Get the Desktop App for Battle.net Now
- All your games in 1 place
- Log in once
- Automatic game updates
"RealPPM Functionality Changes
We changed RealPPM to use your true melee haste or spell haste (whichever is higher) to scale the proc rate instead of attack speed or casting speed. The difference is that things that buff your attack speed (such as the raid buff or Slice and Dice) do not increase the frequency of RPPM procs. True haste boosts, such as Bloodlust or haste rating on gear will improve proc rate. We felt this change was necessary to balance proc rates among different specs."
This log is without unholy frenzy, bloodlust, or any external haste buffs.
On live servers logs are reporting rppm, specifically spark of zandalari, but most likely others, scaling with the 10% melee attack speed raid buff for dks.
Example log: http://www.worldoflogs.com/reports/rt-dry7m82z5k36zow5/details/0/?s=3&e=390
You'll see 5 procs of the Zandalari Warrior buff, which means at least 50 procs of the spark (log says 47, but logs are clearly not perfect right now), the person creating this log reported 5 procs of the trinket, so I can assume that much is true.
50 procs over a 6.5 min period means 7.69 ppm, which divided by the base rppm of 5 gives a 1.54 haste effect, while the player in question has 1.58 attack speed, and a 1.44 meleehaste, the rppm seems to be using the attack speed, not the meleehaste.
This log shows 2346 seconds (39.1 minutes) of megaera, a fight with no fight-specific haste buffs. Overall, that's 8.43 ppm, or an average haste effect of 1.69.
This raider has, once again, a base "attack speed" of 1.58, and a "meleehaste" of 1.44 (1.58/1.1). Dks benefit from haste rating, (his being 19.68), unholy presence, (20%), and the raid buff unholy aura (10%). 1.2*1.1968=1.44, the expected "meleehaste" if UP and haste rating count, while the raid buff (1.43616*1.1) brings you up to 1.58, the amount reported on his character screen under melee haste.
Now, the the 39.1 min period is 7 total tries, each try lasting at least 40 seconds, so for 280 seconds, he had a 30% "true haste" buff from lust. 280/2346 = 12% uptime, and during that uptime he had 1.44*1.3 = 1.87 "true haste", and 1.58*1.3, or 2.054"attack speed" The actual log only shows lust cast twice, but again logs are not perfect, or at least WoL isn't.
Unholy frenzy was used twice per fight, and on the 7 minute fight it was used 3 times. That's 15 total casts, and it lasts 30 seconds total. 450 seconds /2346 seconds = 19% uptime. Unholy frenzy is 20% haste, or 1.44*1.2 = 1.73 "true haste", and 1.58*1.2= 1.90 "attack speed".
Now, I realize that they stack and effect each other, but that shouldn't throw the results very much.
.19*1.73 + .12*1.87 + .69*1.44=0.3287 +0.2244 +0.9936, or an average truehaste of 1.55 throughout the fights. It will be SLIGHTLY higher if you take buff stacking into account, but not by much.
While, for attack speed:
.19*1.90 + .12*2.054 + .69*1.58=0.361 +0.24648 +1.0902 = 1.70 attack speed.
Remember, the actual proc rate recorded was with an haste effect of 1.69. It seems as if the game is still taking the 10% raid buff into account for rppm.
However, I'd also like it to be brought up that, currently stat values say that a spark of zand (ilevel 522) is worse than a relic (ilevel 476), at current haste levels.
Simcraft is showing lower dps, and simcraft uses truehaste, not attack speed, and the stat values we have multiplied by the proc and uptime, added to the static amount, give the trinkets a lower value. This is not the case if the raid buff is added in though.
Edit: To be more clear:
The proc rate we are seeing in-game is higher than expected. It is nearly exactly what expected if you put the 10% haste raid buff into your haste calculations. Every single report and log I've been given of spark show it having a much (3-10%) higher uptime than expected, expected number comes from (5(rrpm)*true haste* 20(duration) / 10(stacks needed)/ 60 (seconds per minute), and is confirmed by sims.
"Fixing" this would, however, make spark worse than t14 trinkets. Is this intended?
Edited by Mendenbarr on 3/7/2013 11:29 AM PST
This same thing is reflected in the rating of trinkets on AskMrRobot.com, too. Based on the system of RPPM, it seems as though the Best-in-Slot for dual-wield Frost is actually the 502 Lei Shen's Final Orders, alongside the Brutal Talisman of Shado-Pan Assault. This raises two questions, similar to the question being asked in the original post:\
1. Is there any intention on reworking the manner in which these trinkets function to bring them to the point where their ilvl actually reflects the quality that it should as compared to those from earlier content. And,
2. Have a trinket from reputation simming higher than those from the actual content adds to my concern of the quality of these trinkets. Is this, too, an intentional addition.
This, of course, is coming from a dual-wield frost perspective, however, I have heard similar concerns from protection Paladins and Blood Death Knights concerning point 2.
I did obtain Spark two nights ago, but I have not tested it extensively. As it stands, I cannot afford to stack any more haste to make this trinket better, and as such, I might switch back to Relic.
A few preliminary results:
As DW frost, spending an hour on a target dummy just auto-attacking net me 19.9% uptime on Spark of Zandalar and 15.6% uptime on Fabled Feathers. I had ran another test but wasn't sure of the total time (somewhere over an hour) and had 19% uptime on Spark and 16.8% on Feathers.
For 2H, I got slightly higher uptime over the hour. 19.9% on Spark of Zandalar and 20.6% on Feathers.
For blood, I only ran a 30 minute test but had 22.2% on Spark and 23.5% on Feathers.
All of these were sitting at 6069 haste right now.
Council of Elders (Spark of Zandalar only):
Tortos (Spark of Zandalar only):
Maegara (Spark of Zandalar only):
Jin'Kun (Spark of Zandalar only):
Durumu (both tirnkets):
Dark Animus (both trinkets):
Iron Qon (both trinkets):
Expected uptime is about 20% on a 7.5 min patchwerk fight (from napkin math and confirmed by simC)
Actual uptime is:
20% on 6:53 fight
26% on 6:26 fight
21% on 8:10 fight
24% on 8:03 fight
21% on 8:03 fight
24% on 4:12 fight
25% on 10:55 fight
Expected uptime is about 22% on a 7.5 min patchwerk fight (from napkin math and confirmed by simC)
Actual uptime is:
25% on 8:03 fight
47% on 4:12 fight
35% on 10:55 fight
Edited by Mendenbarr on 3/11/2013 10:58 AM PDT
I've received reports that trinket's uptime and consistency has changed over the past few days, perhaps to adress this issue, which I applaud.
However, it's very hard to model these trinkets correctly while the information about them seems to be both in flux, and not accurate to the original numbers presented. It would be greatly appreciated if you could give us a copy of the new numbers when you finalize them, or are at least happy with the current iteration.
Edited by Mendenbarr on 3/11/2013 9:30 PM PDT
RPPM trinket formula has changed today, (well notes came out today, it changed yesterday), so I will revaluate expected uptime with actual uptime and see if this remains an issue, or has been fixed. MUCH harder to napkin math out uptime with this change though.
Or someone could just say "We're aware of this and have fixed it/are working on it/don't care"
Edited by Mendenbarr on 3/13/2013 4:03 PM PDT
Threats of violence. We take these seriously and will alert the proper authorities.
Harassing or discriminatory language. This will not be tolerated.