Open Office - sorry, no. I have immediate access to Microsoft Office 2007, which is what I use. Microsoft, being Microsoft, have more or less intentionally designed it so it does not play well with others. Having neither the time, nor energy, to determine what proprietary features of MS Office don't function in Open, I'm afraid it will continue to function only in MS Office. For what it's worth, I do endorse SimCraft. They're good people and actually perform rigorous testing in-game, like I do, to ensure their models are correct.

Gurthalak - several people (who have MS Office) have double checked my work and found I was double-dipping. I accidentally had both CS and DS (should only ever be one exclusively) as potential procs for Gurthalak. That inflated damage by approx 22% (in my gear, which leans towards Haste). I'll upload a bugfix version when my host is no longer being flaky (unspecified error on upload at the moment).

So it's actually LFR Gur, Slicer, Ataraxis, H Sulf, N Gur, H Slicer, H Ataraxis, H Gur.

For comparison N Ataraxis is about 200 DPS above LFR Gur. Enough that I would recommend the upgrade, though H Sulfuras is a further 900 DPS beyond that.


Edit:
Host no longer flaky. Updated spreadsheet: here.

Gear, buffs, trinkets can all influence the situation since a huge portion of its total damage contribution is the proc (the rest being from base damage/weaponspeed). In a raid with less than full buffs or using trinket A instead of trinket B the tentacle could be a lesser or greater proportion than what my personal gearset produces.

I'd hazard a guess that the worse your gear, the better Gur proc looks in comparison. Odd, since your specific gear is at least comparable to mine (barring your trinkets) - may be Human vs Draenei. You're seeing LFR Gur about 250 DPS greater than my gearset, which is well within the realm of possibility. If the spreadsheet says Gur is better for your gear, I'd believe it, although upgrading gear (or reforging better than the built in roughshod system) could change that - and 40 DPS probably isn't even a noticeable margin these days. If you didn't already have Gur I would say not to ignore a Slicer/Ataraxis drop.


For those who didn't see my edit above - bugfix spreadsheet was released.

Thanks for taking the time to explain how it all comes together - I appreciate it. At the very least, now I don't feel so bad about passing on the (N) Slicer that dropped last week so our tank could grab it for his offspec.

(And yes, I've had awful trinket luck since Firelands started. Hopefully that'll change soon, though!)

Comparing where you are and your results and where I am with my results, I'd say a good rule of thumb would be that h slicer and h ataraxis surpass normal gurth around the time you're 395-397 ilvl. That's roughly the in-between point ilvl wise for the two of us, and I'm getting similar results to you only on the other end of the spectrum (I'm finding h slicer and h ataraxis as dps gains instead of losses). I could see arguing that its worth picking them up earlier if possible as well since your ilvl will continue to increase over time.

Checking OP and searching recent thread posts I see no mention of how to handle heroism, especially zealotry plus heroism. Checking some of the top Ultraxion heroic logs the trend seems to be to go to a 1 filler rotation during zealotry with heroism. Personally under heroism with 700 or so haste I have a 2.85s CS cooldown, and as both TV and CS have 1.5s GCD a CS/TV only rotation has no gaps. Without Zealotry active I have a small gap after each spell GCD (due to spell GCD going down to 1s during heroism) - I'm undecided whether I should get enough haste to obtain a 2.5s CS cooldown during heroism and use a 1 filler rotation outside of Zealotry, or go with 1 filler rotation during zealotry+heroism, and then 2 filler rotation during just heroism, or always use a 2 filler rotation. The latter appears to be what Exemplar does in his own logs, but doesn't match what a lot of the top ranked paladins seem to be doing.

When Heroisem:
-without zelotry(t13 2pSet); CS > J > CS > TV > CS > filler > CS > J > TV ...
-with zealotry(t13 2pSet); CS > TV > CS > TV ...

this is what I use. And whitout heroisem I use 2 filler rotation.

As listed in the first post under Haste and the Soft Cap, every test I've performed of:
CS->Something->0.1 seconds of nothing->CS
is less DPS than:
CS->Something->Something Else that pushes CS back 1.4 seconds->CS

Even during Hero.

