Unless I'm missing something, Shadowcraft's calculations for the legendary meta gem are way off - it's showing the legendary would give me 21k dps.
As far as I understand it, we get 324 crit, and 'some dps'.
The damage part of 'some dps' is d = 280 + 0.75*AP
The 'per second' part of 'some dps' comes from the RPPM mechanic. With a RPPM of 32 (that's 21 * 1.535 multiplier for assassination) and at 23.5% haste, the average time to proc is 1.5 seconds. You get the damage when you get 5 procs, so on average every 7.5 seconds.
so, 'some dps' = [280 + 0.75*AP]/7.5.
Say you have around 50k AP, the value of the gem comes to 324 crit + 5014dps.
Alas, as the hits from the gem can crit too, assuming 20% crit chance, that nets us about 1k more dps, so with all these assumptions, the napkin math here gives us a value for the gem of 324 crit + 6018dps.
I hope I've missed something and that the legendary gem is not as bad as I think it is. I look forward to people poking holes in my maths/logic!
Edit 1: used the right RPPM (21 instead of 15), used a more realisitic figure for AP in raid conditions (50k instead of 35k). Thanks Pathal for the corrections.
Edit 2: added the chance to crit. Thanks Viper for input.
Probably just caught in a loop somewhere and I missed it.
EDIT: I'm not sure your base information is right. It's 21 RPPM, but that's stacks. 5 stacks is a proc. So the damage proc is 21./5 RPPM. The Assassination modifier is 21./5*1.535= 6.447 procs per minute base line. With 23% true haste, that's another 21./5*1.535*1.23= 7.92981 procs per minute, or 7.56638557544s intervals. Additionally, I'm pretty sure you will have more than 36k AP in a raid, lets say 50k for now, but I don't know your comp or whatever.
50000* .75 + 280 = 37780 / 7.56638557544 = 4993.13703. Still lower than what Shadowcraft is telling you, but twice as high as you estimated. I think it's because you confused the 15 RPPM for classes not designed to use the trinket, at least, it would explain why you pulled 15 out of the air instead of 21.
EDIT2: Think I got it. Looks like an old bug resurfaced when I used default values for the RPPM proc, caused it to proc ~3 times as often. I'll probably come up with a better solution next patch.
Right in both AP (I was using my outside of raid AP) and RPPM (using the wrong number as you predicted) - I've edited my previous post to correct for these mistakes. I think you might have divided by the wrong number when calculating dps (you divided damage by (procs per minute) instead of by (time interval)), but otherwise we seem to be getting the same result now.
Edited to account for changes made.
In terms of the proc: I glanced through the proc effect code and while I admittedly don't keep up with shadowcraft too much, it looked like it was using the spell crit rate rather than melee crit (due to it being a 'spell damage' proc). The fact that it can crit also should increase the "napkin math" estimate as well.
I always hate guessing what will and wont follow in game patterns. I feel like there were some procs / mechanics in Cataclysm that didn't scale with BG, but I can't remember any of them to save the life of me.
Regardless, I'll fix the BG method to stop it from using a strict whitelist. It's probably more likely that abilities will scale with it in Mists, than it is that they won't scale. I do seem to remember the buff on Shek'zeer scaling with BG.
ICD trinkets are a bit funny with RPPM. The average proc interval starts when the trinket procs, the ICD is basically just a protection to prevent chain proccing while the buff is already up. Where things get a bit odd is the "Time since last proc chance" factor continues to count even the trinket is on ICD so the first attack after the ICD ends will have 10 times typical proc rate.
Are you sure about this? That seems like kind of an odd way for them to implement it and would be a massive increase in uptime. (like, renataki's going up to over 50% proc chance by the time the ICD is off)
It seems more likely that last-chance-to-proc would still be based off every hit and that the trinket simply can't proc while it's on ICD. Keep in mind that last-chance-to-proc isn't supposed to regulate how often the trinket procs, just offer a multiplier to balance out many hits per second with very few hits per second. The actual proc regulator is the new last-proc and average-time-between-procs they added as bad luck protection.
It's basically what happened in the earlier tier already. The secondary stat values inflate the bigger they get, and they are also influenced by the other two secondary stats. As combat, haste EP are going to drop below 1/2 Agi, which in turn means that by gemming agi instead of agi/haste, and agi/haste instead of pure haste, you can counteract that inflation, but only to a certain amount. In BiS gear one will most likely keep the mastery and haste EP values as close to each other as possible.
What happened in your case: Mastery was sitting slightly higher on the EP than haste, while probably most of your gear was still forged to haste. Shadowcraft then reforges every single piece towards mastery instead of haste, resulting in a very high amount of mastery, and a low amount of haste, which in turn results in these massive EP spikes between haste/mastery. Right now it is advisable to change sockets from a haste prioritization towards agility. You can use shadowcraft to do it gem for gem, until you reach the point where haste is still slightly above mastery. You will have to do this for every new piece you get though.
You can use shadowcraft to do it gem for gem, until you reach the point where haste is still slightly above mastery. You will have to do this for every new piece you get though.
I have actually tried this and a jump of the size Kisla mentioned occurs by changing a single gem once you get close to 7.2k haste. Surely something must be going wrong there. Especially seeing how we already went above 7.2k haste in the previous patch and back then Shadowcraft wouldn't make the EP values jump around at that point.