Eternium
Eternium

Announcement

Collapse
No announcement yet.

Trial Clearing Time Research

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

    Trial Clearing Time Research

    Following a discussion about the best trial clear time to maximize XP, I ran an experiment using a Stalker BH with 1995 CLs and a boosted XP of 608%. I picked the Stalker BH because that is my primary class, and that hero has the ability to clear TL 110 in about 3 minutes. I ran each trial to the boss kill, and in most cases, the boss fight was fairly quick. I did die in a couple of TL110 runs, but that did not seem to impact the clear time very much (just a few seconds to reset), so I included those runs in the average. I ran each trial level 10 times and used the average XP earned and average number of seconds to clear that trial level. Then, I extrapolated those averages to estimate what could be expected using a 30 minute XP booster.
    Trial Average XP Earned Average Clear Time (seconds) Seconds in 30 mins runs per 30 mins total XP earned (million) Legendary Gear drops in 30 mins
    80 9.60 77.1 1800 23.3 224.1 70
    90 15.35 97.5 1800 18.5 283.4 55
    100 25.10 130.3 1800 13.8 346.7 41
    110 28.70 183.1 1800 9.8 282.1 29
    What I learned:
    • XP earned from clearing a trial is very consistent with almost no variance.
    • Clear time was also fairly consistent, but there are occasional maps where you have to double back that add a few more seconds.
    Deciding what trial clear target to shoot for in an ANB is a bit of a balancing act, but the data support the position that farming XP beyond TL100 is only worthwhile if you can clear those trials faster than 3 minutes. For this BH, the "sweet spot" was a clear time of just about 2 minutes to earn the most XP in 30 minutes (346.7M). The overall XP dropped off for TL110 because, even though each clear awarded slightly more XP, the higher time to clear meant fewer runs in a 30 minute period.

    However, for an ANB (where gear drops are more important to upgrade gear and to collect salvage materials) a faster clear time might be more valuable in the long run because you collect more gear drops in 30 minutes than a player running higher level trials with a longer clear time. Of the 40 pieces of gear dropped during the experiment most was level 70-72 stuff that was only good for salvage. I did get one level 77 piece, but it was not a set gear piece and had poor stats. I managed to upgrade a couple of pieces of gold farming gear (level 74 boots and a level 76 trinket with XG stats), and I got a level 70 BoM with a scatter shot bonus with XG stats that I also added to my BH gold farmer. I upgraded one piece of Demolitionist gear, but aside from that, the rest went to salvage. Keep in mind that my main account is close to endgame level, so for an ANB hero, I most likely would have been able to upgrade at least a couple of pieces of gear.

    Just for fun I expanded the table to see what would happen after 3 hours of XP farming. Here the difference in materials collected becomes even more noticeable. Running those higher trials at slower clear speeds could mean the difference between having materials to craft versus having to spend gems on materials crates. More likely, it could mean getting a lucky gear drop that will allow you to run trials faster and be able to move up to higher trials sooner than a competitor who is grinding with higher clear times.
    Trial Average XP Earned Average Clear Time (seconds) Seconds in 3 hours runs per 3 hours total XP earned Legendary Gear drops in 30 mins
    80 9.60 77.1 10800 140.1 1344.7 420
    90 15.35 97.5 10800 110.8 1700.3 332
    100 25.10 130.3 10800 82.9 2080.4 249
    110 28.70 183.1 10800 59.0 1692.8 177
    If MoTs are a goal, then you may want to run higher level trials where the set gear drops are higher (e.g., at TL110 73% of the drops were set pieces but that dropped to 60% at TL100), but you have to keep in mind that it will most likely cost you XPs to do that.

    Overall, I am leaning toward adopting a 90 second trial clear time goal for the next ANB. The data suggest that if you are spending more than about 2 minutes to clear a trial, then you are sacrificing XP for whatever other goal you are chasing. The bottom line is that the old clear time goal of "3-4 minutes" seems to be obsolete in the most recent version of the game.
    Last edited by Ozymandius; 08-26-2020, 02:37 PM.

