and no other key) could tag 0.0001 seconds after cooldown ends, or 0.99. Also if you had just pressed another key or were moving/otherwise distracted you may not have the next ability pre-pressed for it to trigger when immediately available.

So rather than 0.1 second latency, it'd be more accurate to set a Delay of 0.25 to 0.3 seconds (the latter I know is an option in your spreadsheet, just not your current setting). This should drive even more realistic results.

P.S. Neraya - if you didn't account for haste in spellcasts (as Redcape did - 1.33 was hasted with his present loaded gear stats) and delay you can indeed have different results. Different weapon speeds (and haste!) will alter the frequency or placement of procs by producing more/fewer autoattack and more/fewer double-proc (autoattack proc immediately after a special ability reset cooldown).

To add to that I did just a little testing preraid while standing around. I got the expected results for my personal weapon speed loss and gain based on running out of range of another pally aura with ret aura active and with fire res active. I got a DK to look at his weapon speed and as I had ret aura active and ran out of range while he was still in the other pally's aura he lost the 3% haste, then regained it when I returned to range with ret aura active. When I ran fire res aura and ran out of range he lost the haste as expected, but when i returned to range he did NOT regain the haste. I am therefore assuming that he also lost the damage bonus as well. The haste returned when I switched to a different aura as expected. I would say that pretty much requires ret to run ret aura until this bug is fixed as it is affecting the whole raid's dps.

God I hate posting from a completely different timezone to you guys.
No I don't have any set bonuses modeled but I'm very sure different dps values for abilities and haste will give varying results. I still think they will be in the ball park of the results I demonstrated unless a set bonus is used that really pushes up the damage value of one of the abilities. I also feel haste is an important factor as I've seen it push different rotations slightly higher at different ratings. I can show some examples if interested but there were a few posts by althor and glutton in the BiS thread that demonstrated this.
Just posting some real numbers in regards to the 2pcT10 below.
Obviously these numbers vary every run as the procs can be wildly different - hence the 200 runs.

As predicted note how few cons and exo we will use DS procs off all melee - makes me definitely want to drop exo glyph and go for the DS glyph.

EDIT: I highly doubt all strikes will proc the DS reset as this is a 600+dps increase.
Also if DS hits 4 targets the chance to proc for that attack alone is 87%.
Would be funny if it proc'd off our seal hits as well 0_0

An ability being bugged is not the same as an ability being "bad". Aura Mastery is a more flexible, more applicable, and generally superior raid cooldown to DSac on pretty much all the content which includes damage of a school resistable by a paladin aura, while sharing none of the massive downsides of DSac (cutting DPS in half, loss of the bubble as a reactive defense ability which is huge on hard modes, 2.5 times the cooldown length, requiring 2 gcds to put up).

Besides that, this bug isn't even particularly debilitating. It isn't like you can't dance auras. In fact, you should for many hard modes.
For example, on freya you could be runnign ret, then dance to conc when the conservator comes out, AM to give the window of being able to ignore the pacify and give the raid more time to adjust, and then simply dance back to ret aura when the AM is concluded.
To use another Ulduar example, on Algalon I always dance to shadow res aura and AM a star collapsing each time my cooldown is up as part of an AM rotation by all the paladins in my raid. When my AM is done I simply swap back to a useful aura, like fire for cosmic smashes, or devo.

Being aware of the Sanctified Retribution bug, you can very easily just work around it without sacrificing the superior power of AM as a tool. Most fights don't require the raid spreading out nearly enough to lose coverage of your aura anyway, but if they did, you could simply dance back to ret aura to reapply sanctified retribution and then return to your useful aura after everyone is reconsolidated.
For example, on Anub'arak kite phases the raid spreads out significantly, depending on strat. It is a simple matter to just swap to ret aura during this phase until the raid has collapsed back together again at the end of the phase and then return to your useful aura.

The point is, any situation where you are talking about losing raid coherency and people going out or aura range is going to be a movement phase. As a melee DPS, there are only limited things you can do during movement phases and you are almost garenteed to have free empty GCDs during such times taken up by movement. Thus, there isn't even a DPS loss to aura dance to compensate for the bug with Sanctified Retribution.

I'm tired of reading that ret paladins MUST run retribution aura because of this minor bug, and that by extension aura mastery is bad. It simply isn't true.

There have only been two fights in ulduar hard modes and ToC hard modes where DSac was a better cooldown than Aura Mastery on bleeding edge progression. General Vezax (no resistable damage) and Steelbreaker (no resistable damage we have an aura for).
At any rate, this is a thread about 3.3, and there's simply no comparison between DSac and AM come 3.3 because we do not have the points to be able to pick up Divine Guardian. So the comparison would be between a powerful raid defense cooldown with a slightly reduced duration, or a gimp party-only cooldown that breaks early unless you continue tieing it to your bubble use. Not taking either one isn't really an optimal choice for progression raiding, as you are sacrificing major utility for almost no PvE gain.

