I agree that my analysis has limitations. I would consider my analysis an approximation.
All analyses will suffer from one critical factor that none of us know, specifically how many people are pushing on a per class basis to get on the leaderboards.
For a better analysis, you will need to think of all, or at least the most significant, relevant variables. In my mind besides greater rift level cleared and time taken, these are:
paragon level
legendary gem level used for the clear’#
augments
hours played’*
The problem that I had from my simple analysis (and later using the json file that let me get the data more easily) is that I could pull greater rift clears, how long the clear took, and paragon levels. I had to manually get the other things like legendary gem levels and augments. Unless there is some other way to get each of these (augments and legendary gems at the least), any analysis can easily be criticized.
Notes:
'# You will need to know what types of legendary gems are equipped. In my analysis in another thread, I corrected for legendary gems that cap at max level under 150 (e.g. esoteric caps at 100) and pain enhancer. For example, 5 of the top 10 barbarians at the time used esoteric. Anther individual (forget which class) used pain enhancer where it was level 25. You can imagine that could introduce critical artifacts into the data.
'* This one may not be so important; however, you can imagine a scenario where 2 classes can achieve the same power, but it takes one class twice as long. This might be nice to know. For example, crusaders in the past averaged the best solo clears (not any more). They are also known in some cases for having “inhumane” hours if you understand my euphemism.
Blizzard APIs do not provide gem levels, augments, or the sets used. Without these data, you cannot effectively consider them. Besides, gem, caldessan levels are 100% in correlation with paragon. A 100s gem levels will not magically clear a 130 gr. If it does, the data will be an outlier data and wont affect the overall.
Therefore, gem and caldessans can be assumed to be more or less in line with paragon. In such analyses, you cannot be very precise but can only be approximate.
In your approach, your fundamentals are wrong: you cannot average grs and give a number like 125.6 gr. You cannot average vsrious gr levels to be representative of a class. You will get weird results which will contradict the reality.
You are right it is an exponential and not linear. That is why I included a rank order. In terms of statistics, there is a major assumption about normality and homoscedasticity of the data. It is not clear to me that these assumptions are met. If not, non-parametric statistical test will be necessary.
I had a small sample size, but this correlation was not 100%. For example, wizards had the highest paragon ~3700 but were much weaker than monks and necromancers relative to their legendary gem levels @ 3400 paragon.
Top 10 are irrelevant. I suggest look at the bottom 10 as well at each GR level. Then your opinions may change. However, looking at thousands of data points, the significance of gems and augments will make minor difference for the mean.
I am not very sure what is the point of your analyses. If you want to show how many GRs worth of buff needed, we have a tried and valudated method already, which confirmed the devs buff numbers in the past and resulted in a very well balanced DPS across classe(back in S12).
I looked at the top 10, top 100, top 1000, median, and range. As I mentioned before, it really does not matter much how the data is parsed in this regard.
Also this:
In another post looking at this era and season only, I took the top 40 clears worldwide as an attempt to get around confounders of legendary gems, augments, and paragon, because I figure these people maxed out their gear well.
Dude, you cannot average different GR levels and give a number for a class as a whole. There are many players who will not even reach 100gr.
The only thing that matters is how easy it is to do a certain GR for an average Joe player of a given class. Average Joe DH doing 110 with their whatever build is only achiavable by a well geared playet snd not by an average Joe barb.
That is the problem with your observations. It is twisting the reality and not correctly giving the mean perdpective.
What greater rift level do you think that we should consider as the “target” GR level for the average Joe for this calculation? The average Joe is not in the top 1000 in the leaderboard unless there are a lot less people playing the game than I think.
And you learned, hopefully, to not make a post unless you know what you are talking about. There is no class warfare, only bull and those whom called out the bull.
My OP clearly shows data that anyone can verify for themselves by going to Blizzard’s website and clicking on “Rankings”. I just made it easier to look at relative top the top 10 solo clears for the last 4 eras and 6 seasons.
We can discuss whether the top 10 is representative or not.
We can discuss whether we should be using arithmetic versus geometric means to think about “average” clears.
We can discuss what things should be considered (paragon levels, legendary gems, augments).
We can discuss about the best model to fit the data.
We can discuss about whether “the best” analysis will use parametric versus non-parametric statistics based on the assumptions of each.
So on and so forth…
To my knowledge, no one has done a perfect analysis. As Prokham confirmed, there is no easy way in his knowledge to get information about augments and legendary gems. I did this manually in another thread for hours played and legendary gem levels. To do it completely “right” is impossible. The next best thing is a reasonable approximation.
If you can compare classes better, I would like to see it so I can learn. It is easy to criticize.
@TuneOut
Here is another global comparison. It uses the top 10 solo greater rifts clears from the current era across 3 regions (the data was grabbed on a different day so that is why the numbers are different.)
Do you think that this analysis is correct or is it fatally flawed and should be ignores?
On topic now. If I remember correctly Blizzards API could be used to pull data from server about clears that are not on leaderboards. There was a website long time ago I think it was diblorank were you could also sort clears by set used.
It does. I downloaded the data for a class as a test. It goes to position 2000. The problem is that file does not list legendary gems or augments. You have to manually look at them one by one.
Maybe it was for clears that where on leader-board but got overwritten buy something else. I clearly remember seeing more than one clear that I participated in. Or maybe that website had its own archive. I did not try API myself.
I do know that! But it’s a little weaker than Thorns and could use a buff. I don’t even think it has to be a big one. Does anyone have numbers on that? Thorns VS DH RG DPS?
Do you think that this analysis is correct or is it fatally flawed and should be ignores?