Nihlathak - Keys?

Just curious - is anyone finding keys from

Nihlathak right now? (Softcore, Ladder, Americas)
I used to find a

Key every 3-5 runs. I think I've done about 40-50 now without a single

Key. I remember there being an issue with this guy in the past. Just wondering if it's bad luck lol.

For me it's the damn Horazon impersonator Summoner who won't drop shet. Welcome to RNGesus.

halflife wrote: 4 weeks ago

Welcome to RNGesus.

This

Make sure to run higher player count games to increase the propensity of key drops

MagicPants wrote: 4 weeks ago

Just curious - is anyone finding keys from

Nihlathak right now? (Softcore, Ladder, Americas)
I used to find a

Key every 3-5 runs. I think I've done about 40-50 now without a single

Key. I remember there being an issue with this guy in the past. Just wondering if it's bad luck lol.

I found 4 keys this afternoon and all runs were 20 minutes or less. Sometimes they drop faster and sometimes they drop slower and there really is no way to know because it's all a roll of the dice. Out of curiosity, how long does it take you to do a single run?

Agreed on the higher player counts. Haven't timed my runs, but prob 2-3 min each. No Enigma yet. No worries, glad keys are dropping, RNGesus tips the other way sometimes. Over the holidays I found quite a few with minimal effort, it was nice.

MagicPants wrote: 4 weeks ago

Just curious - is anyone finding keys from

Nihlathak right now? (Softcore, Ladder, Americas)
I used to find a

Key every 3-5 runs. I think I've done about 40-50 now without a single

Key. I remember there being an issue with this guy in the past. Just wondering if it's bad luck lol.

IIRC, the chance for a

Key to drop is something like 1/12...
So, 1 in 3 runs is extremely lucky, while none in 50 runs is just the same level unlucky.

I get no keys most of the time, and then double drops twice in a row.
It comes down to luck in the end.

ShadowHeart wrote: 4 weeks ago

the actual drop rates should be pretty damn close to the theoretical drop rates.

Shadow nailed it here.

Personally I have always found the averages to normalize and be fairly accurate over a curve of time. I was on my stream a month or so ago when I was doing

Key hunting for my Torch and I reacquired many of the keys in 30 minutes at P7 games. Always RNGesus of course, but drop rates are always about convergence of averages.

Th3ory wrote: 4 weeks ago

ShadowHeart wrote: 4 weeks ago

the actual drop rates should be pretty damn close to the theoretical drop rates.

Shadow nailed it here.

Personally I have always found the averages to normalize and be fairly accurate over a curve of time. I was on my stream a month or so ago when I was doing

Key hunting for my Torch and I reacquired many of the keys in 30 minutes at P7 games. Always RNGesus of course, but drop rates are always about convergence of averages.

To obtain a drop rate average wouldn't the random numbers generated have a preset range? For instance if there is a 1 in 13 chance of a drop then rng would essentially be spinning a wheel with 13 numbers on it, and the rate is essentially reset for each spin. Blizzard has set a system of of bot detection. I read where a dev stated that bots exhibit predictable behaviors and paths, and that they've placed objects along those paths to bring them to a stop. Suppose Blizzard has put in place other bot predictors to mitigate bot behavior? Couldn't someone who engages in an efficient and repetitious Pattern such as

Key farming be detected as a bot? If so couldn't the game number generator automatically increase the range from 1 in 13 to 1 in 26, or even 1 in 52? Mechanical number generators are limited and far more predictable, but software rng can be easily manipulated using basic AI player/bot behavior detection because if a player is effective at farming, then the game AI can't tell them from a bot. There, that should disrupt things a little. LOL

mhlg wrote: 4 weeks ago

Th3ory wrote: 4 weeks ago

ShadowHeart wrote: 4 weeks ago

the actual drop rates should be pretty damn close to the theoretical drop rates.

Shadow nailed it here.

Personally I have always found the averages to normalize and be fairly accurate over a curve of time. I was on my stream a month or so ago when I was doing

Key hunting for my Torch and I reacquired many of the keys in 30 minutes at P7 games. Always RNGesus of course, but drop rates are always about convergence of averages.

To obtain a drop rate average wouldn't the random numbers generated have a preset range? For instance if there is a 1 in 13 chance of a drop then rng would essentially be spinning a wheel with 13 numbers on it, and the rate is essentially reset for each spin. Blizzard has set a system of of bot detection. I read where a dev stated that bots exhibit predictable behaviors and paths, and that they've placed objects along those paths to bring them to a stop. Suppose Blizzard has put in place other bot predictors to mitigate bot behavior? Couldn't someone who engages in an efficient and repetitious Pattern such as

Key farming be detected as a bot? If so couldn't the game number generator automatically increase the range from 1 in 13 to 1 in 26, or even 1 in 52? Mechanical number generators are limited and far more predictable, but software rng can be easily manipulated using basic AI player/bot behavior detection because if a player is effective at farming, then the game AI can't tell them from a bot. There, that should disrupt things a little. LOL

The easier way to think about this is # of games to acquire said item. So by saying 1 in 13, it is really in layman forecasting that you should find at least (which is important), one copy of this item by the 13th game. But with any statistics, there is a standard deviation +/-. The only real way to know this or to validate said run is to complete thousands upon thousand of runs to really empirically validate drop rates. This goes for all items in the game not just keys.