    #2
    That is almost exactly the strategy I used this past Bronze ANB. I got to CL 1851 with over 4 hours left on the game timer then RL issues arrived and I had to stop grinding and finished in T137 with all that time remaining. Arg...

    Good write up, thanks.

    Comment


      #3
      Ozymandius did you keep the time trial stats? Curious what the variance on the times where for trials. We could estimate the confidence interval for the completion.

      Comment


        #4
        NoSuchReality Here are the times for each trial, with the means and standard deviations:
        TL80 TL90 TL100 TL110
        seconds seconds seconds seconds
        72 89 124 154
        65 102 120 195
        73 119 123 208
        78 100 138 199
        79 106 149 184
        74 100 143 171
        94 97 124 163
        75 73 148 174
        75 89 102 187
        86 100 132 196
        77.1 97.5 130.3 183.1
        8.01 12.10 14.63 17.25

        Keep in mind that this is a fairly small sample size. It might be more statistically accurate to calculate a trimmed mean by tossing out the fastest and slowest times as chance variations. That would give you 75, 100, 128, and 185.5 seconds. The conclusion is still the same using those numbers.

        There was a bit of complaining when the Devs added a 30 second penalty for skipping certain trial maps a while back. There are certainly slower maps (e.g., those with lots of long blind alleys), and I know that some players tend to avoid the really dark caves because of poor screen visibility (especially on mobile) and it is harder to pull mobs with all of the rock formations in the way. I suppose you could save time by playing the first 30 seconds to avoid the time penalty then bailing out, but a BH's movement speed tends to make up for some map variations. I found the most difficult maps to be those where you are surrounded by elite mobs at the beginning, and you have to spend some time dodging charges and counterattacks.

        Comment


          #5
          Actually its the outliers I'm concerned about. My outliers tend to be worse, an eyeball estimate will be I'm clearing 1:30, but my bad run plops up at 2:10, 2:15 with a much more skewed eyeball on it. This may just be selective bias during ANB, i.e on your TL80 runs, one may feel like they're running it in 70 seconds after the first three runs.

          TL80 at the worse run is 1103 in 3 hours
          TL90 becomes 1393 in three hours.
          TL100 becomes 1819
          TL110 becomes 1490.

          Thinking ANB, maybe the goal is push trial until you hit the 90-120 seconds mark. If you make a single run under 90 move up, if you make a single run over 120 seconds, move down. When the run again is below 90, move up unril farming TL100 Etc. I will need to test that, hopefully break some of that stall in TL60-70.

          ​​​​​

          Comment


          • Ozymandius
            Ozymandius commented
            Editing a comment
            I mainly ran this test because I did poorly in Bronze using the old "run the highest trial that you can finish in 3-4 minutes" strategy. In the past, I was able to make that strategy work by spending gems on 200% XP boosters all the way, but I am gem poor right now. I only got to about CL 872, even though I still burnt gems on several 100% XP boosters. Discussions with players who are competitive without using gems on boosters are using a much lower "run trials that you can complete in about 90 seconds" strategy to grind over 1200 CLs while only using the free boosters from the event. I figure that it is worth a try, especially since the data support a lower clear time goal than the one that I had been using.

            FYI, I wouldn't really sweat one or two slow times unless I was getting one-shot by mobs; it is the average pattern that matters most. If you treat the best and worst run times as true outliers (i.e., chance variations), you probably get better prediction accuracy.

          #6
          For an ANB it might be meaningful to distinguish Legendary and Set piece drops? Don’t you start to get more set pieces after TL 90 or 100?

          Comment


            #7
            TL 90 drops more set pieces.

            How many gems you plan to spend and how lucky or unlucky your crating is plays in here too. I went with no leg crates in bronze, bracer crafting decimated my silver ore stack, and getting really bad rolls on damage on my staff meant I needed to run 80 mostly. Was clearing in between 45-60s, and needed all those materials. With bracer crafting consuming so much silver ore, I was never low on MoTs.

