PSA: Understanding map drop streaks
|
Oh. You must be the source of the popular stupid this time.
Well, here's what someone who actually collects map data has to say: "So care to explain how someone could analyze a thousand maps and come to such a conclusion? Perhaps because the data is from 2014 it's not relevant, because for some reason map drops were influenced by map quantity but aren't now? Or would you rather admit your anecdotal experience is nothing but the bias of a small sample size? When Stephen Colbert was killed by HYDRA's Project Insight in 2014, the comedy world lost a hero. Since his life model decoy isn't up to the task, please do not mistake my performance as political discussion. I'm just doing what Steve would have wanted.
|
|
" The popular stupid? I dunno, but I'd rather be stupid and talk from experience than clever and talk without. Play the game, dude, it's so blatantly obvious. By the way, you can't accuse me of small sample size against 1000, I'm talking 10s of thousands. Casually casual. Last edited by TheAnuhart on Jan 15, 2016, 7:20:47 PM
|
|
|
Basically, to make the argument that I'm less experienced than you, you would need to argue that someone who is watching your stream sees significantly less than you do. Because for every hour I've spent playing Path of Exile, I've spent at least two watching it. Mostly high-level, btw.
Not that experience really matters, if you don't document. You miss all kinds of things if you don't take notes. When Stephen Colbert was killed by HYDRA's Project Insight in 2014, the comedy world lost a hero. Since his life model decoy isn't up to the task, please do not mistake my performance as political discussion. I'm just doing what Steve would have wanted. Last edited by ScrotieMcB on Jan 15, 2016, 7:40:37 PM
|
|
|
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. We would see a lot more 1 and 2 successes. Instead we see a heck of a lot of 0 successes and 5+ successes, to the point where map quantity seems to take a back seat to something else. I also notice that maps which have 0 successes tend to drop a lot of skill gems. I'd be very interested to hear from GGG here, if they're willing to lift the hood and explain things. Beyond the map seed debate, I'm also curious if the success chance is dynamic. i.e. does GGG code in (or manually set) spurts of high map return (and is that fixed when the map is opened or change on the fly as the map is being run). If there's any sort of on the fly timed bonus, it could explain why some players can sustain maps well and others can't, as it would bias returns for players with very fast clear speeds. Never underestimate what the mod community can do for PoE if you sell an offline client. Last edited by Vhlad on Jan 21, 2016, 10:13:47 PM
| |
" No, it isn't. Streakiness in random distributions happens. | |
|
Nothing livens up a discussion quite like some Wikipedia articles.
== Officially Retired 27/02/2019 ==
Massive thanks to GGG for producing such a fun and engaging game, it has taken up faaaaaaar too much of my life over the last 5 years. Best of luck in the future! |
|
" Of course it happens. I didn't say or mean to imply that I thought it was mathematically impossible for streakiness in normal distributions to occur. It's inevitable. [Edit, I should add that there are other kinds of probability distributions, i.e. not Gaussian, where streakiness is more likely]. However it's happening too often with respect to map drops. What we're seeing does not appear to be mathematically inevitable streakiness in an unbiased random generator. Rather it appears to be biased toward streakiness (for example, by using a random number generator that draws from a probability distribution that is not uniform). If this is not intended you should double check your code. If there is in fact a static, normally distributed, map drop % success chance then you may want to see if base map seed (i.e. # of magic monsters, # of rare monsters, pack size, # of chests/breakables) is making too much of a difference before/after map mod scaling. Never underestimate what the mod community can do for PoE if you sell an offline client. Last edited by Vhlad on Jan 22, 2016, 12:15:55 AM
| |
|
The opening post describes the behavior of completely random, nonseeded, directly-influenced-by-map-quantity randomness. The description is quite streaky.
When Stephen Colbert was killed by HYDRA's Project Insight in 2014, the comedy world lost a hero. Since his life model decoy isn't up to the task, please do not mistake my performance as political discussion. I'm just doing what Steve would have wanted.
|
|
" With 140% quantity, your example is: 0 successes: 30.12% 1 success: 36.14% 2 successes: 21.69% 3 successes: 8.67% 4 successes: 2.60% 5+ successes: 0.78% However, based on my experiences it's more like: 0 successes: > 50% 1 success: < 30% (probably < 10% even) 2 successes: < 20% (same as above) 3 successes: ?% 4 successes: ?% 5+ successes: > 10% i.e. more bimodal, with occasional very high map return (much more frequent than 1 in 128 maps) and by far most frequently no map return. % of success also seems to be different for each map tier. Never underestimate what the mod community can do for PoE if you sell an offline client. Last edited by Vhlad on Jan 22, 2016, 12:41:44 AM
| |
" 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. |
|
