What Games People Are Talking About: A Pie Chart

You may have seen me talking about EN World's HOT GAMES TRACKER recently. It asks the question: What's the current zeitgeist? What are the hottest games being played right now? This isn't a list of sales figures; it tracks what's currently being talked about using a top secret algorithm. Each game is also conveniently linked to a search for discussion about it right here on EN World, should you want to find out more. The spotlight list changes from time to time. The red and green arrows show a game's general trend over the last 90 days - is it being discussed more or less than it was in the previous 90 days?


This page tracks discussion of over a quarter of a million forum members and approaching a thousand blogs on a selection of major independent RPG discussion forums to create an overall sample from a list including EN World, RPGnet, UK Roleplayers, RPG Geek, the RPG Bloggers network of nearly 300 blogs and the RPG Blog Alliance of nearly 600 blogs.

I've extracted some data from that page and turned it into a couple of nifty pie charts. I've presented them below. A couple of caveats:


  1. The pie charts aren't really the the thing. The raw data on the linked page is. If you disagree with the way the data is presented here, I encourage you to look at the actual numbers instead and derive your own conclusions.
  2. This is NOT sales data; it's also NOT what games folks are playing at home. It's exactly what it says it is: a large representative sample of the game folks are talking about online. So be careful what conclusions you extrapolate from that.
  3. The final D&D Next playtest packet was just released. This gave DDN a huge boost. I looked at this data this time last week (sadly, I didn't think to graph it) and Pathfinder was leading D&D Next by nearly 5%. I'll look again at it in a few weeks to see if D&D Next holds its current lead or drops back down to second place again.
  4. The two graphs were compiled a couple of days apart, so the figures changed slightly between them.
  5. I was asked yesterday why 13TH AGE was considered D&D in one graph, and not part of "Other D&D" in the second. That was simply because it was the largest item in "Other D&D" so I slipped it out separately as a clear visual point of comparison. It's not meant to imply the "D&D-ness" or "lack-of-D&D-ness" of it or anything else.
  6. It's ROLEPLAYING GAMES sites and blogs only. I'm sure if we were looking at tabletop wargaming sites, things like Warhammer 40K would be vastly more popular; as it is, that game refers here to the line of RPGs, not the wargame.
  7. This is not a qualitative judgement. Your favourite game is the best.

So, without further ado:

hotgames.jpg


hotgames2.jpg



 

log in or register to remove this ad

Would including all the major company forums be that difficult? Like, maybe as something that could be toggled on and off?

It would be next to impossible. It'd end up simply being a chart of companies with RSS feeds on their forums. Paizo.com has one; wizards.com doesn't. Pathfinder would shoot to the top, and D&D would almost disappear. That only works if you can do them all. And then, you're looking at trying to achieve something closer to a census than a representative sample, which isn't realistically attainable.

The aim is to basically get a statistically significant sample. The statisticians out there can comment more on what and why that is, but it's by far the most appropriate approach.

Now, bear in mind that I'm not a statistican. I'm just doing what I was told to do by an actual statistician. So my explanations may not be entirely coherent! But I've been told very strictly that including company websites would distort the data to beyond meaningless (unless I have the resources to do an actual complete census).

Now, we don't have current survey data indicating the size of the market; so that's a problem. I'm sure it exists, but I don't have access to it. The latest public data is from 1999.

Based on that, we have a confidence interval of within a couple of percent. (That means the 0% ones aren't "none"; they're 0% +/- x%.)

Because it seems like what you have now is going to skew (possibly quite heavily) your results toward out-of-print games and away from games that benefit from strong company website communities.

Fortunately, the data doesn't appear to have done. The top are D&D and Pathfinder, followed by things like 13th Age, Numenera, Edge of the Empire, new Shadowrun. New and current stuff very clearly dominates the chart.
 
Last edited by a moderator:

log in or register to remove this ad

And then, you're looking at trying to achieve something closer to a census than a representative sample, which isn't realistically attainable.

The aim is to basically get a statistically significant sample. The statisticians out there can comment more on what and why that is, but it's by far the most appropriate approach.
I took a fair amount of math in school and I'm well aware of the difference. My aim was to suggest a more representative sample, which also coincidentally would have been a lot bigger. But--of course--if it can't be done, then it can't be done.
 

I took a fair amount of math in school and I'm well aware of the difference. My aim was to suggest a more representative sample, which also coincidentally would have been a lot bigger. But--of course--if it can't be done, then it can't be done.

Sorry! I wasn't intending to try to teach my grandmother to suck eggs! I somehow fell into "repeat stuff I've been told" mode and why adding and enormous data source like, say, Paizo.com would make the data less representative, not more.

When you start adding in company forums too, you begin approaching more talking about attempting a census. Including everything. But no, that can't be done - I can add some but not all, which would break that data completely. At least, not with the sort of resources I have available.
 
Last edited by a moderator:

Sorry! I wasn't intending to try to teach my grandmother to suck eggs! I somehow fell into "repeat stuff I've been told" mode and why adding and enormous data source like, say, Paizo.com would make the data less representative, not more.

When you start adding in company forums too, you begin approaching more talking about attempting a census. Including everything. But no, that can't be done - I can add some but not all, which would break that data completely. At least, not with the sort of resources I have available.
No offense taken... I'm sure you've had more than your share of criticism for this survey, and I wasn't intending to join in that.

