[Elemental] SEIC Spreadsheet - v3 - Page 13

Caelwynar · 10/25/09, 12:42 AM

I don't have to construct a rotation at all, that's the beauty of doing it with a macro. Just write the priority logic to determine the correct spell to cast at the current time, increment the time by the cast time + latency of that spell, and repeat until the fight's over.

You have a point, though. At a 6.5 second cooldown, you will still never get 3 LvBs into a single flame shock, which means you'll have to clip the final tick of the DoT a lot more than you have to now, or lose some of the benefit of the reduced CD. Seems to support the initial assumption that the t10 set bonuses are practically worthless.

The macro version will probably take me a few hours of work to get set up (spread out over however many days it takes to spend that much time on it). I should have something up next week.

zimeron · 10/25/09, 11:31 AM

In a simulation you check to see if the Flame Shock dot will be there after casting lava burst, taking into account haste.

Here's an excerpt from SimulationCraft.

if ( ! p -> active_flame_shock )
        return false;

      double lvb_finish = sim -> current_time + execute_time();
      double fs_finish  = p -> active_flame_shock -> duration_ready;

      if ( lvb_finish > fs_finish )
             return false;

Caelwynar · 10/25/09, 6:39 PM

I've set up the basic code to determine which spells are cast at what times throughout a fight. There's still a lot of work to go (adding damage calculations, overriding certain settings for DPS comparisons, etc), but I'm able to see what the casting rotations look like with and without the T10 4-piece bonus.

With the 150ms casting delay and 871+200 haste values I use for Pahvel, you end up clipping the final tick of every single Flame Shock. The 6.5-second LvB cooldown and 18-second FS duration result in a completely consistent rotation (in my case, FS / LvB / 5LB / LvB / 4LB without heroism, 6/5 LBs during Hero). I won't have any simulated DPS comparisons for a while yet, but I expect them to be pretty depressing. The difference in a 6-minute fight simulation is +7 Lava Bursts, -6 Lightning Bolts, and +1 Flame Shock (with each FS being clipped by 1 tick), which according to the damage values for my character in the 3.2.2 SEIC sheet, comes out to a gain of about 94 DPS.

Long story short, T10's 2- and 4-piece set bonuses combined aren't as good as T9's 2-piece alone, and less than a third of the dps gain of wearing 4 pieces of T9.

Boomedale · 10/30/09, 9:34 PM

So Im a fricking nub and I cant figure out how to change the gear thats already on the spreadsheet. I've been playing with it all day and nothing. Can somebody give me a hand by chance?

Boomedale · 10/30/09, 9:40 PM

Nevermind Nevermind! Sorry... /sigh

Yoha · 10/30/09, 10:37 PM

Originally Posted by Zamir

Even if you resort to simulation, how would you actually construct the 4t10 'rotation'?

I think this may be what you were looking for (or at least sort of to give you an idea); [link] i made the spreadsheet that i linked on the mmo thread.

That is assuming:

1.4 LB cast time
1.05 sec LvB, CL, (and GCD).
- With T10 2+4pc bonuses
(with 0 latency)

Make sure you read the whole post that i made though

Zamir · 10/30/09, 11:08 PM

Originally Posted by Yoha

Make sure you read the whole post that i made though

Actually I was referring more to the problem of four-piece tier 10 reducing Lava Burst's cooldown to 6.5 seconds without changing Flame Shock's duration (discussed here). Two-piece tier 10 is discussed a bit later in the same thread.

That said, assuming LO doesn't trigger the 2-piece, 80 seconds seems ballpark for that haste figure. My own spreadsheet puts it at 84 seconds effective CD for 43% haste and Caelwynar's simulator macro thingy puts it at 88 seconds. As already noted the DPS value is not high (~100 in tier 9 stats), and anything which decreases the frequency with which we cast LB will reduce the value further - such as, say, the tier 10 four-piece bonus.

Boomedale · 10/31/09, 12:41 AM

So either I dont understand this or somthing. I change my hit on the first page and my dps doesnt change at all. Why? I throw up a huge number and it doesnt move. I change it to 0 and it dont move either. Am I missing somthing here?

Yoha · 10/31/09, 1:25 AM

Originally Posted by Boomedale

So either I dont understand this or somthing. I change my hit on the first page and my dps doesnt change at all. Why? I throw up a huge number and it doesnt move. I change it to 0 and it dont move either. Am I missing somthing here?

It assumes you're hit capped even though you may not be. - It hasn't been coded in.

Edit;

Actually I was referring more to the problem of four-piece tier 10 reducing Lava Burst's cooldown to 6.5 seconds without changing Flame Shock's duration (discussed here). Two-piece tier 10 is discussed a bit later in the same thread.