Getting similar numbers now as well on the amount of DS over 5min.

Due to how I had things set up, it was resetting the cooldown on DS before applying the GCD on the ability just used. This had the effect that a proc on DS wasn't resetting DS properly.


You did mention that for the proc in your spreadsheet you wanted "something fast". I.m.o. you can just as well use the built in rand() function then. This very likely even going to be faster than a counter+4tests+test for reset at 11. And it'll significantly reduce the chance of harmonics happening when you start testing with various amounts of haste.

The problem is if Redcape uses the rnd function in Excel then unless he keeps the seed stable and re-uses the same random sequence (which is basically the same as using 4 specifically chosen procs) then comparing one priority and another is apples and oranges.

Priority-set-1 using rnd could streak and proc far higher than 40%.
Priority-set-2 using rnd could also streak and proc far lower than 40%.

Suddenly Priority-set-2 looks like crap compared to 1, but if they'd both procced at exactly 40% 2 actually was superior.

His setup is clean and elegant, if static. It allows appropriate apples to apples comparisons.

I was thinking in the context of 3.3, and figured that the bug would be gone and no longer 'devalue' AM (though I run it as it is anyway). I wrongly assumed blizz would take care of it by the patch without checking if this was the case... Anyone know if it's being addressed (other than the typical "we're looking into it")? Like Brekkie mentioned I too like AM's flexibility and use it all the time; I can only see using it morein 3.3.

I know, I even said in a previous post his solution is a perfectly valid PRNG on condition no harmonic is happening. It's just that with a period of 10 you can fairly easily run into a harmonic that can drastically change numbers in either good or bad way.

I posted a sample of a harmonic in an earlier post (see post #72 in this thread). That one happens at 1.6s delay after a melee ability (GCD+reaction time/lag) and 1.3s delay after a spell using J>CS>DS>Cons>Exo (above 20% so no HoW active, and without T10 bonus). This particular one is a harmonic "bad" as you end up wasting a lot of time due to excessive gaps with nothing to do. If you were to test gear and replaced a bit of gear so you ended up having just the right amount of haste to get in this situation you'd see the gear would be a downgrade because of that harmonic. If you were 'in' the situation then almost any piece of gear taking you out would be listed as an upgrade.

With Redcape's PRNG resetting every 10 times, the chance of a harmonic happening between the PRNG and the swings/abilities used is fairly high. With the numbers used now it doesn't seem to be a problem but if you were to make a graph of dps done (Y axis) on a 5min fight for various amounts of haste (X axis) then you may see some spikes upwards and/or downwards caused by a harmonic occurring. I'm not spreadsheet savvy enough to figure out if this can (easily) be tested based on redcape's spreadsheet.


You're right about the apples and oranges comparison, but that's why you can usually initialise/seed a PRNG to a specific state. But I don't know if the rand() function in a spreadsheet can be reseeded that way, as I said, I'm not that spreadsheet savvy.

There are plenty of top end WMO Rets with Aura Mastery (ex. Qt of Adept, Grant of Blood Legion, Holybeatrix of vodka, Nichts of Dues Vox). I don't believe a Retribution's selection of ancillary support talents correlates well with DPS output in practice. You should continue to spec for whatever your guild needs. For me it happens to be Vindication because it allows our tanks to pick up more threat talents.

The choice of AM or DSac has little to do with dps and more to do with what the raid needs and personal preference.
I cant see anyone picking up DSac in 3.3, as Brekkie said there is no comparison between it and AM without going into DG.
I drop Heart of the Crusader as there is always a combination of either master poisoner, totem of wrath or our prot pally who has picked it up.

There are many target swaps in ICC and ToW will not always be in range of the adds.


With the 3.3 LoH debuff change, it will be a bad idea to LoH tanks (since a Holy Pally cannot LoH them). LoH on a random non-tank is fine.

Yeah....no. The add will have to be about 70-80 yards away from the other target, not being tanked by our pally and not being hit on by the rogue(s). I think I'll take my chances.
The only other places to drop points would be swift ret, vindication or PoJ, none of which I feel comfortable removing as there are 3(4) other sources of 3% crit.

Few new notes from MMO-Champion -

Holy
Sanctified Light: This talent now also has a 33/66/100% chance to prevent Lay on Hands from causing Forbearance when Lay on Hands is used on others.

Retribution
Repentance: This crowd control effect will no longer break early from the damage done by Righteous Vengeance.

I still doubt anyone will take Righteous Vengeance in a PVP build though.