Eternium
Eternium

Announcement

Collapse
No announcement yet.

Demystifying the Optimizer Crafting Perk

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

  • Demystifying the Optimizer Crafting Perk

    I've been looking around the forums but haven't found anyone going into detail on how the most interesting Crafting Perk actually works: Optimizer.

    The text looks reasonable.
    Average attribute value increased by xx.x%.
    But what does this actually do, and will it help us create truly end-game level items?


    I've run a few test to try and shed some light on here.
    Spoiler alert: the perk does not really do what the text claims it does.


    In total I crafted some 200+ lvl5 Uncommon Heavy Caps with Crit Rating as attribute, and noted the values that came out as soon as the box opens to avoid rounding effects. Any [perfect] attributes were ignored since we can't distinguish these from the result of the Finesse perk instead. This resulted in 196 useable results. My Mastery Level is lvl26 so I'm supposed to get a 13.0% increase in average attribute value.

    The [perfect] value for a lvl5 CR attribute is 5.69, and that means the minimum value should then be 4.55.

    The expected average would be 5.12. My result set has an average of 5.1737, only about 1% higher. What's happening here?

    I've normalised all 196 attribute values to be in the interval of [-0.5,0.5], meaning we calculate for every item whether they're better or worse than our expected average. The result can be seen in the first attached picture.


    Something interesting happens here: it seems that we're missing the bottom part of the graph entirely. There's much more values above 0 than below. So where does this 13% Optimizer perk come to play?



    In the second attached picture you see all result values sorted numerically. In a perfectly uniform random process we expect this to approximate a straight line, and it does do nicely, indicating the values indeed still likely follow a linear distribution.
    However, looking at the minimum in the result set we see that it's -0.3699 normalised, or almost exactly 0.13 above the minimum. I'm not going to run any further statistical analysis to calculate the odds of this happening, but it show nicely what the optimizer perk does:

    Instead of raising the average value by xx.x%, the Optimizer perk simply ensures no values get rolled in the worst possible xx.x% bracket.

    De facto, the average goes up with only half the percentage in the perk, relatively to the minimum roll available.

    What does this mean for end-game items?

    At Mastery Level 50, your attributes will always roll in the 76.5-90 bracket instead of 72-90. For reference, a Legendary Box rolls 86-90. But with Crafting we still have that 15% chance of a perfect roll as well ;-).

  • #2
    While you're at it, what are the chances that values will fall below the 80+ attribute mark? I don't have your data to interpolate.
    Mobile Player
    Altheo (M) TL109 ¦ Carlisle (XM) TL104
    (and 2 more heroes at TL101)

    Comment


    • #3
      UmbraDei

      Very nicely done! I am very impressed. We need more like this.

      There are several other things for which the underlying math does not match the in-game description. Devs have not fixed these long-standing 'bugs'.

      Anyways, a lot of people do not understand statistics and RNG but you clearly do, very well. Welcome to the forum!
      VUFO ZEBE ZAQI 1381 -|- Gethi, mage, & Kisheli, XP mage

      Comment


      • #4
        Originally posted by UmbraDei View Post
        Instead of raising the average value by xx.x%, the Optimizer perk simply ensures no values get rolled in the worst possible xx.x% bracket.
        I could have told you that when I read the description of the (skill? perk?)
        ANB #1- Mage Rank 17... T107 in 9:47
        ANB #2- Mage Rank 1..... T116 in 9:54

        Live LB- Mage Rank 3..... T122 in 9:05.
        Live LB- XP Mage ........... T112 in 7:55/ T100 in 3:25.

        Click here to see my gameplay videos.

        Comment


        • #5
          Originally posted by Altheo 42 View Post
          While you're at it, what are the chances that values will fall below the 80+ attribute mark? I don't have your data to interpolate.
          The data shows pretty much uniform distribution, so extrapolating that means that a Mastery Level 50 results in (80-76.5) = 3.5 possible values below 80, out of (90-76.5) = 13.5 possible rolls value, giving 25.93% chance of a <80 roll.

          On a legendary item with 4 attributes that means you'd have (including 15% chance of perfect stats thanks to Finesse perk):
          (1 - (0.85 * 0.2593)) ^ 4 = 36.9% chance every stat is above 80
          and
          1 - 0.85 ^ 4 = 47.8% chance at least one stat is perfect 90.
          Last edited by UmbraDei; 01-04-2019, 12:04 AM.

          Comment


          • #6
            Does it apply when crafting jewelry?

            Comment


            • #7
              Originally posted by EterniumDave23 View Post
              Does it apply when crafting jewelry?
              No. This applies to the new gear crafting system. Jewelry crafting of Uncommons is still random. The only factor you can influence at that stage, is to use gems to get only the desired/selected stat.

              Obviously after the Uncommon step, you are in control.
              VUFO ZEBE ZAQI 1381 -|- Gethi, mage, & Kisheli, XP mage

              Comment


              • #8
                What SHOULD the data have looked like, so as to fit what the description says?

