Eternium
Eternium

Announcement

Collapse
No announcement yet.

Drop rate MoTs (.34)

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Drop rate MoTs (.34)


    This week was running TL 80 every day, ~ 60 races a day. Almost always drop some named or defined item

    However, with this last update (.34) I realized that it is very difficult to drop some item named or defined in TL 80 and 85, even killing Kara in less than 2min.

    Already in TL 90 nothing has changed.
    Last edited by RMeru; 01-18-2019, 11:42 AM.

  • #2
    I have been keeping track of Set items vs. Normal items in groups of runs of Trials.

    Presumably, there is an overall drop rate, which is the probability of an item being a set item. Presumably that probability is a constant for each trial level, and increases with increasing level to a maximum or cap at trial 100.

    I am now keeping track over many many runs and intend to eventually collate the data together to get a better estimate of the "true" drop rate. I happen to be doing that for trial 92.

    I have noticed a group of runs were the drop rate I observe is low (about 30%), and I have also had a group where it is quite high (81% over 7 runs).

    That is the problem with sampling from a population. There can be quite a bit of variability between the samples and you need a massive amount of data to get a high confidence estimate of the overall mean from the population.

    After the variability of % Set Items, then you have additional compounded variability, from the salvage rate of MoTs. From individual items I have seen zero, up to 4 MoTs. Collecting data on MoTs means it's harder to understand because there are two embedded drop rates.

    This is why I am now only collecting data on Set items and total "simplified" resources, and then using UmbraDei data on average salvage rates (for Essences; I don't know that he/she posted average MoT rates from Set items).

    We humans do not accurately perceive variability well at all. We see patterns and perceive changes where there are none. We need data to overcome that tendency.
    Last edited by Coda; 01-18-2019, 02:14 PM.
    VUFO ZEBE ZAQI 1381 -|- Gethi, mage, & Kisheli, XP mage

    Comment


    • #3
      3.34 all set and named items gain 100% MoTs, 100TL - every run I have one set item
      King of Android play
      NODA ZUSI NUHA 1647

      Comment


      • #4
        Coda That's true. I thank you for sharing this information and statistics with us, as I know it's a lot of work.

        Based on my experience over the last few days (before upgrading to 3.34), I received almost one set items per race in TL 80, zeroing the map in less than 2 minutes.

        However, I noticed that after this update (3.34), I almost did not drop set items in TL 80 or 85.

        Already in the races in TL 90, 95 or 100 I receive almost 1 set items per race.

        But you're right, for accurate information, it takes a huge amount of data and testing. Thank you for this work and for sharing with us.
        Last edited by RMeru; 01-18-2019, 03:18 PM.

        Comment


        • #5
          Originally posted by Dyoma View Post
          3.34 all set and named items gain 100% MoTs, 100TL - every run I have one set item
          Thanks for the information! Good to know!

          Comment


          • #6
            I was doing trial 92 and seem to get set items every run or just about. You can't forget about the 25% chance to double your salvage. I don't know if this is per item or if you do a whole dump it counts the whole pile? It be nice to collect 40 or so items and it double the whole lot instead of single items.

            Comment


            • #7
              Let's pretend the "true" drop rate of a set item is 50%. And you get three items. If you label them as S or N, You have these possibilities:
              1. SSS
              2. SSN
              3. SNS
              4. SNN
              5. NSS
              6. NSN
              7. NNS
              8. NNN
              Since "order does not matter" we can combine 2, 3 & 5 as the same, and 4, 6 and 7 as the same.

              That leaves 4 possibilities:
              1. SSS
              2. SSN
              3. NNS
              4. NNN
              But 1 and 4 have only a 1/8 chance, and 2 and 3 have (each) a 3/8 chance.

              If you get a run where you get all 3 S, then is the true drop rate 100% (no).

              If you get 2 runs, where one is SSS and the next one is SSN, is the true drop rate 5/6? (no).

              Keep going.

              This is why you need a LOT of data to estimate with confidence, the "true" mean.


              VUFO ZEBE ZAQI 1381 -|- Gethi, mage, & Kisheli, XP mage

              Comment


              • #8
                Originally posted by Dyoma View Post
                3.34 all set and named items gain 100% MoTs, 100TL - every run I have one set item
                When I collected data on salvaging individual Set/Unique items, the number of MoTs ranged from zero to four. If these results are spread evenly you would have 0,1,2,3,4 which averages to "2".

                Therefore, if you really got (only, but exactly) 1 MoT from every (1) Set/Unique item, that is an average of "1".

                On the other hand, if you are saying that "zero" no longer occurs, but the rest do still occur (1,2,3,4) then we now have an average of 2.5 instead of 2.

                That's the issue with statistics... it's hard to state clearly and accurately what's really going on.

                VUFO ZEBE ZAQI 1381 -|- Gethi, mage, & Kisheli, XP mage

                Comment


                • #9
                  For clarity, there are two averages we are trying to define:
                  1. average "drop rate" (probability) of a Set/Unique item (for a given Trial Level)
                  2. average MoTs salvaged per each Set/Unique item
                  These are two separate probabilities that, with careful data collection & analysis, could be determined separately.