3-2-3-2-2 unless that last FS is out of time for .95seconds (due to adding LB before FS, to allow the extra 3. from looking at my spreadsheet.

Zamir · 10/31/09, 10:01 AM

Originally Posted by Yoha

It assumes you're hit capped even though you may not be. - It hasn't been coded in.

You can't really accurately formulate miss chance for an elemental shaman, as the effect of a spell miss can vary dramatically depending on which spell and when. SEIC calculates a DEP for hit based on 1% hit = 1% DPS which is a reasonable generalisation; if you really want to wildly guesstimate your DPS below the hitcap, just multiply SEIC's figure by 100% minus however much % you're short of the hit cap.

Caelwynar · 10/31/09, 6:50 PM

I'm not sure whether it's worth continuing with the macro version I was working on. I've got it to a point now where it accurately lists every spell cast (with damage) for the entire fight duration, and calculates DEP values, etc similar to Bink's formulas. The DEP numbers from both versions are practically the same.

I wasn't sure I trusted the "just calculate a single rotation" strategy for modelling DPS, but it seems to be close enough.

Zamir · 11/03/09, 9:54 AM

As I've said elsewhere, I think there are definite advantages to your semi-simulated approach - it's more robust and more versatile. If it gives the same result as another method most of the time, that's useful to confirm the validity of the other method. When the results differ, it becomes doubly useful.

But yeah, depends whether you can be bothered or not

◊ masanbol · 11/04/09, 6:44 PM

Getting some bizarre numbers occasionally with the updated spreadsheet. Occasionally my haste DEP values show up as negative when I change certain stats, usually spellpower and haste values. Increasing or decreasing haste rating by 1 sets it back to its normal value, but I'm not sure what in the calculations could be causing this.

Zamir · 11/04/09, 7:15 PM

What precise values will reproduce the weirdness?

◊ masanbol · 11/04/09, 7:28 PM

What makes this odd to me is that changing some glyphs or trinkets that only affect spellpower and not haste can cause it to bug out, making haste drop to -3 in DEP values. If it were related to some rotation weirdness you'd think that changing SP values wouldn't affect it.

Edit: attached a copy of my spreadsheet.

Zamir · 11/04/09, 8:26 PM

What's happening is that at certain haste values, the rotation calculator adds an extra Lightning Bolt in (or takes it away). If you're just shy of that point, the Haste DEP rotation calculator tips it over to adding an extra LB, which causes a longer rotation with reduced FS uptime and a lower ratio of LvB:LB and a slight decrease in DPS. In your example, this is happening during the heroism part of the rotation calc (in the most recent version of SEIC the heroism toggle doesn't work).

I had previously noted that SEIC has small DPS dips at certain haste values, as shown here:

However my understanding was that Bink had figured out a way to eliminate those?

[e] Found the post where that was mentioned: [Elemental] Patch 3.2
[e2] And again here: [Elemental] Rawr Module Development. Reading that may explain why certain spellpower values also cause it, as it affects the calculations which control adding the extra LB.

• Binkenstein · 11/04/09, 9:14 PM

I had.

You deflate the damage on the "extra lb" by the amount of time no FS dot will be up times tick damage divided by 3 (t*d/3), which offsets the addition of the extra LB to avoid that dip.

Zamir · 11/05/09, 12:19 PM

I think what you're saying is that if the rotation time (R) exceeds the duration of flame shock (F), you subtract the value of (R-F)*Flame Shock DoT DPS from the overall DPS of that rotation? If so, was this ever in SEIC? The present rotation calcs don't do that, and it'd seem counter-intuitive if they did as they presently keep FS DoT damage constant for the number of ticks regardless of the rotation time. (I really like the way SEIC builds its rotation, by the way - it's very impressive.)

This has the effect of causing a dip when adding a point of haste results in the rotation including one more LB than previously, which keeps Lava Burst and Flame Shock damage constant but increases the rotation time by the length of one more LB cast. This reduces the effective DPS in that rotation period of LvB and FS, which makes the "extra" LB a very slight DPS decrease.

The only way I can see for that not to happen is either to (a) confine the rotation to precisely the duration of Flame Shock and use LB casts as fractional values or (b) to artificially inflate the damage of that "extra" LB by the effective non-LB DPS of the rest of the rotation. The latter would presumably cause big jumps in DPS at certain haste values, which I suppose is what your method was trying to avoid (assuming I've understood it correctly). Both those approaches seem rather artificial, and at any rate the "dips" caused by minute haste variations in formulation are unlikely to be observed in real situations precisely because they relate to such incredibly specific variables.

• Binkenstein · 11/05/09, 3:41 PM

Ah fooey. I missed that out on the CL rebuild