[outdated] Caustic Arrow Solo Map MFer (20/300+)

"

hankinsohl wrote:

"

cloudscraper wrote:

is there any spreadsheet or probabilities to see whether buying perfect or near perfect andvarii is better or worse than Divining them a bunch of times?

Well, all outcomes on a divine are equally likely. So it's not hard to work this out on your own :-)

If you're in pursuit of a near-perfect Andy, it's almost certainly not worth divining to get it.

It would make sense to divine an Andy if you have a very bad one and just want some soft of improvement - assuming of course that better Andy's are pretty expensive.

If you have enough exalts, another thing to look into is well-rolled Ventor's Gambles. A well-rolled VG will give better magic find than an Andy and has IIQ to boot - which is a bit better than the product of IIQ and IIR suggests.

my Andy are both 55+15%, so pretty low.

Every 80+% Andvarius in Prophecy SC costs around 70 chaoses, which is around 6 divine orbs.


EDIT:
Hypergeometric calc.
21 population size (50<->70)
6 successes in population (65+)
6 = sample size (6 divine orbs at 13c each are 1ex, the price of a good andvarius)
1 = successes in sample (when you get the first good andvarius, you actually stop)

Cumulative Probability: P(X >= 1) 0,91 = 91%

"

cloudscraper wrote:

"

hankinsohl wrote:

"

cloudscraper wrote:

is there any spreadsheet or probabilities to see whether buying perfect or near perfect andvarii is better or worse than Divining them a bunch of times?

Well, all outcomes on a divine are equally likely. So it's not hard to work this out on your own :-)

If you're in pursuit of a near-perfect Andy, it's almost certainly not worth divining to get it.

It would make sense to divine an Andy if you have a very bad one and just want some soft of improvement - assuming of course that better Andy's are pretty expensive.

If you have enough exalts, another thing to look into is well-rolled Ventor's Gambles. A well-rolled VG will give better magic find than an Andy and has IIQ to boot - which is a bit better than the product of IIQ and IIR suggests.

my Andy are both 55+15%, so pretty low.

Every 80+% Andvarius in Prophecy SC costs around 70 chaoses, which is around 6 divine orbs.


EDIT:
Hypergeometric calc.
21 population size (50<->70)
6 successes in population (65+)
6 = sample size (6 divine orbs at 13c each are 1ex, the price of a good andvarius)
1 = successes in sample (when you get the first good andvarius, you actually stop)

Cumulative Probability: P(X >= 1) 0,91 = 91%

Fire_Marshall_Bill has a 66/15 Andy for 60c. So if you simply want an Andy that's 65+ they're cheaper than 1ex. Watch POE for bargains.

Divines seem to be 13-14c as you note above.

But, you can sell a bad Andy for 25c at least.

Cost of a 66/15 Andy is therefore 60c - 25c = 35c.

Chance of rolling 67, 68, 69 or 70 is 4 in 21.

50% chance of rolling 67+ occurs at:

n = log(.5)/log(17/21)

n = 3.3
------
You'd expect to spend roughly 3 * 13 = 39c in divines to roll a 67+ Andy.
-------

So it's close. If you feel lucky, try divining. If not, watch POE.trade for bargains. You might want to broaden your search to include rings with imperfect implicits which you could then fix using Blessings - use similar math to figure out the expected cost to fix up the implicits.
========
Math stuffs...

A hypergeometic distribution describes a situation where you select from a fixed population without replacement. This distribution doesn't apply to the current situation.

For our case you have a 4/21 chance of success each time you use a divine.

In a hypergeometric situation, you have 4/21 the first time and then if you fail, the population is now 20. So the second time you try you have 4/20 chance of success (this is what is meant by "without replacement"). In other words, each time you select from the population in a hypergeometric distribution the resultant population decreases.
-----

To achieve a 91% chance of success when trying to divine 67+, you have

(17/21)^n = (1 - .91)
n = log(.09)/log(17/21)
n = 11.4

So to have a 91% chance of success to roll 67+ you'd need to use a bit more than 11 divines.
------------
Edit:
Actually, I think the way that most orbs work (divines included) is that some sort of change is guaranteed to occur (GGG implemented this so that players would stop complaining when their orb apparently didn't do anything).