            Comment


              #8
              Thank you for gathering the data Ozymandius!

              What you found seems to agree with a similar experiment I did a while back. But now I am utterly confused. Here on the forum I keep reading that XP earned and gear drops from trial bosses improve with higher TL until the cap at TL 120. On discord however, I keep reading that the cap is actually at TL 102. And Montresor says that TL 90 is better for set pieces...

              During the last ANB bronze I "wasted" a lot of time farming TL 110, got only crap from drops. I went back to farming TL 105 which gave me the same amount of XP earned, using the CLs gained as a proxy for XP.

              So, can anybody with more knowledge of the game chime in and confirm where the cap is? Pretty please?
              KUSA GAJE GASU 4903

              CL 2100
              Anisaz, Extra Gold BH, TL 7 (too lazy)
              Keril, confused Mage, TL 7 (can't be bothered)
              Ohad, shieldmaiden, TL 115

              Comment


              • Ozymandius
                Ozymandius commented
                Editing a comment
                bojck I made 10 runs each at TL 80, 90, 100, 110, but my primary focus was recording XP and trial time. Most of the 120 pieces of dropped gear were level 71-72, and only a couple of pieces were "keepers" (and those were only XG farming pieces), so overall I found the quality to be quite poor. I did notice more set gear drops at higher trials, but I did not run anything above TL110 for this research.

                I agree that a more valid test of item quality would be to record the level of each piece dropped, so you can analyze the data more scientifically. If you want to collect the data, I would be more than happy to analyze it.

                Barring some breakthrough from additional research, I would consider drops to be mostly salvage quality gear for materials, with a rare chance at randomly rolling a usable piece.
                Last edited by Ozymandius; 08-27-2020, 01:47 PM.

              • bojck
                bojck commented
                Editing a comment
                yup, we're saying the same thing : ) I'm saying that the chance for +1 level on t120 was 50-55%, starting from item level 70. This means around 1% of drops are 77s. Of those most are not from the set you need, or don't have the stats you need. So if you get a useful drop in a few hundred t120 runs, you're lucky. I think there's no need to experiment with this any further...

                For xp you could test a bit to verify Lodwig's calculations yet again, but they seem to be spot on

              • Montresor
                Montresor commented
                Editing a comment
                So I did the same thing with TL 115, 110 and 105 I think except I did 17 runs each, so 51 items.
                Interestingly the *average* drop stayed very close to 71, but the chance of a 74+ item increased with trial level.

                I did not track set vs non set. At the time I did notice that there were zero new uniques at 105, although they seem to have changed the new item drop rate since then (this was in that dark period between the beta when it was raining new items, and the introduction of glory, and in that period it was like a new item desert)

              #9
              Here's the data I gathered

              Comment


              • Ozymandius
                Ozymandius commented
                Editing a comment
                Wow, if you just go by averages, there isn't really any advantage to running trials higher than 90. However, does look like there might be some benefit for higher level items at higher trials. Unfortunately, it is really hard to create predictive models for events, like a level 77 drop, that are so rare (less than 1% in your sample). Thanks for sharing your results!

              #10
              Anecdotally for low trials, it seems like there is very much a drop buff similar to the trial unlocks for fast clears. Speedy clears in the mid-range TL40-60 drop more legendaries (or at least it seems). Unfortunately, my consistency in trials TL90-TL100 is not good enough to consistently pull sub-60 second clears and my patience is lacking for sandbagging 3:00 to 3:10 clears on the same trial for a sample large enough to compare level of drop, named/set counts and if needed attribute averages as a percentage of max.

              Comment


              • Ozymandius
                Ozymandius commented
                Editing a comment
                I hear you. If things are slow this weekend, I may give it a try to see if I can replicate @Montressor's data.

