05-25-2024, 03:29 PM
(This post was last modified: 06-03-2024, 09:01 PM by wetwilleh. Edited 1 time in total.)
What’s the point of media anyway?
What’s the point of incentives?
What sort of incentives does the league use?
What’s my personal experience?
Enter science
So I collected some data very casually, a random set of posts from 3 or so weeks ago to see what impact the incentives might have had on the media and how that media was engaged with. 39 posts from 4/29 to 5/7. 39 not 40 because someone made a post while I was manually doing the word counting, and 39 not 78 because I got bored about half way through (it was going to be 78 because in the full span of posts I originally grabbed but didn’t do word counts on one post got deleted) and decided I had enough to make a pretty regression plot and pretty regression plots are science.
First lets see if people are responding to the incentives. Here's a plot of the frequency by word count for the 39 selected posts:
Cool. We do see a little blip with 3 people braving the dizzying length of 2500+. One person fell short in the desolate 2000-2500 wasteland, just short at 2467 by my count, though that one post was one of the best performing posts in the window. Coincidence? Probably. How about the three 2500+ posts 2 cleared that mark by a far margin, they weren’t trying to squeak in by a beak. And these posts all did quite from a view “perspective”, but not as well as the one immediately below it, and only a hair better than the 1500 to 2000 bin. Here’s a table:
Time for the money shot:
I regressed views on word count and we see the same story. Our one brave martyr falling just short of the 2500 line had a great view count, the three that soldiered on had solid view counts. The bulk of our evidence comes from medias articles with between 300 and 1500 words and in that range a lot of the juice comes from a few high fliers, but in general the trend holds more words more views.
So if its not more words what’s driving these views? Maybe it has something to do with content. And don’t forget the other incentive lever the league can pull. One of these posts even gives up the trick, aptly labeled “2X”. Mock draft media was bonused at the time. And in the data is more popular:
So no easy way to see if this created incremental posts on the topic, but we do see that those media articles discussing the draft were more viewed. It’s a great way to drop some @ symbols and that certainly helps (future analysis might look at the correlation between @ symbol usage and views/comments), but I think its something people are interested in interacting with whether they’re in the draft clad or looking to get a sense of who is coming into the league.
Let’s get that variable in our regression! If one variable regression is good multiple regression is even better
Heck yeah, get that R^2 up there. Our univariate view had around 16.7% variation explained, now we’re in the 37% neighborhood. Adjusted value took a little hit since in this view I included an additional variable ‘subtitle’ that is not very predictive. But the regression output shows that both word count and “draft” titles come in significantly. If you're trying to predict your views at home, you get ~86 views just for posting something, an additional 71.8! if your title contains the word draft. And then each of your carefully chosen words would be expected to add ~0.04 additional views, or 4 additional views per 100 words. Isn't science great.
Let's take a look at the correlation matrix to see how the variables correlate to one another:
Word count and “draft” titles significantly correlated to views, subtitle a nothingburger, and thankfully word count and “draft” titles aren’t highly correlated to one another a good candidate for inclusion in a multiple regression model.
What have we learned
Other stuff I didn’t mention yet
So what now?
- Well it’s a way to earn money
- It helps keep people engaged with the league
- And can create a deeper sense of community
What’s the point of incentives?
- Ideally it would create more media that meets the above goals
What sort of incentives does the league use?
- As a newer player I’m not the best to give an exhaustive history but the ones I’m familiar with are different flavors of “2X bonuseseses” and the 2500 word threshold for an extra bonus.
What’s my personal experience?
- There have a been a few times where I’ve been writing and wondered if I should stretch to hit the 2500 word milestone. In doing so I might have made the article less fun or coherent to read, but as a slave to the sim dollar I soldiered on in accordance with media goal #1 make that money.
- As a consumer of media I prefer someone who makes an interesting point more succinctly. If someone really invested a lot of time and research and or has an interesting perspective that’s what I personally am most interested in, but if they can make a point effectively in 600 or 1200 words all the better.
