Somebody phone the one developer working on D3 at Blizzard and get him to fix this issue ASAP please. While you are at it, please also send the laundry list of bugs and other requests that should be fixed/implemented.
Has anyone actually reported this to the bug forum? Is that even an appropriate venue?
I’ve noticed this, too. Thought I was just being superstitious. But if other folks have noticed it, too, maybe there’s something to it.
After this thread I tried a bunch of gambling and rates were around 15-17% for landing a legendary for me. Although example size is small, I’m somewhat satisfied after that because a small spread was expected.
Looking at numbers shown in this thread however, seems like something is arguably off here. If that’s what people’s experience from the game after unlocking Kadala’s buff from Altar, then I’d question the rates.
Jewelry and weapons are always expenny to buy from Kadala and they fail to give you a large example size to observe. So it’s natural that you feel something is off when you only buy rings or amulets from her.
A ring is two times the cost of an armor piece or off-hand, weapons are three times and amulets are four times the cost. If you can buy 60 armor pieces or off-hand from Kadala before a full dump, then you’ll have way less items when you attempted anything else instead of cheapest 25 bloodshard cost. This could give you illusion of something is off while rates are set straight.
It occurs to me that the issue is partly about the item weights that Kadala has gambling rates for bloodshards spent on the session. Some legendary items just refuse to appear unless you spend a threshold amount in a single session.
In short; be sure you are buying the cheapest item for an observable and large enough sample size, and be sure to hit GRs repeatedly without ending the session for any reason. By all means feel free to write a report and note the support with your findings if you really think something is off. If Kadala’s item weights could impact this, instead of doubling the chance, developers should simply half the cost rates.
If that’s true, that’s profoundly unfortunate. I’ve been trying to gear up my Wiz by running a GR or three on my main, then gambling on the Wiz 
Probably better to just play straight on the Wiz then, even if it’s far below T16?
Depends on how far below T16 you have to go. I personally wouldn’t go below T11, though preferably T13. But you should be able to get enough gear to do T11-13 for wizard from Kadala, upgrading rares (might require doing round or two of bounties every now and then) etc.
There are averages to obtain the item from Kadala, but those are only averages and weights appear to be equal for every item slot besides one or two exception per item type.
If there are, only limited amount of legendary items of the same slot available to player at the start of the session from Kadala (its seed or whatever) until they spent some randomized usual amount, then Kadala supposed to still yield legendaries with 20% chance at the cost of repeating the results. I just try to come up with a plausible excuse to why people encounter this, while it barely makes sense even for myself.
People usually do what you do. Run on their main then dump their shards by alts. I’m just trying to come up with an excuse why people still see 10% chance instead of 20%.
Jewelry and weapons are always expenny to buy from Kadala and they fail to give you a large example size to observe. So it’s natural that you feel something is off when you only buy rings or amulets from her.
That’s well understood, you’re just not looking at the numbers.
Aside from gambling rings almost from the beginning, in my given example I gambled 850 rings from her, and (very slightly) less than 10% were drops. That’s right in line with the standard drop rate. You’d have an argument if it was ~15%, but not what it ended at. The sample size is enough when you consider that the drop rate is supposed to be 20%. That’s one-in-five. It does not require tens of thousands of samples to come to a general conclusion.
Not only that, every other armour piece (with boots being a possible exception, and I can’t vouch for weapons or amulets) were clearly dropping at around 20%. Many times when spending 2k shards on, say, gloves, my inventory was almost full. None of that is reflected with rings which I have gambled heavily since the beginning of the season. Never has it reflected a double drop rate.
Instead of perhaps of making general, wand having arguments in the face of actual given data, you could do a test yourself. Gamble 500 rings (25k shards) and come back with your results. I would wager your results will match closely with mine and others.
I’m looking at numbers and that’s sort of my impression as well. I am just trying to ensure if that’s my superstition or not. I tried a bunch of chest armor about an hour ago, tons of it and rates looked like 10-14% to me at best per dump attempt. Spread still looks abit off for me. I’ll gamble rings and report back tomorrow or so.
edit: I dumped a few rounds of 1500 bloodshards into different items from Kadala. According to my findings, rings, chest armor and boots seems like 10% only, while helms and bracers are accurate on 20% chance. I think Blizzard needs to patch this, but knowing their hesitancy about mid-season resets on consoles it will be a hotfix or nothing.
Especially if we pool data.
Combine your 83/840 with my 52/500 for rings.
That’s 135/1340.
The probability of getting 135 or fewer legendaries, on 1340 gambled rings, if the true legendary gamble probability for rings was 20%, is 1 in 15,396,402,281,177,711,378,222.
Just curious, where do you get these astronomical calculations? Can you please show the math? Or is it explicit content because you are pulling it out of your Wizard hat?
It’s just the p-value calculation for the binomial distribution.
If you have some statistics software, you can get it easily, e.g. in R:
> pbinom(135,1340,0.2)
[1] 6.495024e-23
Otherwise, you could calculate it manually summing over the appropriate range of binomial pdf, e.g.
sum from x=0 to 135 (1340 choose x)*0.2^x*(1-0.2)^(1340-x)
Wolfram alpha can calculate this if you punch it in there
If you want a more detailed explanation, it’s actually fairly intuitive if you break it down. The binomial distribution (and related geometric and negative binomial distributions) are all fairly intuitive to calculate if you understand basic probability math.
I’ll illustrate it for simpler numbers, e.g. x=2 successes on n=20 trials with p=20%. If the chance of success is 20%, then the chance of failure on a given trial is 1-p, in this case 80%.