It makes sense that if you can't add WotC then you can't add anybody.

It also makes sense that if you did add company websites then you would need to increase the non-company site portion of the survey to census levels in order to account for that. (Though in that case I would imagine that your best course of action would be to simply present two sets of data: one that included company site data and one that excluded it. That was where I was going with the toggle suggestion.)
 

No offense taken... I'm sure you've had more than your share of criticism for this survey, and I wasn't intending to join in that.

It makes sense that if you can't add WotC then you can't add anybody.

It also makes sense that if you did add company websites then you would need to increase the non-company site portion of the survey to census levels in order to account for that. (Though in that case I would imagine that your best course of action would be to simply present two sets of data: one that included company site data and one that excluded it. That was where I was going with the toggle suggestion.)

It's not just increasing non-company site portions, by going to company websites, you're adding a lot of bias to the sample. Naturally, a preponderance of discussion on sites like WotC and Paizo will be on their own products, making them particularly biased sources of data. Sites with a particular focus in their history, like ENWorld with its original D&D focus, already add bias to the sample despite being open to discussion of any games because the historical focus generated a sample of gamers with a systematic bias. A company website would be like injecting that sort of sample bias on steroids.
 

Small nitpick that someone else may have mentioned: It probably doesn't make much (if any) difference in the results, but you seem to be counting 13th Age in both the "D&D and Variants" and the "non-D&D" groups.
 

This seems like it would be a great source of hundreds of discussion points for a first course in statistics!

I think I would actually just avoid talking about the kind of sample you have - you have a non-random convenience sample of some not-well defined larger population. If that population is supposed to be all gamers who post on line then you are suffering from a huge source of bias due to undercoverage because your sampling frame is systematically missing all of the company web-sites. Its even farther away from all gamers. Is it better to just say you have a census of "The most popular independent RPG sites and Blogs"?

As far as what you're measuring. You are only getting the count of "people who post about games" if your algorithm only counts the number of distinct posters over the 90 days once each. Further you would need to assume that most people don't have at least two boards they use regularly. You are even farther from getting the number of "people who play each game" (which one of the charts has as a label) since a lot of us seemingly post in threads for systems that aren't related to what we are currently playing. Going with "Most discussed RPGs" like you say in several places seems appropriately nebulous enough that it isn't misleading.

And then once you have your sample and a rough research question you get all the arguments about what data is most appropriate. If it's threads, is a one post thread as good as a 1,000 post one? If it's number of posts, what happens when a few 3.5ers who are on-line at the same time manage to crank out 20 pages of posts in a day and a half arguing about wizards versus fighters? If there are lots of PF on-line games and OD&Ders like to play in person should all on-line games be excluded, or are you just going for total on-line traffic about the games? If its threads, how does "All D&D count"? If its post, how does a post talking about 4e as a counter example in a thread about a PF question count? What happens if I don't like the auto-linking ENworld does when I type Pathfinder so that I use PF instead? For posts, how does posting they wish they could give XP count versus actually gettting XP? etc...

Hmmmm.... I'm getting more and more glad it isn't my project to run and that I just get to see the cool results. (Which I really enjoy, thanks!).
 
Last edited by a moderator:

Small nitpick that someone else may have mentioned: It probably doesn't make much (if any) difference in the results, but you seem to be counting 13th Age in both the "D&D and Variants" and the "non-D&D" groups.

The two lists are completely independent of each other. Anything in one list doesn't affect the other. Numenera and Star Wars (on the Hot Games page) are on the Sci-fi chart additionally.

Unless you mean the pie charts specifically. All they are are the tables on the Hot Games page pasted into Excel, and me hitting the "chart" button. On those, I slipped 13th Age out separately too because it was the largest in that group, and it gave a good visual key as to size.
 
Last edited by a moderator:

Is it better to just say you have a census of "The most popular independent RPG sites and Blogs"?

Indeed! That's almost exactly what the Hot Games page says! "This page tracks discussion of over a quarter of a million forum members and approaching a thousand blogs on a selection of major independent RPG discussion forums to create an overall sample from a list including EN World, RPGnet, UK Roleplayers, RPG Geek, the RPG Bloggers network of nearly 300 blogs, the RPG Blog Alliance of nearly 600 blogs, and Reddit." Though I use the word sample, not census. Extrapolating sales data - or anything else - from that would not be an appropriate thing to do.

http://www.enworld.org/forum/hotgames.php

As far as what you're measuring. You are only getting the count of "people who post about games" if your algorithm only counts the number of distinct posters over the 90 days once each. Further you would need to assume that most people don't have at least two boards they use regularly. You are even farther from getting the number of "people who play each game" (which one of the charts has as a label) since a lot of us seemingly post in threads for systems that aren't related to what we are currently playing. Going with "Most discussed RPGs" like you say in several places seems appropriately nebulous enough that it isn't misleading.

Yup! Discussions. That's exactly what the page says! :)
 
Last edited by a moderator:

Indeed! That's almost exactly what the Hot Games page says! "This page tracks discussion".

Sorry, I should have made it clear I was aiming at your post #21 ("statistically significant sample" and whether there's any need to bother with confidence intervals) and not at the actual HG page. You can't get much more careful in a description than saying "top secret algorithm"! :)
 

Remove ads

Remove ads

Top