Crafting statistics (Gotta love the math here.....)

Posting this out of extreme annoyance at the math that I'm being screwed out of.. I crafted 22 exceptional rings (Imagine how many it took to get those at 25% exceptional rate). So I decided to upgrade these rings. My enchantment is GM'd and was at 102 (You'll see it go to 104 from the SS).

This puts me at a 95% chance to add the first + (None of the rings broke from the first enchant)
83% chance for the second enchant
46% chance for the 3rd enchant

"Alright, i like my odds. From these straight forward numbers I will lose approximately 16 rings."
Here's the straightforward math. 22 rings * 0.83 = 18.26 so 18 rings left after the first enchantings. 18 * 0.46 = 8.229 so 8 rings should be left in the end.

The reality of what i got was this:
8 rings breaking at 83% chance making my chance for successfully enchanting a second time at 73.7%.
11 of the remaining 14 rings broke at 46% making my 3rd enchant a whopping 21.5% chance of success....

Less than half of my success rate.... Seems legit...
I hope a Dev or admin reads this and can explain this math to me please because from the numbers presented, it was extremely off on a fairly sizable selection to pool stats from.

Here's the screenshots showing the breaks. Please feel free to leave comments of your own beloved % experiences. Maybe we can get changes in the game to make it more realistic to the numbers that are given to us.

1: Broke at 46%
2: Broke at 83%
3: Broke at 46%
4: Broke at 46%
5: Broke at 46%
6: Broke at 83%
7: Success at 46%
8: Broke at 46%
9: Broke at 46%
10: Broke at 46%
11: Broke at 83%
12: Broke at 83%
13: Broke at 46%
14: Broke at 83%
15: Success at 46%
16: Broke at 46%
17: Broke at 46%
18: Broke at 46%
19: Success at 46%
20: Broke at 83%
21: Broke at 83%
22: Broke at 83%

 

Your sample size is small and you're also cherry picking within the sample size. You performed above expectation on your first enchantments and so you just threw those trials out. Not that it makes a huge difference, but it's hard not to produce anomalies if you pay more attention to samples based on results.

 

I'm not cherry picking at all. You saying I performed above exception with my first enchantments means very little when it's a 95% success chance. 1 ring should have broken, all of the statistics are there right there. A sample size of 22 exceptional rings rings is a fairly decent amount, unless you want to craft 400+ rings to have a sample size of 100ish exceptionals. I honestly do not understand what point you are arguing and would like clarification on how I am cherry picking and how this listing of facts is anything but factual.

 

Actually, no. It's not a very significant sample at all. You don't have to take my word for it though. The math of variance isn't a matter of opinion.

Leaving out the first enchantments doesn't change your expectation much, but it does clearly show some bias on your part, even within this tiny sample. I really doubt that you would have reported your results if your results had been at or above expectation, not only because of that bias, but because no one ever does. There are no threads where someone comes here to report binking 12 exceptionals out of 20 attempts even though things like that are a fairly common occurrence also.

 

How does it show bias when i clearly state that no rings broke on the first enchant... it's a 5% difference thats well within reason, my goal was to focus on the 46% variances and how i got less than half. You seem really hellbent on slandering me for no reason when i am reporting my results. Please try to refrain from insulting others on their posts.

 

@Knightmare2 your math seems correct. There is a branch of mathematics called combinatorics that is often applied to dice rolls & card games.

According to a standard formula the chances of a ring making it intact through 3 enchants (based on your %) would be :

95 % x 83% x 46% = 34.2%


34.2% x 22 rings = ~ 7.5 rings that are intact & triple enchanted

The sample size doesn't really matter.

 

@Numa Exactly, I have a minors in math and it doesn't take that to know statistics.

 

It kind of does... take, for example, coinflip. I can flip a coin 22 times and even though the chance is roughly 50/50 (depending on design of the coin).... It's possible to have it come up heads everytime. The liklihood of success is useful in explaining the possibility of the outcome you receive, but unless the number of flips/enchants/etc is quite high you can get a pretty large margin of error.