D2DropDash has awesome visuals to help depict this - below is an example with /players7, 200% MF, and Nightmare

Andariel
vs. Cumulative view
Chance to drop 1 copy
The general idea here around the 'convergence' of averages which is very important is that you expect somewhat of a bell-curve overall based on RNG of item generation and hitting the appropriate quality checks. But over time, the averages should normalize. Some gain the item faster, some it takes significantly longer etc.

Th3ory wrote: 4 weeks ago

mhlg wrote: 4 weeks ago

Th3ory wrote: 4 weeks ago

Shadow nailed it here.

Personally I have always found the averages to normalize and be fairly accurate over a curve of time. I was on my stream a month or so ago when I was doing

Key hunting for my Torch and I reacquired many of the keys in 30 minutes at P7 games. Always RNGesus of course, but drop rates are always about convergence of averages.

To obtain a drop rate average wouldn't the random numbers generated have a preset range? For instance if there is a 1 in 13 chance of a drop then rng would essentially be spinning a wheel with 13 numbers on it, and the rate is essentially reset for each spin. Blizzard has set a system of of bot detection. I read where a dev stated that bots exhibit predictable behaviors and paths, and that they've placed objects along those paths to bring them to a stop. Suppose Blizzard has put in place other bot predictors to mitigate bot behavior? Couldn't someone who engages in an efficient and repetitious Pattern such as

Key farming be detected as a bot? If so couldn't the game number generator automatically increase the range from 1 in 13 to 1 in 26, or even 1 in 52? Mechanical number generators are limited and far more predictable, but software rng can be easily manipulated using basic AI player/bot behavior detection because if a player is effective at farming, then the game AI can't tell them from a bot. There, that should disrupt things a little. LOL

The easier way to think about this is # of games to acquire said item. So by saying 1 in 13, it is really in layman forecasting that you should find at least (which is important), one copy of this item by the 13th game. But with any statistics, there is a standard deviation +/-. The only real way to know this or to validate said run is to complete thousands upon thousand of runs to really empirically validate drop rates. This goes for all items in the game not just keys.

D2DropDash has awesome visuals to help depict this - below is an example with /players7, 200% MF, and Nightmare

Andariel

image.png

vs. Cumulative view

image.png

Chance to drop 1 copy

image.png

The general idea here around the 'convergence' of averages which is very important is that you expect somewhat of a bell-curve overall based on RNG of item generation and hitting the appropriate quality checks. But over time, the averages should normalize. Some gain the item faster, some it takes significantly longer etc.

A 1 in 13 chance only means each time you do a run you have a 1 in 13 chance of a drop, but it doesn't mean that in a second run it becomes 1 in 12, and a 3rd 1 in 11, and on until by the 13th run you get a drop. It means that for each run you have a 1 in 13 chance or about 7.69% chance on each run. I agree, the larger the sample the more accurate the statistical result and yes there will always be some deviation. What I was getting at is that software rng is easily manipulatedn because and event range can be altered on the fly, and that in turn means the probability of an event happening will also change.

mhlg wrote: 4 weeks ago

Th3ory wrote: 4 weeks ago

ShadowHeart wrote: 4 weeks ago

the actual drop rates should be pretty damn close to the theoretical drop rates.

Shadow nailed it here.

Personally I have always found the averages to normalize and be fairly accurate over a curve of time. I was on my stream a month or so ago when I was doing

Key hunting for my Torch and I reacquired many of the keys in 30 minutes at P7 games. Always RNGesus of course, but drop rates are always about convergence of averages.

To obtain a drop rate average wouldn't the random numbers generated have a preset range? For instance if there is a 1 in 13 chance of a drop then rng would essentially be spinning a wheel with 13 numbers on it, and the rate is essentially reset for each spin. Blizzard has set a system of of bot detection. I read where a dev stated that bots exhibit predictable behaviors and paths, and that they've placed objects along those paths to bring them to a stop. Suppose Blizzard has put in place other bot predictors to mitigate bot behavior? Couldn't someone who engages in an efficient and repetitious Pattern such as

Key farming be detected as a bot? If so couldn't the game number generator automatically increase the range from 1 in 13 to 1 in 26, or even 1 in 52? Mechanical number generators are limited and far more predictable, but software rng can be easily manipulated using basic AI player/bot behavior detection because if a player is effective at farming, then the game AI can't tell them from a bot. There, that should disrupt things a little. LOL

As someone who has made my own mods for original D2 by playing around with, among other things, the files with the droprates, yes, it is a simple X/Y chance for things to drop.
Or at least for the current tier. As there, hitting any other number out of Y, drops to the next treasure level where it's another simple X/Y for each thing there and so on until it hits lowest treasure level tier.

