The Pitfalls of D&D Beyond Data

[ClaytonCross]

I think this might help.

In probability, all the probabilities must add up to 100%. So no matter how I want to delimit a particular population, if that population contains members that don't fit the normal categories associated with the chosen delimiter then the solution is to add in a category "not X" so that these members of the population have a place to get accounted for in our categorical breakdown (otherwise the probabilities wouldn't add to 100%). In probability the population determines whether the "not X" category needs added. In probability you would use the same kind of title for a graph where you did not have to add in the "not X" category and for a graph where it needed to be added in. In probability you always are required to clearly specify your population.

So from my perspective which leans heavily on mathematical probability understanding.
If their population of active characters only included characters with subclasses then I think we both agree that the proper graph title would be "Subclass Distribution (Active Characters)". In this context a graph of a population of active characters, some without subclasses, would then also be titled the same way because it's still a graph of the same thing no matter whether some members of the population don't have subclasses. The only difference is that the first graph wouldn't have a "no subclass" category because it wouldn't be needed but the later graph would have a "no subclass" category because it would be needed.

Hopefully that helps clarify where I'm coming from.

I understand what you are saying. What I am saying is that the delimiter of subclass makes the population only representative of characters subclass they will reach 100% of that defined population. This throws off your comparisons which is (I am guess) the point of your distress. That doesn't make them flatly wrong. They chose my method over yours. However, I do think my suggested option would work better for both of us and solve that problem.

 

[FrogReaver]

Ok, then progress was made to a common point but we are still adjacent not aligned. As I under stand this.

"Subclass Distribution (Active Characters)" for you means "Subclass Distribution (active characters with or without subclass)" to exclude you "Subclass Distribution (active characters with a subclass only)" - I added "only" for as much standardization and to be as specific as possible in attempt to avoid confusion. I know that's not exactly what you said.

"Subclass Distribution (Active Characters)" for you means "Subclass Distribution (active characters with subclass only)" to include you "Class Distribution broken down by subclass when applicable (Active Characters)"

We can both find both answers... however, my belief that I am correct is based on the point of the slide as written with no implications is "Subclass Distribution". From my prospective, if you don't have a subclass your just talking about a class and implying a greater scale on a slide shown to general audience is counter intuitive. It is my belief that since they have a class slide and they don't list classes without subclass on the subclass slide its it clear to me that D&D Beyond is using my method. I understand your saying they could have use your method, but having a class slide and a subclass slide with classes on it seems redundant. Could they have done it? Absolutely, but it seems incorrect to me by causing the problems your complaining about. They said they are not using your method and kept it to a "it means what it says" without implications train of thought for mass audiences. The assertion of the thread is that it is incorrectly presented, however since we are both able to demonstrate our method works then they could not "win" someone was going to say it was wrong. People looking from your side and people looking from mine. But saying it means exactly what it says no more no less provides them with the ability to reduce confusion as easily as possible. Making this method the better of two correct answers. The better answer being what I am calling correct in this case, even though your method is functional.

I do think better than both would be the mentioned: 1 "Class Distribution(Active Characters)" and 12 "X class Distribution by subclass when applicable (Active Characters)" with Total # / Non-Multiclass # / Multi-class Number # under the name and percent. I think looking at these slides requires not inference or implications and due to the high number of low level characters the Class without the subclass would be clearly visible.

I don't agree with the large middle portion obviously, but this whole post is well stated. You summarized my position very well. However, I have enough of a background in probability and statistics to know that the category "not X" gets added to any categorization summary of a population simply when that population has some members that don't fit into any of the categories. I have enough background to know that adding in that category "not X" never changes the title that would be used or anything else about how the data is presented.

So arguing that you can't or shouldn't add "not X" into a probability graph seems very uninformed to me. It's required that students do just that every day in probability and statistics courses across America. In fact one of the biggest "gotcha" type homework/test questions in probability is to give you a collection of for example 10 balls of various colors and 1 book and 1 pencil and ask you to graph the probability distribution of colored balls for that collection. You will get counted off if you don't make a category "not a colored ball" and include the pencil and book in that category.

 

[FrogReaver]

I understand what you are saying. What I am saying is that the delimiter of subclass makes the population only representative of characters subclass they will reach 100% of that defined population. This throws off your comparisons which is (I am guess) the point of your distress. That doesn't make them flatly wrong. They chose my method over yours. However, I do think my suggested option would work better for both of us and solve that problem.

No. My issue is that their defined population was "(active characters)". you dispute that. It's fine. The post you are citing was to inform you why I think that. But thanks for not interacting with any of my reasons...

 

[ClaytonCross]

I don't agree with the large middle portion obviously, but this whole post is well stated. You summarized my position very well. However, I have enough of a background in probability and statistics to know that the category "not X" gets added to any categorization summary of a population simply when that population has some members that don't fit into any of the categories. I have enough background to know that adding in that category "not X" never changes the title that would be used or anything else about how the data is presented.

So arguing that you can't or shouldn't add "not X" into a probability graph seems very uninformed to me. It's required that students do just that every day in probability and statistics courses across America. In fact one of the biggest "gotcha" type homework/test questions in probability is to give you a collection of for example 10 balls of various colors and 1 book and 1 pencil and ask you to graph the probability distribution of colored balls for that collection. You will get counted off if you don't make a category "not a colored ball" and include the pencil and book in that category.

So I am willing to allow your "I'm experienced" attempt of "authority on the subject" if you can recognize that while that might flavor your view, its entirely possible that is not the right approach when providing information to the general populace and not students/teachers of probability and statistics. Your saying your way is the "studied correct way" while I am arguing that for the general public they are providing information based on not requiring interpreting subtext.

Do you disagree that the average person would look at a chart titled "Ball Color Distribution" see the colored ball counts and be fine, then see the "not a colored balls" and think what the hell? Why put that there? Isn't that what makes it a "gatcha"?

Your prospective is from that of someone claiming to be presenting to trained professionals. D&D Beyond is not presenting to trained professional but instead the untrained masses. As I said before, expecting intuitive leaps is a lot harder to defend against the untrained masses than "It means what it says". Particularly in a community that widely understands RAW vs RAI. The RAW carries more weight. This isn't a "your wrong" statement, its a "even if your right their is very good reason why you with a background in this are the largest and strongest opposition to this". Basically its your stated expertise that makes you not a valid demographic target for explaining to the masses. It then makes since why the method described is wrong to you and yet D&D Beyond went with the direction that the most people would understand at face value from a "business customer service perspective" instead of a "probability and statistics expert perspective".

So I can see now why you say its wrong, progress has been made. I do however disagree that D&D Beyond should be forced to those standards when its reasonable based on their target community aka customer base, to go with a "RAW" approach which they are more likely to understand and widely excepted. More so because it has been proving that to in fact be the case. I mean their have been ALOT more people posting against your view than for it and it took us a while for you to explain it to a point where I can follow your reasoning and even then, I understand but I don't agree.

 

[ClaytonCross]

No. My issue is that their defined population was "(active characters)". you dispute that. It's fine. The post you are citing was to inform you why I think that. But thanks for not interacting with any of my reasons...

I understood that YOU define it that way. My argument has and is that THEY don't. They even clarified that to be the case. BadEye spelled it all out for you very clearly. So I get why your saying that. I do. But that is not what they are doing. They are breaking the slide by subclasses but only polled active characters (with subclasses). That is a fact of the data they pulled and they have confirmed that. You saying that was not the "correct method" I respond to that in my last post so I will not try to separate what that into this post too.