The RNG deception.
" I think the key is learning how to think in terms of inverse probability - i.e. figuring out the odds of losing, rather than focusing on your chances of winning. For example, if each roll has a 1 in 1000 chance of winning, what are the odds of getting at least one win in 1000 tries? It's hard to make a ballpark estimate and most people's intuition would be overly optimistic. If you invert the odds, each attempt has a 999 in 1000 chance of failing. The probability of getting zero wins in 1000 tries is simply (999/1000)^1000. The scientific calculator on your smartphone can tell you the result: about 37%. The odds of at least one win in 1000 are thus 63%. This means that less than 2 out of 3 people are likely to succeed after 1000 tries, leaving a sizable number of unlucky players feeling like they were robbed. Last edited by RogueMage#7621 on Jul 10, 2013, 2:54:39 PM
|
|
|
This is one of those situations when something is labelled as deception when perception does not match reality.
But is the perception valid to begin with? Intuitively, if this were in fact a FAIR random system, each possible link combination will have equal probability. Each of the 5 possible segment connections (6 nodes, 5 segments) is binary (link/no link) so that each permutation has 1/31 probability. If this is a FAIR random system, then the results should follow a binomial distribution. There are a total of 32 combinations but the transition from one combination (1) to others (31) is being counted. So the probability of getting "6-links" or 5-segment connetions should therefore just be p = 1 - (1 - 1/31)^R where R is the number of fusing rolls. 1 roll, 3% chance for "6-links" 10 rolls, 27% chance 50 rolls, 80% chance 100 rolls, 96% chance But this does not match reality at all. It doesn't follow the binomial distribution stemming from the perceived FAIR random system. The reality is that GGG (and they are not alone on this) knowingly skew the distribution of the results. It is still random...but with a skewed and biased distribution designed to be aligned with a desired outcome when applied over a large population. |
|
|
Again... it falls on GGG to make these types of numbers available. We can make up stuff all day, and try to explain our reasoning to people and what assumptions we make, but in the end they don't understand what we've said.
GGG should come out with a complete disclosure of the process and what the probability of different outcomes are. Leaving up to "random" without defining the terms is actually illegal in many places. You'll see lotteries in the US and scratcher tickets all post their odds. They have to. GGG doesn't have to, but they should. It has nothing to do with combinations of possible links by the way... it's an arbitrary number that determines the number of links. It shows up as a 5L or a 6L, but it doesn't determine each link independently. That's been shown. Last edited by Shagsbeard#3964 on Jul 10, 2013, 3:15:56 PM
|
|
" The position of each segment connections vary as well. So it's not just an arbitrary number as you put it. How the connections for example for 3L or 4L playout is important as well so each link has to be determined (rolled) independently. This is how one controlls (or skews) the distribution, via stacked RNG. Stacking RNG 5-times for "6-links" will make it more improbable than stacking RNG twice for "3-links". But what is the base? 2, 6, 12? |
|
|
Each link is not determined independently... so you're wrong. I'm not going to try to explain why.
|
|
|
Spoiler
http://www.pathofexile.com/forum/view-thread/73004
Just FYI for the sake of discussion on links and RNG. Mark replied on how links are rolled. I need more purple titles Last edited by Ocylix#1741 on Jul 10, 2013, 4:30:45 PM
|
|
|
A less frustrating system would be one where there's a chance to add or remove a link. Not go from 4L to 0 in one fuse. There would be a sense of progression/regression when using fusing orbs.
Say you have an item with 6 sockets. The chances to add/remove a link (between any sockets, not necessarily consecutive ones), could look like this, depending on the links that are already present on the item: 0 link: 100% chance to add 1 link: 60% chance to add; 40% chance to remove 2 links: 30% chance to add; 70% chance to remove 3 links: 5% chance to add; 95% chance to remove 4 links: 1% chance to add; 99% chance to remove <-- use Eternal before attempting this 5 links: 0 chance to add; 100% chance to remove (that's your 6L, obviously) With a system like this, there would be more meaning to an item that drops with links already on it. Say a 4 socket item drops like this: 0-0-0 0 or like this: 0-0 0-0 You would know you are 30% chance away from getting your 4L with 1 fusing. It's more fun than total pure RNG. Last edited by Thalandor#0885 on Jul 10, 2013, 4:58:56 PM
|
|
|
They're unlikely to change a system that works fine.
|
|
"you're assuming a binomial distribution when in fact you can make an argument for a hypergeometric distribution instead (sampling without replacement vs sampling with replacement). you're assuming the link roll is being made once for all the links in the item it seems. each link is rolled for separately. either way, it does take quite a bit of data to fit a distribution and its parameters. Last edited by floredon#6631 on Jul 10, 2013, 5:40:14 PM
|
|
"You're both wrong; it's a geometric distribution. Period. 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#2697 on Jul 10, 2013, 7:01:28 PM
|
|
