the background card art doesn't help convey any information
It does actually. It conveys costs. The alternative is to have each region be a different shade of grey, which necessitates the reader consulting a legend and trying to figure out which shade of gray is the $4, which is the $5, etc. Conventional wisdom these days is that you shouldn't have a legend in your graph unless there's no alternative. Legends are an inconvenience to the reader. This hypothetical legend (which isn't really hypothetical, because I tried it originally when designing the graph) especially would be horrible because we're asking readers to distinguish between 8 different shades of the same color.
Now, it can be argued that it would be best if each region only had
one card image instead of four, but that would have required stretching the card art to obnoxious proportions (or asking RGG/Donald X for a super high res image for 8 different cards that could be zoomed really far into, which (1) might not even exist, (2) I am probably not privy to even if I ask politely, because they don't know me from Adam, (3) they may not be legally allowed to disclose anyway depending on the terms of the contract between the publisher, designer, and artists, and (4) "You're asking us for this huge favor just so you can make a graph??"). The only other alternative is to make each region an unrelated color, such as having $5 = Orange, $4 = Green, etc. This is a bad move because it fails to convey how the region above it is exactly $1 more than the region below it. Also it's ugly and requires a legend. Don't underestimate the value of a visually appealing graph. People will stop looking if it's ugly, as it reminds them of math homework. A pretty visual keeps them looking long enough to digest the info.
The point is taken though. I might play with contrast to make the differences between each region pop more. You probably can't tell, but I already made it so that each region is slightly more transparent than the one below it, in attempt to make them more distinguished. All I can tell you is, if you saw the original side by side with this one, you'd know it worked. But your feedback that it's still hard to read is helpful. I'll see what I can do. But I'm not convinced that ditching background images is the right way to go.
In your post you say you want to compare the distributions of sets, but it looks like you are plotting the cumulative distribution of card costs over time
No, I didn't say that. The title of the post is "Cost Distributions over the Years", and in my post I said "it's a time axis, but with set icons instead of dates as labels", as well as, "with the x-axis being time and the y-axis being %". I think you may have just misread it. The reason the labels are set icons instead of dates is because, if the axis labels tells somebody that in October 2010 the $5 cards went from being 3.75% of the card pool to 5.71%, their next logical question is "what's the significance of October 2010?". The answer being "That's when Prosperity came out". So why not just cut out the middleman and put the set icons- the relevant information- in as the labels? It's less cluttered, more pleasing to the eye, and it answers the questions that people will naturally have anyway. This doesn't mean I'm trying to compare the distributions of each set though. One wouldn't even use an area graph for that, one would use a bar graph.
While this is also interesting, it will make it harder to notice significant differences of composition in later sets
Which is not the purpose of this graph at all. To use an analogy, that's like complaining that
this graph, while also being interesting, will make it harder to notice significant differences in how each individual city or county voted. Making a graph illustrating the cost distributions of each individual set
would be interesting, and I would be willing to do it if you and/or others are interested. But that's not this graph's intent.
I'd be tempted to group the sets into four or five time periods
I am not a fan of this idea. What value do we gain by washing out detail?
I'd also consider grouping the costs into three categories
Again, what value do we gain when you can just do this manually with your own eyes? The $3 and $4 regions are right next to each other on the graph. Each region is ordered by cost. So you can just add them together visually. By bucketing data, I
force all readers to bucket them, whereas if I leave all data unbucketed, those who want bucketing can bucket with their eyes (at least with the y-axis. You can't do it as well with the x-axis due to the bars being spaced out, which is why I'm making another version with promos bucketed for people like Donald X who don't care to see them separated out).
something like: low (2 and below), mid (3-4) and high (everything else)
Look at the graph though, the 3-4 region would be
massive. Besides, there's a non-trivial difference between $3 and $4 cards in Dominion. What do we gain by washing out detail?
I would also try to slot Potion and Debt costs into these bins in sensible ways
The only sensible way to consider Potions is by using Complex or "Imaginary" numbers (ie two-dimensional numbers). Doing that means making this graph three-dimensional, which is a no-go, especially if you're already concerned that the graph is hard to read as-is. Plus, the number of potion costing cards in the game is low enough that they would appear very small on a graph. If there were a sequel to Alchemy which added more potion cost cards, we could gain something from analyzing potion cost cards. But for now, there just aren't many of them, and that step into the 3rd dimension would cause more confusion than it's worth.
Author
Topic: Cost Distributions over the Years (Read 4078 times)