So, in the above, the number of actual outcomes on a divine is 20, not 21, because the current value is excluded.

But... I might be wrong here wrt divines/blessings. I'm pretty sure I used some sort of orb in the recent past and once used found that the target item didn't change. At any rate... whether the number of possibilities is 20 or 21, the above is going to be pretty close and if it's 20 and if you care you can adjust the math accordingly.

Fence post anyone?

Thanks a lot for serleth and han for the answers.

So far the statistik is o.k.

And your right. i stoppt every time if i had a 5l.

But if i look at statistik its totaly equal if i fuse a new item
or use the fuses on a item i used 1000 fuse bevor.

so i used till now not over 4000 that s was wrong but 2867 fuse.

and 1978 chance orbs. so i thougt it was ovre 4000 fuse but it was 4over 4000

trys to get a item. and if the average is 1000 fuse, then it says nothing about

the single try. Thats correct. BUT if some account have in 4000 trys 5 or 6 successfull hits and many many players have nothing AND its a computergenerated rng than it can have some reasons
why it works so.
My friend has 6 l 3 items with 3200 fuses and chanced over 10 items with less then 1500 chance orbs.
If i have 1200 fuses i give him my maraketh bow and the fuses and we will see if he can 6l it.

If this will happen i am really shure the computerbased rng from ggg is something what has to do with server load, people online, gateway times and so on.
If i look the drops (they are also rng) then i can prophecy :-) that if one unique drop in a map there will be more. And if one map drop then there will be more map drops.
Thats not only thinking i wrote it for 150 maps on paper and there was a significant conformity with my presuposition.

But all in all your right. More trys will get more chances and at any moment i will get my 6 l. so i try now all 6 socked items one time to 6l and then i vendor it.
so i have the chance and a little profit :-)

@ichbinwerichbi
Keep this in mind. Although the average cost of 6-linking may be 1200 fuse (presumably... I'm not sure if GGG has actually confirmed this), sometimes it will take less and sometimes more. And given the large number of players in POE, it's not surprising that you'll find some "very lucky" players who seem to 6-link with few fuses quite often and others who spend lots of fuses to no avail.

But this is just the way randomness works.

In your case, your lucky friend has no better chance to 6-link than you do.
------

With regard to unique drops, the chance is determined by the product of your magic find and the map's. If you're in a map with high IIQ/IIR then that map has a greater chance for uniques to drop than a map with low IIQ/IIR - and yes, it's more likely that several uniques will drop in that first map than in the second one.

Another factor is the number of mobs in the map. The more mobs, the more chance that a unique will drop.

But it's all randomly generated according to map packsize, map IIQ/IIR and your IIQ/IIR.

@Those who completed the 300 maps from map bosses challenge

How did you complete this challenge? Just run a bunch of maps or farm Merciless Piety?

At a addition.

I am verry intersted in the algorithm and kind of implentation from ggg in case of the rng generator. If there is some information (i dont found a explanation from cris or someone else how they implented this) i am interestet in a link or something like this.

"

Another factor is the number of mobs in the map. The more mobs, the more chance that a unique will drop.

But it's all randomly generated according to map packsize, map IIQ/IIR and your IIQ/IIR.

Ok thats seems to be a factor. I wrote only white maps (i run for the uber lab challenge).
My IIq/iir factor is in all this maps the same. And there a really extremly different drops (what i cant explain with rnd) Id would be interesting to see if the statistiks from ggg show this. They have the data and can answer this question. (but i can understand that they dont answer (-: )

"

But this is just the way randomness works.

Thats correct. But only over all players and time. That says nothing about the randomness per server load or per player online or something like this.

AND there is no really rnd generator in computercode. All are in a way not random. because they are programmed, and the processor load or ram load can interact with that.

not to say that over millions from occurences the average is rnd.

thats the reason i am interested in knowing how ggg this implemented

(i like such discussions, also if there is no chance of a result, because ggg can not exhibit there algorithm)

"

hankinsohl wrote:

@Those who completed the 300 maps from map bosses challenge

How did you complete this challenge? Just run a bunch of maps or farm Merciless Piety?

oh my, don't remind me please. It was my last challenge to be completed and not as easy as I thought it to be when I read the announcement of the challenges.

