Saturday 12 August 2017

Analysis of Greater Rift Clearing Efficiency Using S10 Data: Comparison between the top and bottom performers (Wizard & Barbarians)

Saturday, August 12, 2017  sQren  



Here is a greater rift clearing efficiency chart that I made using Season 10 data for Barbarians and Wizards (Took me 3 hours). Hopefully, this will provide a perspective to the community not to argue among each other. And I hope this reaches the devs. well since they could use the calculated data in order to balance the classes properly among each other.

Background:

In this thread, I will try to compare the barbarian and the wizard and their power efficiency in clearing high greater rift levels. This will aid to show the existing power creep between the barbarian (the weakest class) and the wizard (strongest class). This will outline the class that are in need of major buff in order to achieve equilibrium at end game.

The Methodology:
  1. Using Season 10 leaderboards (as of today), I selected the top 210 barbarian players from greater rift 100 to greater rift 81 levels.
  2. Wizards needed to be selected differently since they are able to clear way higher than the barbarians so I selected 10 highest performing and 10 lowest performing Wizards at the compared greater rift levels. The total number of the selected Wizard players are 152.
  3. The parameters used in the analysis are Greater Rift Level (A), Clearing Time (B), and Paragon Level (C)
  4. For each toon, a greater rift clearing efficiency coefficient (P) has been calculated as follows: P=A/(B*C) The resulting numbers are scaled up with 1M in order to ease the comparisons.

The Explanation of the Formula:

According to this formula, the efficiency in clearing grifts of a toon is defined as the highest possible greater rift level to be cleared as fast as possible with the lowest paragon level.

Results:

When we look at this chart, we can observe the scattered barbarian performance indicators as blue while the same indicator values are shown in orange for Wizards.

Analysis of Results:


  1. All wizard clears are above 88 in the 1st 1000 leaderboard clears
  2. All barbarian clears are above 50 and 75 in the 1st 1000 leaderboard clears
  3. The clear efficiency coefficient distribution of barbarians at 81 are very similar to the wizards' at 92, which suggests approximately 11 greater rift difference.
  4. Using regression, the means for the barbarians and wizards are plotted in blue and orange respectively.
  5. When these means are compared, it can be seen that the barbarian clearing efficiency mean at 81 is equal to the wizards' at 92 (about 11 grift difference). Also, the barbarian clearing efficiency mean at 85 is equal to the wizards' at 95 (about 10 grift difference).
  6. NEW: Barbarian requires approximately 250 (1250 STR) paragon levels more at mid-late 80s while this becomes 500 at mid-late 90s (2500 STR), which could be even more considering there are only a handful of +90 barbarian clears compared to the wizard +90 clears. Purely paragon comparison suggest at least 6 greater rift levels head start of wizard class. However, paragon comparison is not a good representation for efficiency. As shown in the initial graphs, Wizards finish greater rifts much faster and more effectively at lower paragon requirements, which is expected to amplify the 6 greater rift difference to higher values, as suggested in this study to 10-11 greater rift levels.

Due to the exponential nature of the greater rift difficulties, the clearing efficiency of classes seems to decrease accordingly, but barbarian takes the biggest hit due to the lack of DPS buffs especially after 81-82 greater rift difficulty level (Clears are more consistent only up to 83 after which the clears get significantly less). This can suggest that average barbarian has a cap of -83 Grift performance (unlike some streamers suggesting +90, +95).

Verdict:

Barbarians need a buff equivalent of 10 greater rifts. From the given chart, it is obvious that if the barbarian class is buffed as such, the expected rift clearing performance of the class will overlap the Wizards'.


How you can use the information in these graphs?

Case study examples:

You can use these approximations in order to gauge your greater rift potential considering your paragon levels, greater rift level and desired clear time. For example,
  • Assume that I would like to clear 95 at 14.59 min=14*60+59=899 seconds. Then what I do is to go to the power efficiency graph and find the mean value corresponding to the grift level:

At 95 grift level, class clear efficiency is approximately 75. When we rework the formula for the paragon level, C=(95x1000000)/(75x899)=1409 Paragon levels required for the slowest possible clear (note this is average Joe approximation, better clears can happen due to the associated RNG)
  • Assume that I would like to do grift 100 at 899 seconds. As we can read from the graph, the power efficiency level is about 60 at 100 grift level so: C=(100x1000000)/(60x899)=1853 Paragon levels required for the slowest possible clear (again average Joe approximation).

I am assuming this 60 value stay more or less constant after 100 so let's check Archael's clear: Archael cleared 101 at 13m, 20.950 seconds (800.95) at 4600 paragon (don't know its exact value right now but profile is at this value). So what our graphs estimates for Archael: C=(101x1000000)/(60x800.95)=2102 Paragon clear requirement. So Archael well above satisfied this value. As a result his exceptional clear time.
  • Let's check the new top leader: Clear 103 at 14m 56.416s=896.416sec C=(103x1000000)/(60x896.416)=1915 Paragon level required for barely clearing the grift where this person has 3600 paragon (again well above, but a bare clear).

There will be another one of these analyses within 1 or 2 weeks for Necromancer, Wizard and the Barbarian in order to show the new skew in the existing power creep towards Necro and Wizard. However, S11 data will be very biased since many are favoring towards the Necromancer (given that is is a new addition).

Thanks for reading.

Posted in: