Please upgrade your browser for the best possible experience.

Chrome Firefox Internet Explorer
×

is RNG really RNG

Carba's Avatar


Carba
08.22.2019 , 08:13 AM | #1
it does not feel to me that the RNG function follows the binomial distribution, particularly in the tails. For example, take the Ossus daily to Reprogram Scrambled Excavator Droids. I find over the past six months that I have gotten four successes in a row four times and ten failures in a row four times. That is across 8 characters over six months, or roughly 200 missions. If p=50%, then I would expect four successes in a row to happen about 12 times and ten failures in a row once. There is no value of p that fits my experience very well unless the RNG has a very fat right tail rather than following the binomial cumulative distribution.

TonyTricicolo's Avatar


TonyTricicolo
08.22.2019 , 08:23 AM | #2
The system is rigged. The house always wins.
TONY NOVA SWTOR Shadowlands/Star Forge
-STRYDER- SWG Sunrunner/SWGLegends
Referral Link: http://www.swtor.com/r/vNnj3G

olagatonjedi's Avatar


olagatonjedi
08.22.2019 , 12:03 PM | #3
Quote: Originally Posted by Carba View Post
it does not feel to me that the RNG function follows the binomial distribution, particularly in the tails. For example, take the Ossus daily to Reprogram Scrambled Excavator Droids. I find over the past six months that I have gotten four successes in a row four times and ten failures in a row four times. That is across 8 characters over six months, or roughly 200 missions. If p=50%, then I would expect four successes in a row to happen about 12 times and ten failures in a row once. There is no value of p that fits my experience very well unless the RNG has a very fat right tail rather than following the binomial cumulative distribution.
It is my belief that the RNG isnt per character, but is server-wide for a specific occurence, which cannot be individually monitored. It also lends credence to how randomness has been experienced by many people with GC crates as well. The more people that are doing that specific randomized task, the more streaky results you will notice, and likely someone else in the galaxy far far away is experiencing the opposite streak. Examples of this streakiness are also seen in quests like the Seeker Droid line - very few people "using up fails" causes those quests to often take forever to get a positive hit.

At least that is the relative conclusion i have come to.

AwesomeTacoCat's Avatar


AwesomeTacoCat
08.22.2019 , 03:09 PM | #4
What are you implying, that each attempt at a task, such as repairing the scrambled droid, is NOT independent? That seems extraordinarily unlikely, that the game records previous attempts and bases the probability of success on previous outcomes.

So for a binary event itís almost certainly the case that the individual outcome is governed by a Bernoulli distribution. Hence, a sequence of events must be governed by a Binomial distribution.

There are at least three reasons why you might have seen some extreme sequences of successes and failures. First, a different value of p (other than 0.5) may lead to the observed outcomes being more likely. Second, 200 is a relatively small sample size - itís not that unlikely to see a bunch of outliers. Third, the probability of a success may depend on other factors. For example, your gear may matter (so that like better accuracy or better gear in general changed he probability of success). So perhaps your 4s in a row were on geared dps toons with high accuracy and the 10 failures in a row were on poorly geared healers with no accuracy.