I’m looking for a sanity check on the veracity of my below observations on set conversions and whether anyone has any tips for forcing the game to actually give statistically expected RNG.
Background:
I’ve done at least high-hundreds if not thousands of set conversions and I consistently encounter 2 different scenarios: either I get a reasonably random sequence of items or it oscillates between the same 2 or 3 items over and over and over again.
It’s especially noticeable when I already have 4 set items and am trying to get a 5th. For each item conversion of a 6 piece set, I should have a 40% chance to get one of the 2 remaining items I don’t have. This means that if I do 5 set conversions, I should have a 92% chance, and if I do 10 set conversions, I should have a 99.4% chance of getting at least 1 of the 2 remaining missing items.
However, every season (for the last 4 seasons), in multiple regions for multiple accounts, I will do 10+ set conversions in a row (when I already have 4 set items and am trying to get a 5th) and I will not get the 5th one because it keeps oscillating between the same 2 or 3 items I already have. This should only happen 1 out of every 167 sessions, yet it happens to me every single season in multiple regions for multiple accounts, so there seems to be something seriously wrong with the randomness of set conversions.
K, here is what is really going on. Blizz uses Skynet caliber AI with the full intention of screwing people over. It knows exactly what you need and will not give it to you.
It is the same when I want to roll elemental damage on my bracers. I will get all the elements I don’t need at 20% multiple times before I get the one I do need and it will be only 16%.
The only way to defeat their Skynet AI, other than sending someone back in the past to destroy it, is to use reverse psychology. Try to not get the one you actual want, and you will immediately get it.
I’m not sure the math is correct. You can try it with a 6-sided die. How many rolls does it take to get a 1 and a 2?
Use a shaker cup to maximize randomness.
It’s a 5-sided die because the item itself is excluded from the pool of possible items it can be converted to. I used https://homepage.divms.uiowa.edu/~mbognar/applets/bin.html to calculate the probabilities (thanks to BehindTimes for recommending this site).
I understand that unusual behavior is possible for small sample sizes. However, as sample sizes get larger, RNG results should begin to converge with expected statistical results. I believe what I described above is a large enough sample size to start to draw some plausible hypothesis about unequal weighting (or possibly even dynamic weighting) for set item conversions, but I acknowledge that I am not a stats or math expert, so I wonder what others think would be a large enough sample size to introduce reasonable doubt.
You still have 60% chance to fail it. When you calculate it that way you’ll see that pseudo random number generators are not kind or fair. Since PRNG simulates a total random setting, each attempt comes down to 60% failure to 40% success regardless of previous ones failing or not. If RNG were to add a pity counter you supposed to get it in 2 out of five tries each time; there’s no pity here because it’s not pseudo random but a pseudo random number generator that simulates total randomization.
For example failing ten times with 40/60 is in the possibility of 0.6^10 = 0.0060466 ~ %0.6; that number only represents its occurrence across the player base and how common is this. That number only represents the percentage of player base who went through failure 10 times when they have odds of 40% to succeed. However, a failure of ten times still possible at the rate of one in 165 players or attempts with 60% failure. That doesn’t mean it ain’t gonna repeat or chain up higher to twenty or hundred.
Each of your failed attempts do NOT add up to the chance of success. True randomization can easily generate the same result with small variations on minor variables because it’s on the verge of possibilities. True randomization don’t care if an outcome repeated in the last few attempts or not. Check your calculations and think about the possibility of failure instead of success.
The examples you list, did you actually track all the rolls off 100 or so samples, or going off what you remember?
I have had this happen with Urshi. I swear I get far, far more 5 misses at 60% than 5 upgrades. When I have actually charted 60% upgrades over a few hundred, the upgrade chance is usually between 58%-62%, and the strings of 5 misses almost the same as 5 upgrades.
Convert set frustrates me too as I have rolled 20+ times to get piece I actually want, but it doesn’t always happen that way. Next season, we have 2 conquests Years of War and Masters of the Universe that will take me 32 sets to do them all. I will keep track and give my results, track yours and we will see if we have a Weighted Convert Set conspiracy or not.
I spend one time about 1 billion in gold and still didn’t get the roll I wanted. Look at some of the gear and what is on them you can roll. On some of the gear you have up to 20 different items you can roll. More than half of them you would never ever use in a million years. But the stuff is there.
Just yesterday I needed a socket on a ring. It took me about 60 ties to get it. One thing I do is walk away and even leave the game. Then go back in and try again. Yes sometimes it works and sometimes it doesn’t.
That is just like Kadala and the Altar. It says when you do this item in the Altar. You have better chance to get more leg item. On all 4 accounts in the family it isn’t working. We are all seeing the same thing. About 95% of the time we are still getting about 99% yellows & Blues.
But it is a video game and you really have to live with what you get. The Good, Bad and the Ugly!!!
But, I think that @yessquire’s point was that it doesn’t matter how many times you flip a coin or roll a die, you rarely get what you want.
Some days you may succeed most of the time. Some days you fail miserably.
I just hate that Blizzard weights certain aspects of the game. (Items, Properties, Stats, etc.) I mean, RNG is bad enough. But, “stacking the deck” or “loaded dice” is worse.
Ya, I don’t play Wiz much. But I did in S28. Tried to convert a set piece like 20 times to get Tal boots, then looked it up and saw the set doesn’t have boots.
I noticed this too in S28 when I was going to try to make all solo LBs. I charted most of it this season for some of the classes. I stopped when Blizz’s crappy servers killed my HC DH. Most gear pieces are around the 20% legendary rate for Kadala. A few gear pieces for a couple of classes are a lot lower. Think I am going to try again this season on SC and will post my results after the season.
Convergence to the mean is expected in a truly random system, but let’s not forget that practically all computational random number generators are pseudorandom, at best.
Confirmation bias, my friend. Confirmation bias. Like my working theory that Myriam gives me the rolls I want when I cuss at my screen.
Have you tried “using” a different Character Class? That always seems to work for me.
Just did a quick check with my “start of season” Necro and it worked like a charm for the DH gear. Did only one (1) test with recipe 4 and it was cycling through all 6 gear pieces (number 6 being of the type I started with).
“Just for fun” tried recipe 2 on a suboptimal DH ancient item - The Necro gave me one really useful ancient item on the first try!
Exactly. Which is why I said I charted it to actually verify if my crazy conspiracy theory was crazy or not. I’ve checked many of the “RNG broken” theories in the past. It keeps the season from getting to boring to fast if I have a goal.
I use this for all 3 Prime Evils (Kadala, Myriam, Urshi). BTW, what is your record 4 letter word combo bonus? Think I have gotten in the 20s before.
Each session you open supposedly give you a new hash information or a new seed of sorts for randomization. It can work but no guarantees. I usually try to land a low weight or rare sights of affixes to bargain my position with Mystic. If I need a socket, I try to land skill modifiers and it eventually lands a socket on possibilities. Everyone have their weird methods but it’s just randomization.
As a former football () referee, I’m well-trained in the art of expletives and I’m probably up into the hundreds. I’d love to post an inspirational NSFW song from Psychostick but I’d just get banned.
I expect they meant verify but autocarwrecked bit them in the derrière.
I shower Myriam with gold to get what I want. This season I was up to almost 54 million per reroll to get from 48% to 50% on a pair of Stone Gauntlets and it took days of persuasion (having a temporary stop at 49%). She’s indeed an enchanting, and expensive, acquaintance.
) referee, I’m well-trained in the art of expletives and I’m probably up into the hundreds. I’d love to post an inspirational NSFW song from Psychostick but I’d just get banned. 