Enter science
So I collected some data very casually, a random set of posts from 3 or so weeks ago to see what impact the incentives might have had on the media and how that media was engaged with. 39 posts from 4/29 to 5/7. 39 not 40 because someone made a post while I was manually doing the word counting, and 39 not 78 because I got bored about half way through (it was going to be 78 because in the full span of posts I originally grabbed but didn’t do word counts on one post got deleted) and decided I had enough to make a pretty regression plot and pretty regression plots are science.
First lets see if people are responding to the incentives. Here's a plot of the frequency by word count for the 39 selected posts:
Cool. We do see a little blip with 3 people braving the dizzying length of 2500+. One person fell short in the desolate 2000-2500 wasteland, just short at 2467 by my count, though that one post was one of the best performing posts in the window. Coincidence? Probably. How about the three 2500+ posts 2 cleared that mark by a far margin, they weren’t trying to squeak in by a beak. And these posts all did quite from a view “perspective”, but not as well as the one immediately below it, and only a hair better than the 1500 to 2000 bin. Here’s a table:
Time for the money shot:
I regressed views on word count and we see the same story. Our one brave martyr falling just short of the 2500 line had a great view count, the three that soldiered on had solid view counts. The bulk of our evidence comes from medias articles with between 300 and 1500 words and in that range a lot of the juice comes from a few high fliers, but in general the trend holds more words more views.
So if its not more words what’s driving these views? Maybe it has something to do with content. And don’t forget the other incentive lever the league can pull. One of these posts even gives up the trick, aptly labeled “2X”. Mock draft media was bonused at the time. And in the data is more popular:
So no easy way to see if this created incremental posts on the topic, but we do see that those media articles discussing the draft were more viewed. It’s a great way to drop some @ symbols and that certainly helps (future analysis might look at the correlation between @ symbol usage and views/comments), but I think its something people are interested in interacting with whether they’re in the draft clad or looking to get a sense of who is coming into the league.
Let’s get that variable in our regression! If one variable regression is good multiple regression is even better
Heck yeah, get that R^2 up there. Our univariate view had around 16.7% variation explained, now we’re in the 37% neighborhood. Adjusted value took a little hit since in this view I included an additional variable ‘subtitle’ that is not very predictive. But the regression output shows that both word count and “draft” titles come in significantly. If you're trying to predict your views at home, you get ~86 views just for posting something, an additional 71.8! if your title contains the word draft. And then each of your carefully chosen words would be expected to add ~0.04 additional views, or 4 additional views per 100 words. Isn't science great.
Let's take a look at the correlation matrix to see how the variables correlate to one another:
Word count and “draft” titles significantly correlated to views, subtitle a nothingburger, and thankfully word count and “draft” titles aren’t highly correlated to one another a good candidate for inclusion in a multiple regression model.
What have we learned
- Incentives work. We have actual science that shows 3 people out of the 39 I randomly chose wrote 2500 words, while only 1 poor soul chose to write 2000-2500 words.
- People love to talk draft. There’s probably some other content types that people prefer to consume and those might good candidates for bonuses, or maybe people would write them anyway (a possible area of future study, we would need a bigger data set but could avoid the tedious word count bit so long as the titles are indicative of content).
Other stuff I didn’t mention yet
- The third goal is creating a deeper sense of community and one thing I didn’t touch on yet is that I’ve seen a few instances where people have written collaborative posts and shared the spoils. This is one of the possible benefits of the 2500 hurdle. Getting people to team up in a group to turn what might have been 3 1000 word posts into 1 3000 word post. Teamwork makes the dream work.
So what now?
- I'm going to care less about hitting the 2500 threshold, in the grand scheme of things it’s a tiny drop.
- Maybe media 2X bonuses should be more frequently rotated. If you want people talking fantasy draft or playoff push or trash talking or historical analysis throw a bonus that way. If nothing else it’s nice for people wanting to generate media to get an idea on something to talk about.
- I'm going to try to include the word “draft” or something that looks like draft in all of my titles/subtitles from now on.