              • NoSuchReality
                NoSuchReality commented
                Editing a comment
                I’ve often wondered if the drops are independent or if the three drops are a single instance. Most of our analysis have assumed independent. I.e does @Montressor’s data have 27 data points per level or 51.
                Last edited by NoSuchReality; 08-27-2020, 08:02 PM.

              • Ozymandius
                Ozymandius commented
                Editing a comment
                Given that I have never observed any pattern in the drops, I would count each piece as an independent event. I just ran Trial 90 ten times and got 2 level 77 drops out of 30 pieces of gear, so it really illuminates the problem or trying to uncover a pattern in extremely rare events like rolling a level 77 piece.

                I had 60% set gear in yesterday's runs in Trial 100 and 73% in Trial 110, and it looked like a pattern. But I only had 36.7% set gear in today's runs of Trial 102. That is way more random variance than can be tracked by running a trial 10 times.

              #11
              I updated the table adding stats for trials 85, 95, & 105:
              Trial Average XP Earned Average Clear Time (seconds) Seconds in 30 mins runs per 30 min booster total XP earned Legendary Gear drops in 30 mins
              80 9.60 77.1 1800 23.3 224.1 70
              85 11.83 87.6 1800 20.5 243.2 62
              90 15.35 97.5 1800 18.5 283.4 55
              95 19.30 110.1 1800 16.3 315.4 49
              100 25.10 130.3 1800 13.8 346.7 41
              105 27.71 150.0 1800 12.0 332.5 36
              110 28.70 183.1 1800 9.8 282.1 29

              You can see a clear trend of increasing XP by trial level which starts to level off between trial 100 and 105 and becomes almost flat after trial 105. So even without taking clear time into account, you can see diminishing returns for running higher trials.

              Click image for larger version  Name:	trend line.JPG Views:	0 Size:	18.9 KB ID:	179958


              The points that are slightly off the line probably have more to do with the variation from only running each trial 10 times. I am sure that increasing the number of trials would make the pattern even clearer, but I don't see the need for further analysis given how clear the trend line turned out.

              Comment


              • NoSuchReality
                NoSuchReality commented
                Editing a comment
                The second chart should be normalized for per second or collective 30 minute booster.

                Per booster would allow you to add the 2nd scale and plot drops as a 2nd line.

                Still amazed at how linear your completion times look across 30 levels. Actually it’s 10, 10, 10, 20, 20, 30. Might need to check mine, seems like my turn up is more abrupt, but I may be thinking over three minutes.
                Last edited by NoSuchReality; 08-29-2020, 03:43 AM.

              #12
              Hi, sorry for the newbie question: do you think the chart can apply to warrior as well (with different finishing times, obviously)?

              Comment


              • Ozymandius
                Ozymandius commented
                Editing a comment
                Yes, the trend holds for all three classes of hero. I probably need to update this post to reflect the changes in drop rates since Legendary drops are no longer automatic at TL80, but the main takeaway is that if you are farming trials for XP & materials the best strategy is to run the highest level trial that you can complete in 90 seconds or less.

                FYI, as far as I know, the XP still tops out between TL100 and TL105, so it does not make sense to farm any higher than TL105.

              #13
              Thanks for the very useful data. Another interesting variable to include in your analysis would be the drop rate of Books of Learning. As with the legendary drops, the increased number of runs should also pay off with a higher rate of book drops. However, if the higher level runs consistently drop Greater Books, the increased time may be worth it from that perspective.
              Last edited by jdh0721; 04-28-2021, 03:20 PM.

              Comment


                #14
                As far as drops are concerned, two months ago Rob Spooner ("Wispy") produced four very informative YouTube videos, running TL90/100/110 and 120 each 150 times, listing the drops and, after Salvaging, the MoTs and Essences.

                The mobile link to the TL90 video:
                https://m.youtube.com/watch?v=bhToE3L8lJk
                Last edited by Medamanaman; 04-29-2021, 05:53 PM.

                Comment

                Working...
                X