mhlg wrote: 4 weeks ago

Th3ory wrote: 4 weeks ago

mhlg wrote: 4 weeks ago

To obtain a drop rate average wouldn't the random numbers generated have a preset range? For instance if there is a 1 in 13 chance of a drop then rng would essentially be spinning a wheel with 13 numbers on it, and the rate is essentially reset for each spin. Blizzard has set a system of of bot detection. I read where a dev stated that bots exhibit predictable behaviors and paths, and that they've placed objects along those paths to bring them to a stop. Suppose Blizzard has put in place other bot predictors to mitigate bot behavior? Couldn't someone who engages in an efficient and repetitious Pattern such as

Key farming be detected as a bot? If so couldn't the game number generator automatically increase the range from 1 in 13 to 1 in 26, or even 1 in 52? Mechanical number generators are limited and far more predictable, but software rng can be easily manipulated using basic AI player/bot behavior detection because if a player is effective at farming, then the game AI can't tell them from a bot. There, that should disrupt things a little. LOL

The easier way to think about this is # of games to acquire said item. So by saying 1 in 13, it is really in layman forecasting that you should find at least (which is important), one copy of this item by the 13th game. But with any statistics, there is a standard deviation +/-. The only real way to know this or to validate said run is to complete thousands upon thousand of runs to really empirically validate drop rates. This goes for all items in the game not just keys.

D2DropDash has awesome visuals to help depict this - below is an example with /players7, 200% MF, and Nightmare

Andariel

image.png

vs. Cumulative view

image.png

Chance to drop 1 copy

image.png

The general idea here around the 'convergence' of averages which is very important is that you expect somewhat of a bell-curve overall based on RNG of item generation and hitting the appropriate quality checks. But over time, the averages should normalize. Some gain the item faster, some it takes significantly longer etc.

A 1 in 13 chance only means each time you do a run you have a 1 in 13 chance of a drop, but it doesn't mean that in a second run it becomes 1 in 12, and a 3rd 1 in 11, and on until by the 13th run you get a drop. It means that for each run you have a 1 in 13 chance or about 7.69% chance on each run. I agree, the larger the sample the more accurate the statistical result and yes there will always be some deviation. What I was getting at is that software rng is easily manipulatedn because and event range can be altered on the fly, and that in turn means the probability of an event happening will also change.

@mhlh and @fulcrum75 My comment is about convergence of averages and the fundamentals of statistics. More specifically, probability distribution.

1:13 or ~7.65% (chance per run - single game) (x) 13 Games or clears = 100%. Ergo, by the 13th run, you 'should' expect on average to find at least a single copy of a Destruction

Key. But as we all know, this is not always the case given the above comment around std. deviations against the Mean. This is why drop rates and calcs are directional and computational proxies as RnG will be RnG.

Take Nishi's calculator below which perfectly depicts convergence:
100% agree, that averages are what they are - a simulation of chance vs. compounding; I didn't say this above. 1:13 is very much 1:13 individually per game, not 14% in run 2, or 21% in run 3. However, against the bell-curve it becomes important as there is a likelihood of acquiring multiple copies better than the simulation would suggest.

In other words, by X number of total area clears & runs, based on a +/- std. deviation of average, you expect to find it in 1:X.

D2DropDash who by the way is an engineer at Google and also an economist depicts the point I am making as it relates to the # of copies of said item (by # of runs - the earlier screenshots I shared) across said convergence/bell curve. You may have a 1:13 chance to get AT LEAST one

Key, but you very much can get multiple in 13 runs.

In another example, you once again see how # of runs (Area Clears - the only Drop Calc to demonstrate the data this way) in addition to D2DropDash relates to ratio of chance:
You cannot overlook this as it relates to convergence. While they are not one and the same, they are correlated and intrinsically connected.

And TCEX mechanics and item generation is an entire other conversation. TCEX operates as a galton-board mechanic in D2 RNG with a variety of things that go into treasure class selection, quality checks, base checks, # of picks, NoDrop, modifiers, and then of course the 6 item

Cap limit. All of these inputs of course influence the chance to drop and your probability of doing so.

We are all in agreement around overall what X:Y mean as a straight ratio and you can indeed equate this on 'averages' to # of full clears and the chance (not guarantee) of finding AT LEAST a single copy of said item.

You may have a 1:13 chance to get AT LEAST one

Key, but you very much can get multiple in 13 runs.

Any statistic is possible, I've had as many as 3 keys drop at once.

My point is that software rng isn't always that random because its rng numerical range can be increased and decreased on the fly. For instance magic find wouldn't work if D2r software was static. All software incorporates internal logic mechanisms or switches which control different functionalities based on a specific set of conditions that automatically trigger when those conditions are met.

mhlg wrote: 3 weeks ago

conditions that automatically trigger when those conditions are met.

The Slot machine that is D2 that we have loved for 20 years . Good discussion my friend. We are all saying the game thing, I just wanted to provide some clarity around my earlier commentary.

In the end, the data is helpful, but it is about the finite chance of all intersections being as you astutely nailed, met.

If I may follow up on this. I did finally get one

Key drop today. Haha. One more for my set. I've never done Ubers, but I'm gonna give it a try soon.