For a very basic example of how sample sizes work with statistics see: http://nbviewer.jupyter.org/urls/gi...916b7dde89acd0f51d00710e12c/sample_size.ipynb

I wouldn't be concerned with this result using this sample size. 1,000 or 10,000 with this variance would be much more disturbing. (Blame RNGeesus for your bad luck).

 

Right, i'll get back to you when i've crafted 1k exceptional rings and enchanted all of them to the third tier.
It's easy to just blame sample size and I understand it's a small sample size. Numbers are numbers and the facts are merely above.
Everybody is entitled to their own opinion and beliefs whether they are right or wrong.

 

Putting the statistical math aside. We're having user experience issue here. The percentage of success chance is not reliable against the RNG, because we can't assume that the seed will eventually distribute the stream of random numbers. So while seeing 80℅ success chance, we could be failing 10 out of 10 due to bad seed.

Perhaps it may be help to add the character based successful percentage based on the individual on the acreen as well?

Or maybe chance this success chance into the number of dice rolls. If you reach certain sum, then you win. Otherwise you lose.. since we're gambling our gear away anyway, so might as well seeing some dice rolling.

 

This!

 

A great suggestion. I believe it would console some people if we would at least see the percent that was "rolled" for the gear other than, oh it broke, oops the next one broke, oops let's repeat this 8 times in a row.

 

couldnt someone go on the q and a server and craft 1000 rings then? Get a bigger sampling? Maybe hav a dev front him the mats for crafting 1000 rings? Sounds boring as hell, but it might be necessary to solve this effectively.

THE HENDOMAN

 

Edited offline save file perhaps.

And yes, the RNG seed needs to be ironed out a bit

 

Yikes.. 3/5 of my last 5 bow crafts wer3 excep... then i did armor yesterday n got 3/7 except.. did masterwork x2 on all3 bows.. did enchant x2 on all 3 bows.. and masterwork x1 on each armor. 0 breaks. and .. 2nd masterworks were about 67%.. firsts only 90%..

Sorrry.. i stole your good rolls :/

 

Afaik it's a known Unity issue that seeds don't update if you don't tell them to. That's just hearsay from someone that actually ran accross problems with randomness in Unity, I don't know the specifics. It's not that overall with a huge sample size the math doesn't add up. The issue is that the distribution in small sample sizes is not as one would expect with true randomness. E.g. you get 8 crits in a row and then no crits at all for the next 30 shots. I don't think it's a huge issue in combat atm, but it seems to make the randomness in crafting even more frustrating.

 

Using this tool because I'm lazy: http://stattrek.com/online-calculator/binomial.aspx
with a probability of success of (0.95*0.83*0.46)=0.36
in 22 attempts, the probability of having no more than 3 of your rings survive is 2% which is not unrealistic. The chance of having more than 7 get through, as suggested by @Numa, is 57% so is far from guaranteed.

For comparison: the chance of having no more than 30 survive in 220 attempts is <0.000001% so sample size absolutely does matter.

N.B. this is not to say everything is working perfectly because I can't possibly know whether that's the case, this is a purely mathematical treatment of the observed results.

 

The RNG, seem to have same problem with Skills. A lot of time, i got a Fizzle Spell, 2 or 3 time in a row, with spell i have at 92% . i dont move, i'm in Leather.. and this is happen to me, a lot of time.. i fizzle spells, 2, 3 time in a row, spell i have at 90% and more..

 

RNG complaints have been going for the last 3 releases. You have the sample-size statisticians vs the crafters who have a beef with the implementation of the instantaneous die-roll result for success vs failure of hours of grinding. No one can argue the statisticians because no one is going to go through the pain of gathering components and crafting 1000 sets of whatever to try and prove the point.

Applying pure randomness on a basis of expected statistical results over time doesn't work with the expectation I have to see a reasonable reward for my effort in some short-term window. This method of generating results isn't fun whatsoever.

IMO the determination of masterworking and enchanting success needs to be rethought using a creative approach that balances effort, crafting skill and reward.

Using pure RNG for success and failure sucks given the statistical variations over a medium sample size and the known Unity issues with RNG. The current approach is a cheap way out of building a more balanced and creative scheme to making the crafting fun.

I dropped Path of Exile like a hot rock because of their BS RNG.

 

Sample size is perfectly fine and his outcome matchs the outcome of my ~400 attempts.