Gear drop rates are changed from last update.

2Next

Replies

  • Kyno
    32087 posts Moderator
    Gifafi wrote: »
    YaeVizsla wrote: »
    The last update was three days ago. There is no way you have a remotely decent dataset in that time.

    Time doesnt matter. Attempts do.

    If he did 1000 sims before and drop rate was 33% and in last 3 days he did 50 sims and his drop rate was 6% he has 3nough data to make a claim

    50 sims is enough? lol?

    What do you consider is enough?

    it depends on some variables of the curve. if its a very standard curve 30, but the next rule of thumb when its less normal, 100. but those are just rules of thumb, not 100% standards.
  • YaeVizsla
    3448 posts Member
    edited May 2020
    What do you consider is enough?
    That is a more complicated question.

    However, your example- a 6% drop rate in 50 attempts- is not even 99% confident that the drop rate is a positive number, and is low enough that even 1 hit is a large swing in the rate. That dataset is only three hits.

    It's also important to keep in mind there is an enormous playerbase. Relying on people who had bad runs to self-report is not useful because somebody is expected to get bad runs. Your example of 50 rolls is a couple of days. Thousands of players are expected to get a bottom 1% bad couple of days this week. Several players will get one in a million bad couple of days this quarter. Somebody is going to get a one in a million bad couple of days this year.

    It would have to be a substantial dataset properly compiled to be at all convincing. And like all things in statistics, there is not a magic number that's "good enough." The larger the number, the more persuasive.

    Further, the statistical standard of 30 before you can draw anything generally assumes trials are difficult to get. That is not the case. There are hundreds of die rolls per day. Larger sample sizes are very easy to compile. When I tracked my Geonosian Brood Alpha drops, I got 145 trials in a week on my own, and that's off an energy we get half as much of and which costs twice as much, on a node that costs 60% more energy than the fleet nodes we generally grind gear from. 50 is not terribly persuasive, but get me an actual, properly compiled 50 trial dataset and we can start talking.

    But that's academic, as the number of properly compiled trials supporting the OP in this thread is zero. Just one screenshot a fifth the size selected in a biased manner and an informal sentiment poll.
    Still not a he.
  • YaeVizsla wrote: »
    What do you consider is enough?
    That is a more complicated question.

    However, your example- a 6% drop rate in 50 attempts- is not even 99% confident that the drop rate is a positive number.

    I understand you are a die hard supporter of CG for all your previous posts but really?
    How could the drop rate be a negative number?

    I studied, tutored, and even was a TA for statistics at a cal state... but I'm learning a ton of new info here on forums it's very interesting.

    Can anyone tell me more about these negative drop rates? Honestly I dont know how that's even possible...
  • YaeVizsla wrote: »
    What do you consider is enough?
    That is a more complicated question.

    However, your example- a 6% drop rate in 50 attempts- is not even 99% confident that the drop rate is a positive number.

    I understand you are a die hard supporter of CG for all your previous posts but really?
    How could the drop rate be a negative number?

    I studied, tutored, and even was a TA for statistics at a cal state... but I'm learning a ton of new info here on forums it's very interesting.

    Can anyone tell me more about these negative drop rates? Honestly I dont know how that's even possible...

    Its called a confidence interval, would have expected someone who taught others statistics to know that. Obviously with drop rates we would know to limit the drop rate to 0, but just applying a standard confidence interval calculation to a 6% drop rate over 50 observations would included 0 as a possible option.
    Looking for a new guild? Come check out the Underworld Alliance on Discord:https://discord.gg/wvrYb4Q
  • YaeVizsla wrote: »
    What do you consider is enough?
    That is a more complicated question.

    However, your example- a 6% drop rate in 50 attempts- is not even 99% confident that the drop rate is a positive number.

    I understand you are a die hard supporter of CG for all your previous posts but really?
    How could the drop rate be a negative number?

    I studied, tutored, and even was a TA for statistics at a cal state... but I'm learning a ton of new info here on forums it's very interesting.

    Can anyone tell me more about these negative drop rates? Honestly I dont know how that's even possible...
    Given that your initial post about this stated that 50 sims was enough to make a claim, I’ll chime in with this:

    a) 11 sims is enough to make a “claim”, because that’s what the poster did. If someone wants to make a hypothesis on a tiny amount of data there’s nothing stopping them.

    b) How many sims would be enough to draw a conclusion? It’s certainly >50.
  • It blows my mind that our player base is more knowledgeable on statistics and variable tables that I presume CG code writers are writing notes to be used in a latter patch.
  • YaeVizsla
    3448 posts Member
    I understand you are a die hard supporter of CG for all your previous posts but really?
    How could the drop rate be a negative number?

    I studied, tutored, and even was a TA for statistics at a cal state... but I'm learning a ton of new info here on forums it's very interesting.

    Can anyone tell me more about these negative drop rates? Honestly I dont know how that's even possible...
    They can't. That's the problem.

    If a standard 99% confidence interval includes values that are LITERALLY IMPOSSIBLE, then that should be a strong indicator that the data set is too small to draw meaningful conclusions.
    Still not a he.
  • thedrjojo
    949 posts Member
    edited May 2020
    A 99% confidence interval is wrong (doesn't contain the actual value) 1% of the time as well... That's better odds than pulling a 330 shard pack, but those happen from time to time too... When you have millions of battles a day, a sample of even a 100, there will be ones that fall out of the CI...
  • Nikoms565
    14242 posts Member
    2 quick points:

    1) CG has stated repeatedly that they don't/haven't changed drop rates on gear.

    2) Anyone that has actually tracked data over a sufficient data set proves it.
    In game name: Lucas Gregory FORMER PLAYER - - - -"Whale blah grump poooop." - Ouchie

    In game guild: TNR Uprising
    I beat the REAL T7 Yoda (not the nerfed one) and did so before mods were there to help
    *This space left intentionally blank*
  • Fanatic
    415 posts Member
    edited May 2020
    It blows my mind that our player base is more knowledgeable on statistics and variable tables that I presume CG code writers are writing notes to be used in a latter patch.

    What relevance does CG's ability to do statistics (or there lack thereof) have anything to do with this discussion? But do you really think they've made a billion dollars without understanding statistics? Do you really think they don't have a good idea of what percentage of their players spend, and how much, and can't with reasonable certainty predict how much revenue a new feature will generate (within a margin)?

    They have shortcomings, and faults, but I don't think lack of understanding statistics is one of them.
    Post edited by Fanatic on

  • Can anyone tell me more about these negative drop rates?

    Shards that you have are deleted.
    Make Bronzium autoplay opening an option.
Sign In or Register to comment.