The probability of getting 2 successes on the first 2 trials, followed by 18 failures on the remaining trials is 0.2*0.2*0.8*0.8*0.8*0.8*0.8*0.8*0.8*0.8*0.8*0.8*0.8*0.8*0.8*0.8*0.8*0.8*0.8*0.8, or more succinctly, 0.2^2*0.8^18. In general terms, this can be expressed as p^x*(1-p)^(n-x) where n is the total number of trials, x is the number of successes, and p is the probability of success on a trial.
Now that’s specifically for the case where you get success on the first two trials, followed by 18 failures. There’s lots of other ways to get 2 successes on 20 trials. You could have successes on trials 2 and 9, with failures on the other 18 trials. That would have probability 0.8*0.2*0.8*0.8*0.8*0.8*0.8*0.8*0.2*0.8*0.8*0.8*0.8*0.8*0.8*0.8*0.8*0.8*0.8*0.8. It should be easy to see this is the same product as above, just rearranged, and simplified it’s still 0.2^2*0.8^18.
We could do it for other combinations of trials, e.g. successes on trials 3 and 12, but it should be easy to see all of the possible combinations of 2 successes on 20 trials have the same probability, 0.2^2*0.8^18.
So to get the total probability of getting 2 successes on 20 trials, we could count up all the combinations of pairs of trials with 2 successes. There’s lots of them, Successes on trials {1,2}, {1,3}, {1,4}, …, {18,19}, {18,20}, {19,20}. It’d be pretty tedious to count up all these manually, especially for larger numbers like 135 successes on 1340 trials. But it turns out there’s a nice simple way to calculate this. It’s called a combination, and it’s calculated as 20!/(2!18!) = 190, i.e. here are 190 different ways to get 2 successes on 20 trials. This is commonly called the choose function, i.e. (20 choose 2) = 20!/(2!18!). To find the number of combinations of k successes on n trials, it’s just (n choose k) = n!/(k!(n-k)!.
So then to get the total probability, we can multiply the number of combinations by the probability of an individual combination, i.e. (20 choose 2)*0.2^2*(1-0.2)^(20-2) = 0.137. There’s a 13.7% chance of getting exactly 2 successes on 20 trials.
In more general terms for arbitrary n,x,p we have (n choose x)*p^x*(1-p)^(n-x). And that’s it, that’s the pdf (probability function) of the binomial distribution. The probability of getting exactly x successes on n trials with probability p is (n choose x)*p^x*(1-p)^(n-x).
To get the probability of k or fewer successes on n trials with probability p, we sum over x from 0 to k. This is just the p-value for the binomial distribution, i.e. sum from x=0 to k (n choose x)*p^x*(1-p)^(n-x).
For our case with 2 successes on 20 trials, we’d be summing the probabilities to get 0, 1, or 2 successes.
sum from x=0 to 2 (20 choose x)*0.2^x*(1-0.2)^(20-x) = 0.206, i.e. there is a 20.6% chance to get 2 or fewer successes on 20 attempts if the true probability of success is 20%.
In this case, with n=1340, k=135, p=0.2, we have what I showed above:
sum from x=0 to 135 (1340 choose x)*0.2^x*(1-0.2)^(1340-x) = 6.495x10^-23
I have more news to report
Conclusion: The Kadala Buff has been applied to some, but not all, gear slots.
I tracked the following Kadala gambling. (In addition to my original post)
1054 “Kadala Trials”
92 Kadala drops
8.73% = Observed Success Rate
692 “Kadala Trials”
130 Kadala drops
18.79% = Observed Success Rate
515 “Kadala Trials”
50 Kadala drops
9.71% = Observed Success Rate
For each gear slot, lets estimate 95% Confidence Interval for its “Kadala Success Rate”.
8.73% = Observed Success Rate
0.87% = Standard Deviation
95% Confidence Interval for "Boot-specific Kadala Success Rate"
8.73% plus/minus 1.7%
7.03% to 10.43%
18.79% = Observed Success Rate
1.48% = Standard Deviation
95% Confidence Interval for "Glove-specific Kadala Success Rate"
18.79% plus/minus 2.91%
15.883% to 21.7%
9.71% = Observed Success Rate
1.89% = Standard Deviation
95% Confidence Interval for "Chest-specific Kadala Success Rate"
9.71% plus/minus 3.71%
6.00% to 13.42%
Kadala is bugged … but its predictable and repeatable.
Too bad nothing will be done about it. Devs do not read these forums.
Trying to get a helm and a chest. Helm is 20%, chest is 10%. Yeah, sad.
Rings must be very low. I frequently get an inventory full of yellows and blues when going for rings.
Off-hands (for necros at least) seem to give even more than would be expected with the buff. Tons. On further review, what appeared to be tons was pretty much right on mark for 20% for off-hands. It just seemed like tons compared to rings.
A sample size of 751 gambles is nothing. Try a sample size of 100k.
A sample size of 751 gambles is nothing. Try a sample size of 100k.
The (purported) drop rate is 20%, clown. It’s not .001%. You don’t need “100k” samples to come to a conclusion. The smaller the chance, the bigger the sample size needs to be. That simply isn’t the case here.
to throw some more numbers in here:
gambled boots 548 times (13700 shards) (on the character that unlocked the altar)
got 48 legendaries
that is ~8,75% leg rate
chances of getting that if it would be 20% is 4,4e-13
or
1 in 2,272,727,272,727
assuming kadala is broken and is using 10%, would put the chance of getting 48legs or less at 18%, way more plausible.
there is bad RNG and then there is almost statistical impossibilities
and looking at the numbers posted here, something is wrong with the Kadala-altar unlock
So in terms of just hunting my last primal to unlock full altar, I guess the takeaway is to gamble for gloves? Since they seem to be closer to the expected 20%?
Absolutely! The drop rate for gloves is close to 20% (I’ve gotten at least one double primal from her gambling for gloves). The drop rate from chests is definitely not 20% Tested with both wiz and dh.