PSA: Understanding map drop streaks

"

GeorgAnatoly wrote:

"

Qarl wrote:

"

Vhlad wrote:

The thing is I don't see a normal (Gaussian) distribution of maps. It's binary, feast or famine.

The streakiness is too extreme for there to be a static global % map drop success scaled by map quantity.

No, it isn't.

Streakiness in random distributions happens.

Just want to point out that the legitimacy of the mathematics doesn't change the fact that not getting maps to drop simply isn't fun.

There is legitimacy of mathematics, and there is legitimacy of code that tries to mimic a specific process. :)

"

Qarl wrote:

"

Vhlad wrote:

The thing is I don't see a normal (Gaussian) distribution of maps. It's binary, feast or famine.

The streakiness is too extreme for there to be a static global % map drop success scaled by map quantity.

No, it isn't.

Streakiness in random distributions happens.

Just come out and say the only factor to map drops is the mods on the map and that there is no other (and far more determining) factor such as a varying seed.

Or don't.

"

TheAnuhart wrote:

"

Qarl wrote:

"

Vhlad wrote:

The thing is I don't see a normal (Gaussian) distribution of maps. It's binary, feast or famine.

The streakiness is too extreme for there to be a static global % map drop success scaled by map quantity.

No, it isn't.

Streakiness in random distributions happens.

Just come out and say the only factor to map drops is the mods on the map and that there is no other (and far more determining) factor such as a varying seed.

Or don't.

So you're asking them to lie to shut the player base up concerning the thousands of map drop posts in the forums ? I don't think GGG will come out and purposefully lie to us just to shut us up. I guess it's obvious that I believe there is a "far more determining" factor in map drops such as a varying seed per login, character name, or some other wildly swinging factor.

Shame I'm late to such an interesting post.

I really only see this two ways, either GGG is appealing to a core demographic or they are trying to appeal to everyone. If choosing between the choices of a select group who have been given an enthusiasts grade game, tailored towards a hardcore market I would easily be in favor of the developers choice.

The point is deeper I realize, but I doubt GGG is intentionally out to loose player retention at high level content, the comparison is interesting though.

"

Qarl wrote:

"

Vhlad wrote:

The thing is I don't see a normal (Gaussian) distribution of maps. It's binary, feast or famine.

The streakiness is too extreme for there to be a static global % map drop success scaled by map quantity.

No, it isn't.

Streakiness in random distributions happens.

<3

And thanks again for another interesting look at the math of mapping, Scrotie. Always interesting to read.

"

Qarl wrote:

"

Vhlad wrote:

The thing is I don't see a normal (Gaussian) distribution of maps. It's binary, feast or famine.

The streakiness is too extreme for there to be a static global % map drop success scaled by map quantity.

No, it isn't.

Streakiness in random distributions happens.

True.

Do you think that's a good or desirable thing, though? Or do you think that is just what happens? And what do you think the effect on your average player is?

As you design this game, you should know that there are ways to dampen the streakiness of RNG (and I am not talking about guaranteed map drops). The question is whether you want to use them and improve your mapping system, or whether you want to keep things as they are, and the system as it is feels, I got to tell you, very unrewarding.


@Scrotie: Nice post explaining the mechanics behind things, but I am still not satisfied. The problem is not your math, though, but the way the map system works.

All you need to know is that T9=76, more than enough to get you to 100

Players are frustrated because their expectations aren't matching reality.

Good luck getting people to change their expectations.

"

Qarl wrote:

"

Vhlad wrote:

The thing is I don't see a normal (Gaussian) distribution of maps. It's binary, feast or famine.

The streakiness is too extreme for there to be a static global % map drop success scaled by map quantity.

No, it isn't.

Streakiness in random distributions happens.

There's another factor at work here as well. Let's say someone tosses a coin in the air 20 times and only gets heads once. So he gets 100 friends and they each toss the coin in the air 100 times and between all 10,000 coin tosses, they get 78 heads. It could still be random clustering right? So they get a Facebook campaign that goes viral called "RNG Broken?" and 2 million people sign up and each conducts 100 coin tosses. The net result? 1.1 million heads.

Is Randomness broken at that point? No, because what I didn't explain was that the coin has 100 different possible results, so the results are very possible.

As players, we don't know the exact drop rate chances, the total number of successful drops, the number of unsuccessful attempts, or any of a multitude of related factors. Unless a group tries to systematically account for all the variables to get a representative sample, we can only make intuitive guesses.


In the first link (Clustering Illusion on Wikipedia) I would agree with their conclusion and logic, but their "hot hands" example of basketball players is a poor choice - for the very same reasons as players evaluating PoE's randomness of drops.

A basketball player may be more rested, confident, be less covered (because of another player being more heavily covered), have an advantage over a particular defenders ... etc on any given day, which could be non-random factors affecting whether he or she has a "hot hand" on a given day. The reverse is true as well, with injuries, or a coach who scripts a good defense against an opponent, etc helping a player have a 'cold hand' at shooting. There are far too many variables, many of them unknown.

That statisticians think they can adequately apply complete randomness to sporting events is what I like to call the mathematicians fallacy. They really want to believe, but a competitive sporting event isn't random. A ball will certainly take an unpredictable bounce here and there, but if you could measure all the variables, the bounce would be highly predictable. The stray atoms that hit it here and there aren't going to affect the macroscopic behavior.

Hmm, I was trying to find some video of a basketball player that had a very "hot hand" - never missing a shot in years....

Unfortunately, it seems there is no video of the basketball shooting robot they used to have a Pacific Science Center in Seattle. It was similar to the one Carnegie Science center has, but IIRC the mechanism for holding the ball was a little bit different. The Carnegie one has degraded over time, but at one point it was a perfect example of where randomness didn't apply to something that people think has a random element.

Athletes intentionally trying to do something isn't random, so ruling randomness "In" is a very silly idea. One that I think even Monty Python would be embarrassed to entertain.

In any case - I think people look for randomness and fail to see randomness all too often.

The apparent streakiness, or clustering of drops does make for exciting game play at times.

@johnKeys

The larger the sample size, the less streaky things become.

Flipping less than 4 heads out of 10 coins is a 17.2% chance.

Less than 40 out of 100, 1.76% chance. (I use this to make money when I go out drinking. Actually had to give out a $20 once; rest of the time I got my dollar.)

Less than 400 out of 1000, 1 in about 11.1 billion. (The chance of less than 470 is 2%.)

I don't feel like making a spreadsheet on my phone with 10000 rows.

What you are talking about is so astronomically unlikely that it is downright foolish to think it has ever occurred, or ever will. Assuming PoE doesn't last for over 100 years.

Edit: to contribute to the actual topic, if you think map drops are streaky individually - and they are, individually - get running 10-20 of them, and not so much anymore. I might even run the numbers.