                  An open question is, does #2 depend on the quality level of the item, meaning how high are the stats of the item.
                  VUFO ZEBE ZAQI 1381 -|- Gethi, mage, & Kisheli, XP mage

                  Comment


                  • #10
                    Originally posted by Coda View Post
                    For clarity, there are two averages we are trying to define:
                    1. average "drop rate" (probability) of a Set/Unique item (for a given Trial Level)
                    2. average MoTs salvaged per each Set/Unique item
                    These are two separate probabilities that, with careful data collection & analysis, could be determined separately.

                    An open question is, does #2 depend on the quality level of the item, meaning how high are the stats of the item.
                    1- So far, for me, the Set/Named drop rate (compared to regular Legendary) *feels* like it isn't less than: t81=~33%, t91= ~50%, t101= ~66%. How does this compare to the data you've collected so far?

                    2: Are we sure it isn't 1 or 2, being doubled by Crafting Mastery up to 2 or 4 at times? I don't recall seeing a 3, but I do typically salvage 30-40 Legendaries at once, instead of one at a time.

                    Open question: How would you measure this? For example, 30 Set/Named item batches from ~80 vs 30 from ~100. How large of a sample size (repetition of 30 vs 30 comparison salvages) would it take for you to feel confident in the result? This example does take for granted the drops are "better" at higher trials, in terms of an internal number modifying the stats (kind of like crafting at 0 Mastery vs 50). If there is no such internal number linked to items, would we be looking at segregating items with (primarily) scores in the 70s vs items with multiple 85+ scores?

                    And then , of course, there is the worry our devs are doing their own internal datamining with ALL the data, and subtley adjusting drop rates as needed, invalidating our data.
                    Last edited by Trouble; 01-19-2019, 03:44 PM.

                    Comment


                    • #11
                      Trouble
                      ​​​​​​​
                      I don't know how much depth you or others care to go into. It's just a game. What we are trying to do is estimate the mean of a population from taking samples, and a confidence interval of the mean, so we know how close the estimate is to the "true" value.

                      You can google about the statistics of this yourself. Here's one example of an explanation:

                      http://sphweb.bumc.bu.edu/otlt/MPH-M...ple_print.html
                      VUFO ZEBE ZAQI 1381 -|- Gethi, mage, & Kisheli, XP mage

                      Comment


                      • #12
                        Coda I understand. I was asking your opinion, because you seem to intend on digging deeper than I am.

                        Comment


                        • #13
                          Originally posted by Trouble View Post
                          Set/Named drop rate : t81=~33%, t91= ~50%, t101= ~66%. How does this compare to the data you've collected so far?
                          My data is approximately similar.

                          Originally posted by Trouble View Post
                          Are we sure it isn't 1 or 2, being doubled by Crafting Mastery up to 2 or 4 at times?
                          No, not sure. Good point.

                          Originally posted by Trouble View Post
                          And then , of course, there is the worry our devs are doing their own internal datamining with ALL the data, and subtley adjusting drop rates as needed, invalidating our data.
                          I agree. Hopefully they aren't that cruel.
                          VUFO ZEBE ZAQI 1381 -|- Gethi, mage, & Kisheli, XP mage

                          Comment


                          • #14
                            the only problem is we need to compare drop rates with people who have equal or close ML. Also the chance to double salvage. I get 25% at ML50 and I think they might come out with higher ML because it shows you what the next drop rate is even though you are maxed. Also, I don't know if that 25% to double your salvage is per item or per the whole salvage. Statiscally you should get the same 25% amount of double but if it is per total salvage you get a crap ton more then 3 regular salvage. If it is per item than statistically 1/4 of your items get doubled but which items is just random when it comes to named/set items vs regular legendaries. I am running TL98 right now but 92 was working out to be the same drops just about when it came to named/set items.

                            So if you would like, I am on board. We just need to agree on an amount to salvage, like 40 items and do we do all set/named items and all legendaries seperately or half and half or just whatever you got? I say we seperate them which in that case probably lower than 40 items, maybe like 20 or so. Take SS of them drops so you can save the image and make a excel or whatever way you can document it on paper too. Then we agree on trial levels. They don't have to be the same trial but if we do the same trial we can get more data for that trial quicker. I don't think it matters but do different trial bosses have different drop rates? I highly doubt it but am wondering if anyone has noticed any of that

                            Just count the number of runs you do, take screen shots of your salvage, make a typed up excel copy, run the same trials as each other and probably be best if we seperate named/set items from regular legendaries. If you want to salvage items 1 at a time, we can do that too, just to get a better idea of what the max or near max values are. I was just thinking taking 20 items or whatever and dividing my slavage by 20 to get a overall average. LMK what you want to do because I am actually in the process of doing this.

                            Thank you,
                            Andrew

                            Comment


                            • #15
                              Originally posted by andrewcb88 View Post
                              the only problem is we need to compare drop rates with people who have equal or close ML. Also the chance to double salvage.
                              That's not quite right. One needs to compare salvage rates at equal ML level, or, do the math to normalize between two that are dissimilar.

                              However, the drop rate of Set/Unique items is independent of Mastery Level and as far as we know only depends on Trial Level.

                              This is why I have been keeping track of drop rate and using UmbraDei data on average salvage rate (for Essences) ... because it removes at least one variable. Again, he/she did not post data on MoT salvage rates from Set/Unique items, only Essences, but the same concept applies.
                              VUFO ZEBE ZAQI 1381 -|- Gethi, mage, & Kisheli, XP mage

                              Comment

                              Working...
                              X