                If it's raised ALL the values, hasn't it increased the average? Or are you saying there should still be low values and those are offset by more HIGHER than average values to attain the overall increase?
                CL 2K+
                Mage- Ladim(#9 Live LB), Proxima(XP)
                TL119 in 9:14 on mobile (iphone8)

                Mage Guide

                Comment


                • #9
                  UmbraDei Thanks once again. Very helpful!

                  Comment


                  • #10
                    Originally posted by lama75 View Post
                    What SHOULD the data have looked like, so as to fit what the description says?

                    If it's raised ALL the values, hasn't it increased the average? Or are you saying there should still be low values and those are offset by more HIGHER than average values to attain the overall increase?
                    Assuming the quote in the OP is from the game description and not paraphrased, you'd expect an increase of 13% of the displayed value. But some quick math shows that isn't at all feasible. Baseline average is only 10% lower than max, so even at 13% you'd expect more than half the rolls to be max. And if that bonus scales to 25% at lvl50, every roll would be max.

                    Rather than "Average attribute value increased by xx.x%" , it seems to be "Minimum attribute value increased by xx.x% of Max-Min, which in effect increases the average by half that".

                    Or to simplify, just move the decimal one spot to the left. Not exactly accurate, but close enough.

                    Comment


                    • #11
                      Originally posted by Buppy View Post

                      Assuming the quote in the OP is from the game description and not paraphrased, you'd expect an increase of 13% of the displayed value. But some quick math shows that isn't at all feasible. Baseline average is only 10% lower than max, so even at 13% you'd expect more than half the rolls to be max. And if that bonus scales to 25% at lvl50, every roll would be max.

                      Rather than "Average attribute value increased by xx.x%" , it seems to be "Minimum attribute value increased by xx.x% of Max-Min, which in effect increases the average by half that".

                      Or to simplify, just move the decimal one spot to the left. Not exactly accurate, but close enough.
                      Ok. So a followup question here is, would it be better to properly implement as the description says, and reduce the benefit? Or leave it as is?

                      Another question is, should this be considered a bug? This analysis makes me wonder if the lack of high quality crafting reported by others at ml50 might be due more to improper implementation.
                      Last edited by lama75; 01-05-2019, 02:46 PM.
                      CL 2K+
                      Mage- Ladim(#9 Live LB), Proxima(XP)
                      TL119 in 9:14 on mobile (iphone8)

                      Mage Guide

                      Comment


                      • #12
                        Originally posted by lama75 View Post

                        Ok. So a followup question here is, would it be better to properly implement as the description says, and reduce the benefit? Or leave it as is?

                        Another question is, should this be considered a bug? This analysis makes me wonder if the lack of high quality crafting reported by others at ml50 might be due more to improper implementation.
                        It's not really a bug, just very suggestive wording. One could argue that the average comes "25% closer to the max value, seen relatively over the 72-90 min-max interval an attribute can roll". So technically there exists an interpretation where the game's wording is correct. It just happens to be the interpretation that results in the least possible benefit for us players.

                        Comment


                        • #13
                          Originally posted by Purple Potato View Post
                          I could have told you that when I read the description of the (skill? perk?)
                          Well then why didn't you so he didn't have to do the math!

                          I dont read them...i just kill things
                          Lord Emperor of Bridge Navigational Systems Engineer for:
                          Eternium 2: Rise of the bounty Hunters
                          see post:https://forum.makingfun.com/forum/et...pply-for-a-job
                          and https://forum.makingfun.com/forum/et...414-eternium-2

                          Do NOT hire Xagan! He works for the enemy!!!

                          Friend code: GEBA WABE WADI 4184

                          If it's spelled bad. Horrible Grammar. Random words. No photo I was talking about. Or cut off mid sentence...
                          It's my tablet, I'm sorry

                          Comment


                          • #14
                            Originally posted by Purple Potato View Post
                            I could have told you that when I read the description of the (skill? perk?)
                            Could you now? I realize I'm no native speaker (then again, neither are the developers...), but to me the Perk description seems to clearly state twice as much as the effective value increase...

                            Comment


                            • #15
                              Originally posted by Jose Sarmento View Post

                              Could you now? I realize I'm no native speaker (then again, neither are the developers...), but to me the Perk description seems to clearly state twice as much as the effective value increase...
                              I read it and thought that the minumum value is increased by 25% (of the range, so if the range is 0-1000, the range is then 250-1000)
                              Same as the minimum value change from gear dropped at trial 100. 50% increase in the minimum range. (72-90 becomes 81-90).
                              ANB #1- Mage Rank 17... T107 in 9:47
                              ANB #2- Mage Rank 1..... T116 in 9:54

                              Live LB- Mage Rank 3..... T122 in 9:05.
                              Live LB- XP Mage ........... T112 in 7:55/ T100 in 3:25.

                              Click here to see my gameplay videos.

                              Comment

                              Working...
                              X