Well, this is a question we get asked a lot and answer a lot, so we thought we would do a little write up about it and then go through our data with you.
UPDATES: Last Change 27/01/2021:
- “How did you work out Loot Scroll and Agris values?” section added to include new calculations based on new Loot Scroll info.
- “Why can’t we get all the data ourselves” section updated with the new total specs and data numbers.
We want this data to be community-driven, so we looked at what the community has supplied to us and used that. Before talking about the methods that were used it’s important to make clear that whilst we have received a good amount of data, it must always be edited/reviewed to ensure that the outliers don’t get through.
For data gathering, we have used google forms and spread it around the community, both in discords, forums, and Reddit. To get a mass amount of data, hundreds, and hundreds of submissions. In this, we asked for ap values, class, and other details (and if you’d like to submit your data please do so here)
The data was checked for blatant errors (usually adding or removing a 0 makes a difference, someone saying they get 900 or 90,000 loot at Aakman, for example, is clearly wrong), and further obvious fake submissions.
Then for each class/spec/zone we ordered the numbers by AP values and started to compare the data. This usually gave us a clear “range” of data. Though we could have put the average we chose not to. We felt that a list like this should show more of what you could get if you played optimally rather than just an average. Realistically most people that come here want to be “above” average and will be, hence why you are here, and therefore this also fueled the decision to show the higher end of the numbers.
To confirm the higher end numbers we cross-referenced our data with that from public records. This included checking class discords and youtube to see if people could get these numbers and had some proof of this. Alongside our own personal testing numbers to back this up further.
The other question that we get is how do we do rankings for the PvE data and then asking if we take into consideration the effort of playing a particular class (e.g. Ninja’s grinding versus a Guardian’s).
We don’t take a class’s effort to grind into consideration as it is very opinion-based, where some people think high APM(Actions per Minute) is a lot of effort, others think that slow classes and less movement play styles are more mental effort to keep them engaged.
Furthermore, our PvE lists are based purely on numbers, they get put into a table and ranked, from highest to lowest. We then check the highest number to the lowest number to see the set deviation. If the set deviation is over a certain limit then we choose what range of rankings to use. A good example of this was Aakman grinders, the difference between the top tier and the bottom tier was so much we had a D/F tier.
This way it keeps it fair, and we try to have a split that represents a bell-shaped curve where we try to have the ‘B’ rating as an average of scores, Whilst there are many other ways to do data management as we believe that this is both a simple yet effective way to rank this.
What was left out of the scope?
With all data projects, a scope was defined and more importantly what we wouldn’t look at (to start with). We have chosen to not include what main hand weapon the user has, the same as not using the level of off-hand. As most will know Blackstar is realistically the ultimate PvE weapon, and Kutum at PEN is also up there, which will increase loot numbers by a decent amount. With this being said to obtain meaningful data we would require much more data than what we currently have available.
Additionally PC spec, connections, and other things like that (more external factors) were also ignored since we would have no control over these and fewer ways to check if they are factual.
How did you work out Loot Scroll and Agris values?
We’ve reworked this section entirely after a lot of testing regarding the figures.
As most people know loot scrolls increase your loot but not quite by the amount they say.
This is mainly true of when your loot is increased by 50% (or 150% or 250%), this is because mobs are unable to drop partial amounts of loot and the number that drops is always rounded down.
What this means is the following; if you were at an area where mobs drop between 1-4 trash items, a 50% increase would give you 1, 3, 4 and 6. This, therefore, lowers the 50% increase in loot and is much closer to 35-40% increase instead.
Whereas 100% loot scrolls do work as they say – doubling will always give you a whole number, and therefore the loot you gain will double. This has lead to us changing some calculations from the data submissions to take this into effect:
- 35% for Blue LS
- 100% for Yellow LS
- 135% for Agris (No Scroll)
- 200% for Agris with Blue LS
- 235% for Agris with Yellow LS
There is a misconception that Yellow LS and Agris don’t stack… this is partially wrong. It’s easy to check, you can even do it yourself and you will get more loot with both Yellow and Agris (235%) than Agris and Blue LS (which is 200%) BUT there is an issue in specific areas because the maximum loot is capped.
This means you CAN get over 200% but won’t be able to hit 250% (as shown above it will be 235%).
Other things that we have noticed is that the loot scrolls affect different areas slightly differently, this is based on the base trash loot that they drop, but this will be something we look at going forward.
Why can’t we get all the data ourselves?
In the perfect world we would, of course, want to get all the data ourselves but let’s look at the numbers to clarify how this isn’t feasible:
- 40 specs (including Hashashin and Nova)
- 4 Areas (and more will be added, so the number increases even more)
- 4 AP Brackets
- 6 Loot Options (No LS, Blue LS, Yellow LS, Agris with No LS, Agris with Blue LS, and Agris with Yellow LS)
This comes in at a whopping 3,840 bits of information. Even if we didn’t look at the different loot options it would still be 640 (for a perspective that’s 640 hours, that’s 26 full-days) with separate grinds to get a single set of numbers. To make “meaningful averages”, and we would want at least 5-10+ sets for each to compare and be able to make min/max statements.
On top of all of this, we would need to get all of the classes to 62 (for an equal playing field) as well as being able to have solid grinding knowledge on the class and all gear levels we currently include.
There are even more factors on top of this, but the short answer is that it’s not feasible for our small team, nor would it make any sense as with each patch we would have to retest it all again.
This is the reason we decided to try and use a crowd-sourced approach to our ranking data and also get the community involved, balancing both the quality and quantity of the data.
We could go much deeper into this and if you would like more information please leave a comment below and any other questions we can look at answering etc.
Note: We’ve taken a look at feedback since our initial release, and have now updated the data tables to include the AP range, and the possible loot range within each sample to give more transparency.