PSA: Understanding map drop streaks

Binomial Distributions. Learn to love them!

"

The problem is that frustrated players don't provide large sample sizes...they just leave the game lol. I think the game SHOULD be balanced to medium to small-medium sample sizes...because human emotions > statistics when it comes to player retention. i.e. when it feels bad or is not fun...people leave and that's bad for everyone :[

This isn't feasible. If you reference the Bonus Section I added to the OP, even if the expected return per map averages to 1.2 -- so 20% more than needed for long-term sustain -- there's a 30+% chance of getting no map. There is no way to balance random drops for extremely small sample sizes in a game which relies up thousands of random loot drops per hour, so whether GGG should or not is a moot point.

"

ScrotieMcB wrote:

"

The problem is that frustrated players don't provide large sample sizes...they just leave the game lol. I think the game SHOULD be balanced to medium to small-medium sample sizes...because human emotions > statistics when it comes to player retention. i.e. when it feels bad or is not fun...people leave and that's bad for everyone :[

This isn't feasible. If you reference the Bonus Section I added to the OP, even if the expected return per map averages to 1.2 -- so 20% more than needed for long-term sustain -- there's a 30+% chance of getting no map. There is no way to balance random drops for extremely small sample sizes in a game which relies up thousands of random loot drops per hour, so whether GGG should or not is a moot point.

There is a way. Other companies do it.

You can cut off the low end of the tail of the probability distribution curve. For example, in Hearthstone every time you open a pack and reveal a card you have a % chance to get a legendary. Rather than leaving it entirely up to RNG, Blizzard tracks your legendary frequency and if you're really unlucky they force 1 legendary every 38 packs (and the counter resets).

Systems for RNG safety nets are present in many games. Without them, RNG driven events can be absolutely devastating for a significant number of players. i.e. 2.5% of players are 2+ standard deviations from the mean in the low end of the distribution curve. If the success chance of some event is already super low, the investment/time difference for that 2.5% (1/40) can absolutely turn them off your game, permanently.

I used my words carefully. I said there was no way to balance random drops for small sample sizes, without ruining large sample sizes. You are talking about layering on deterministic - that is, nonrandom - systems for drops.

97.5% random is random enough to still be random.

Edit: I'm saying they're still random drops. Adding a safety net to balance/limit random drops doesn't stop you from calling it a random drop system. Up until the 2nd last card (or pack) it's still random, with the same probability for success as the first attempt. If one deterministic event kicks in just before your investment doubles the expected investment (which is what Blizzard does with Hearthstone, i.e. 5.37% chance of legendary per pack (~18.62 expected packs per legendary) with a forced legendary if you haven't found one by the 38th pack), I'd call that a way to balance random drops for small sample sizes without ruining large sample sizes. Blizzard does the same thing in D3 with legendary drops (tracking non-idle time or monsters killed or vessels opened instead of card packs).

GGG should be doing similar things in many cases. For overall customer satisfaction and retention, it's smart business. Pure RNG is just too abusive.

Blizzard wouldn't be such a profitable behemoth if they didn't do so many things right.

I don't think you're understanding what I said.

But in any case, I'm not sure how to feel about deterministic safety nets. Including the one in HS (my primary game, PoE is #2 for me).

"

ScrotieMcB wrote:

I used my words carefully. I said there was no way to balance random drops for small sample sizes, without ruining large sample sizes. You are talking about layering on deterministic - that is, nonrandom - systems for drops.

They could scale down the draw pool to allow for something that feels better to a smaller sample. Say for example rather than making maps able to drop from any monster, chests etc make them only able to drop from rare or unique mobs. This would allow for higher overall drop rates in a smaller number of samples at the expense of the possibility of a larger quantity of maps dropping. But this itself could be a good thing as then you can more realistically introduce a drop system that allows smaller variance in map level dropped relative to the one ran.

But in reality concerns over the economy would probably stop something like this from being feasible.

Why do people always add maths to these things... you know, these RNG based things... you're just wasting time.

"

Why do people always add maths to these things... you know, these RNG based things... you're just wasting time.

Yeah, that's like using facts and logic to win an argument. Who does that?

"

Why do people always add maths to these things... you know, these RNG based things... you're just wasting time.

That was the dumbest thing I've read this month