Also as noted in that area, to the best of my knowledge I'm the only one to actually perform such testing. Thus it is more a Theory than a Scientific Law as it has not been externally validated. Similarly there has been no countervailing proof or mathematics. Anyone going with a 1 filler rotation is presently following their gut, which on rare occasions is correct. If someone believes that 1 filler is superior, I would love to hear with what tool they've performed it or see the hard math here on the forums.

Again - 1 filler could be superior, but I have never seen any proof or even testing that displays it is. I don't claim infallibility, but do claim more than anecdotal 'Well, person X does it in this fashion.'

Edit: I suspect any attempt to prove one or the other superior either on bosses or test dummy will fail - hidden within the massive RNG of our spec (Crit, DivPurp, AoW, weapon procs, trinkets, etc). Paper or simulations may prove the truth, but it'll be hard to find in practical application, which is downright annoying and frustrating.

Jagro: TV, using the terminology of the haste section in the OP, is a filler for the sake of this discussion.

Exemplar, you're restating that introducing any dead time into the rotation is a DPS loss, but I don't see the applicability of that statement to herosim. During heroism you will always have a sub 3s CS CD (you need 105 haste rating or thereabouts to reach 3s CS CD with WoA/SotP/hero). So the only time you're given the choice of sitting around and waiting for possible dead time is following a spell GCD rather than a melee style GCD.

Melee filler:

00.00 CS
01.50 CS GCD ends, melee filler GCD begins
03.00 melee filler GCD ends, potential CS GCD begins.

Spell filler:

00.00 CS
01.50 CS GCD ends, spell filler GCD begins
02.50 spell filler GCD ends, either cast another filler or wait for CS
02.xx potential CS GCD begins

If we agree on the premise that having any dead time at all is to be avoided at all costs, that still leads to the conclusion that we should use 1 filler during heroism unless that filler is a spell, and if it's a spell then we should use 2 GCDs. It also leads to the conclusion that during Zealotry + heroism, when we have the ability to make all GCDs potentially melee GCDs, we should be using 1 filler (TV) as this does not introduce any dead time.

It's not quite so simple as that: if you instead stick to 2 filler at all times you do actually cast more abilities in the same timeframe, thanks to the spells being 1s GCD. You cast a lot fewer CS, and consequently TVs, though, due to losing 1/3rd of a TV to gain a 1/15th chance at TV for every CS that you replace with a second filler (again, not so simple as that considering judgement with 2pt13 as the filler but anyway). A 1 filler rotation (2 filler if spell) would net you 28-29 GCDs during heroism, while a 2 filler rotation will net you a couple more depending on RNG.

I don't know if your spreadsheet takes all this into account. Lacking MSOffice I'm afraid I can't have a look for myself. I'd have thought SimCraft ought to be able to handle this though; unfortunately I don't believe Theck's matlab project supports variable CS cooldown and haste's effect on spell GCDs as the first isn't applicable to prot and the second has negligible effect for prot due to only having extremely weak and low priority/long CD spell-gcd abilities.

No, I can confirm that the MATLAB code doesn't currently support either of those effects. However, it wouldn't take a lot of modification to do so. This point hasn't gone unnoticed by me; in fact, I've been toying with the idea of expanding the code to support Ret in 5.0. Having the FSM code as a cross-check, particularly for things like the rotation comparisons, would probably be a valuable tool.

That said, there are some difficulties involved. For starters, the FSM code currently runs with 3 time steps per GCD. It supports more than that, of course, but the state space (and thus processing speed) grows considerably as the time resolution increases. I anticipate that for ret, you'd want something on the order of 0.1s or 0.05s resolution, or 15-30 steps per GCD. We've made a lot of optimizations to the FSM code, but that might still take quite a while to run.

In addition, dynamic haste effects are problematic. They'd have to be coded directly into the FSM, on an effect-by-effect basis, and they'd inflate the state space significantly. However, we may be able to ignore them, and just evaluate rotations at approximate haste values (i.e. with and without BL, possibly at several different benchmark haste values).

Finally, there's also the matter of adding all of the Ret abilities, talents, and what not to the code base. Not a particularly challenging problem, but time consuming nonetheless.

However, as much as I'd like to add retribution support, the truth is that my job and life have gotten busy enough that it's unlikely I'll have time. If there's someone interested in giving me a hand with the 5.0 code, preferably someone proficient (or at least familiar with) MATLAB and C#, there's a much better chance that we can make it happen.

