Average of like 27% uptime on Bad Juju across all our H Council pulls. I'd say that's pretty good!
Average of like 27% uptime on Bad Juju across all our H Council pulls. I'd say that's pretty good!
So normal Juju > 510 bottle, would you say? Sounds nice to actually be able to use normal trinkets aswell.
Hi. I would like some help. I'm bad at theorycrafting and would just like a few replies or solid answer. Just got 522 bad juu juu. Replaced terror in the mists 500ilvl and using relic and bad juu juu. Think that is appropriate choice? Also when comes time for second trinket, Think the 522 valor hit trink is better then relic? Im not certain ill be able to reforge all the hit away but can get pretty dang close. Just sounds odd that chance hit trinkets with high agi proc will be out solid agi at all times.../shrug
So based on the spreadsheet, only normal Rentakis (thunderforged) beats HC Bottle 2/2 upgraded? Seems blizz screwed that one up
how does bad juju compare to the valor trinket? I'll try to get a heroic rentakis to replace my h bottle
I removed public access to my spreadsheet until I get new math. It's not accurate at all ATM. Sure, 10% increased chance is easy to put in, but I'm not touching the "unlucky streak protection". Apparently, RPPM trinkets' uptime got raised by...say, 40-50%? Pretty crazy.
Old Gods made me do it.
How is your calculations/spreadsheet compared to those of Conjor? Here's his topic!
It's same basic formula for no icd trinkets, actually - except I don't process "confidence". The thing is, I think that "Unlucky Streak Protection" is really important, but he considers it of no consequence.
And his spreadsheet doesn't consider ICD, while they DO matter.
---------- Post added 2013-03-15 at 01:14 PM ----------
Meh, I'll just put that 10% modifier, and make it public again. But keep in mind that real uptime will be like 20% higher judging by the logs that I saw.
Last edited by Thiron; 2013-03-15 at 09:08 AM.
Old Gods made me do it.
Im not really good with Mathematics so i woud like to ask I have Renataki's 522,Hc bottle 510 and VP trinket I'm using atm VP+Bottle should i switch bottle with renataki's 522?
I got LFR Rentaki Tuesday and went back to LFR last night with some friends for the hell of it. On the 3 boss fights in there Rentaki was up 23, 27 and 27 percent.
With blanket assumptions on the icd and proc rate, all you need to do is add the icd to the expected interval between procs and use the same formulas.
The thing is if you want to capture the entire picture and put into into a continuous framework you will need to use some form of confidence. Trust me, I've tried. The math either explodes to where you can't get a continuous solution in the real domain or becomes unmanageable.
--EDIT--
I'll try to explain in a bit more detail.
As you go longer and longer without a proc, whether it is with an icd or not, you slowly increase the instantaneous rate of procs (the "protection" they added). However, that increase does not affect the previous procs, nor does it increase the "average" proc rate across the board, nor does it affect the proc rate of future procs. This is because the procs are inherently dependent on each other. This means that you need to look at each proc independently of the rest, and go from there.
If any website out there is saying the proc rate went through the roof (50+% on some) they assuming to many things about the independence of the variables at play.
For example, lets say we are in the middle of the fight somewhere, and I just got a proc. Great, now my (time since last proc) is zero. My chance to proc is now (RPPM) until t=3/(2*RPPM), (that is, we have gone 1.5*the expected time before we should have gotten a proc). Then it starts going up linearly by some factor, call it a*t*RPPM. As soon as it procs, all this gets reset. We go back to our base frequency, ect. It has no effect on procs that happen between t=(0, 3/(2*RPPM)] nor any future procs.
If you truly want to do anlysis with the changing RPPM rate, you need to go into nonhomogenous Poisson distributions, and limit your K=1 or 0. At that point you need to use confidence to figure out what the maximum time before the next proc will be. There is no way you can find an "average" proc rate because they are dependent on eachother.
I should note that I havn't tried to apply Cox processes to this yet. But tbh that seems like major overkill.
There is a reason Blizzard uses simulations all over the place. It is because finding closed form solutions like this are often impossible / stupidly difficult.
One reason for the difference between the average uptime numbers produced in game and the math for general average uptime is that the protections put in to improve the long time between procs have a disproportionate effect on the beginning of a fight. So the average uptime over the course of the fight will be better, especially on shorter fights. This is because there is usually at least 3-5 min between pulls, which means every RPPM trinket will have a high chance of proccing early in the fight. After that early fight you'll see a more normal uptime scenario. This is more noticeable in the trinkets with very poor proc rates like Bad Juju than it would be on a trinket which would normally proc more frequently.
I haven't updated my site yet with formulas which attempt to reproduce this case, but this is probably the bulk of what people are seeing. I would expect to see a small uptime increase over the course of the fight independent of that, just because there is a boost in the case of poor luck but not a delay in the case of good luck, which I think would result in a very slight trend towards higher average uptime, but advanced formula math like this is not my strong suit and I'm not really sure how best to incorporate that into a formula.