Partly, because I spent an absurd amount of time in the lab instead of maps and partly because I didn't track it until the last week. The drop rng of bosses didn't help either.



So when I started out at like 57/300 I didn't quite know this ofcourse. I would chain easy layout maps and rush to the boss. Later I joined some piety rotations. They were very tedious because most of the time, the guys wouldn't know how rush rotas worked, leave mid-run or destroy their instance by using a portal in my instance f.e. --> back to map rushing.
Then I saw something on reddit 'bout atziri runs netting more maps, went in there and ran a dozen or two dozen sets. Medium map returns, but still.
After I started to alch & corrupt arenas/courtyards/museums/etc. (nice side effect: more sac frags for even more atziri runs).

personal verdict (TLDR kinda):
- piety not worth at all without a good group, too much waiting
- caer rotas not worth the hassle of buying / assembling
- atziri is medium ground, entry gated with the sacs tho.
- easy layouted maps with multiple bosses, preferably twinned and corrupted net you most return on maps and feed you with more atziri frags. this should be very easy for MFers, we normally swim in Alts ^^

does this help you, hank?

"

Chronodroid wrote:

does this help you, hank?

Yup. Thanks Chrono :-)

I need to get some more Sacrifice to the Goddess fragments... so I'll be plowing through linear maps on my MF toon to get those - I'll probably achieve the 300 maps challenge just as a side-effect.

I guess my question was a bit more "Informational" than "practiacal."

BTW...
I'm still amazed at how fast both you ans Serleth got to 40/40.

/bow :-)

"

ichbinwerichbi wrote:

"

Another factor is the number of mobs in the map. The more mobs, the more chance that a unique will drop.

But it's all randomly generated according to map packsize, map IIQ/IIR and your IIQ/IIR.

Ok thats seems to be a factor. I wrote only white maps (i run for the uber lab challenge).
My IIq/iir factor is in all this maps the same. And there a really extremly different drops (what i cant explain with rnd) Id would be interesting to see if the statistiks from ggg show this. They have the data and can answer this question. (but i can understand that they dont answer (-: )

"

But this is just the way randomness works.

Thats correct. But only over all players and time. That says nothing about the randomness per server load or per player online or something like this.

AND there is no really rnd generator in computercode. All are in a way not random. because they are programmed, and the processor load or ram load can interact with that.

not to say that over millions from occurences the average is rnd.

thats the reason i am interested in knowing how ggg this implemented

(i like such discussions, also if there is no chance of a result, because ggg can not exhibit there algorithm)

Your questions are interesting Ichbin. Thanks for helping to keep the game interesting for me :-)

It is possible (though I think very unlikely) that there's some sort of programming error or design intent such that if a map drops a unique, it's more likely to drop more uniques, after factoring out packsize and map IIQ/IIR.

The correct way to assess such a hypothesis exists in the mathematical field of statistics.

In this field, you form a hypothesis - your's is that it's more likely that uniques drop in a map (after factoring out packsize, map IIQ and IIR) once at least one unique drops.

Next you pick a confidence level. 99% confidence is common.

Finally, you use the maths to figure out how many maps you'll need to run to build confidence in your hypothesis.

I'll likely look into this some more to fill in the details... I'm not a mathematician by profession - but I love math and am pretty good at it.

My sense is, though, that you'd need to run quite a few maps to achieve 99% confidence that your hypothesis is correct. Likely many more maps than you have run so far.

Complicating the analysis is that you cannot hold variables constant. In other words, the number of mobs in one white map will almost certainly be different than the number of mobs in a second white map, even if they're the same base map (e.g., two white Tropical Island maps). Further muddying the waters is that white mobs drop less stuff than blues which drop less than yellows which drop less than exiles which drop less than bosses. For example, let's say that a map contains an exile. You kill it and a unique drops. Now you're ahead of the game. This map will probably drop more uniques than some other map which doesn't contain exiles. Then there are drops from chests... and prophecies... etc. to factor out. In other words, ensuring that you're reaching 99% confidence in the hypothesis is complicated by a bunch of external factors which are impossible to control and which are going to be hard to adjust for.

This isn't to say that you're not correct. Instead it's saying that in spite of the results you've seen so far, it's more than likely down to just randomness.