Actually, not everyone pushes sub-3 second CS CD during Hero. The camp that is very happy with Crit can still reforge so that their CS would be slightly above this value. That said, it should be a sliver of time beyond 3 seconds (rough modeling was around 3.05) and not worth worrying about.

So, the terminology of "filler" seems to be confusing people. Our core rotation (yes, rotation) is Crusader Strike and Filler. Filler is defined as anything that takes place between Crusader Strike (the filler is where we get into Priority). Void of any Haste whatsoever, this filler is 3 seconds long (4.5 seconds - 1 GCD for the CS itself = 3 seconds). As Haste is added the filler space grows shorter.

In normal (non-Heroism) circumstances the filler is greater than 1.5 seconds and less than 3 seconds. This provides the option to use 1 GCD and dead time, or 2 GCD and delay a CS by some tenths of a second. My testing has found that a filler space as low as 1.6 seconds should still be used for 2 GCD, even though it pushes back a CS by 1.4 seconds (3.1 - 1.5 CS - 1.5 Fill1 - 1.5 Fill2 = negative 1.4).

Now we get to the special circumstance of Heroism. In Dragon Soul gear this will drop the CS CD to somewhere around 3 seconds - perhaps a fraction less or more. Pretending we get this perfectly flat value of 3 seconds, your filler becomes 1.5 seconds (3 - GCD = 1.5). For melee abilities this is perfect for 1 ability with absolutely zero delay. Ideal. For spell abilities (of which we use relatively few) this leaves 1 second and 0.5 seconds remaining. Again, my testing finds it is superior to use this 0.5 seconds for another attack (assuming one is available! It will not always be so!) rather than leaving it as entirely empty space.

Thus under Hero we would neither be "2 filler" nor "1 filler", but more of a "1.5 filler" hybrid scenario. My main point remains that leaving dead space is detrimental. Adhering strictly to 2 filler or 1 filler should both be inferior to tailoring it based on the circumstances and minimizing dead air.

The above was napkin math for clean examples - a lot of the time any potential second filler is actually a spell. The spellGCD being a flat 1 second to perhaps 1.2 seconds means less overall CS loss. An 0.1 second dead air usage pushes CS back by 0.9 to 1.1 seconds rather than 1.4 seconds. I suspect this is part of the reason my models continued to incline towards filling all dead space.

Yes, Simcraft should be able to model this. You would need to heavily modify the entire priority setup with references to remaining time on CSCD to force the 1 filler scenario - I believe the default would use 2 abilities if there is even a fraction of a second remaining when a GCD ends. I've done some heavy modification like this before in Simcraft as their timing for trinket and AW/Zeal usage is (or was when I did it) far from ideal (like popping AW/Zeal just before Inq expired or using a click trinket when CD completed and AW/Zeal were not ready due to HP/Inq reasons). I honestly do not recall if I modified to test the filler situation. Either it matched my spreadsheet (0.1 dead air is bad) or I did not test - for the community I'd go with the 'did not test' choice. I know the tool can support this and provide reasonably accurate results. If someone decides to test, I'd enjoy them posting the spell logic so others could copy/paste it into Simcraft themselves or further hone it for accuracy.

Again, filler is space between CS. If you have 1.5 or less usable seconds between two Zealotry CS then the only possible rotation is CS, TV, repeat (barring a DivPurp that requires further TV be used). If there is no free time then there is nothing to fill.

Not using TV would mean the next CS would generate wasted HP (3 of them!) and using two filler abilities (one being TV) would be delaying the next CS (and all subsequent CS) along with its HP (again, 3 of them!) by a full GCD. You must use the TV or you cannot use another CS.

It's simple math. 1 empty GCD = 1 filler. 1 empty GCD != 2 filler. I'm not sure how a 'top ranked paladin' could do more DPS by delaying CS and TV with extra fillers under a Hero+Zealotry scenario. I'm really unsure why there is even a question about Hero+Zealotry.

Hero without Zealotry is an entirely different matter. With spells at 1 second GCD, your 1.5 seconds becomes:
1.5 empty spellGCD (1.5 seconds) = ???. My tests showed the answer was 2 fillers (one spell, one anything). This is debatable and the debate is open.