# Holiday Data-palooza: Site statistics from 2017

Other Person: Hey, King, remember that one time you said you learned all that stuff about the Stack Exchange Data Explorer?

kingledion: ...yeah...

Other Person: Well I have some questions about how this site got used in 2017...

kingledion: Like....what kind of questions?

• What users generated the most responses?

• What users had the highest score per post?

• What users had the longest posts?

• (Michael Kjorling) What users make the most effective use of their words?

• (Green) Which posts are most viewed?

• (James) Guys, am I awesome? Please?

• (Separatrix) Who gets questions answered at the highest rate?

• (Green) What is the relationship between views and votes?

• (Separatrix) What were the top voted questions and answers of the year?

## Finit

As of 1200 Eastern time on Monday 18 Dec, I'm calling the Data-palooza off. See you again this time next year!

• Net post score divided by post length. Let's see which users were able to get their points across well in few characters! For this metric, a higher score would be better; it means you either got more votes (x/y is larger when x is larger), or you spent less characters to get the same number of votes (x/y is larger when y is smaller). I know that net post score isn't a perfect measure of post quality, and that total post length isn't a perfect measure of how easy the post is to read, but they tend to correlate fairly well. This might need to be grouped by approximate post length.
– user
Dec 14 '17 at 15:53
• I suppose "How awesome was James in 2017" is too specific? :) Dec 14 '17 at 16:28
• Are these statistics for 2017 only, or for the lifetime of the site? Could you provide a site wide average to compare against? Dec 14 '17 at 17:06
• Could we also get averages for occasional users, and frequent users? Dec 14 '17 at 18:22
• Would it be possible to get a link to the queries themselves? Dec 14 '17 at 18:31
• @James How on $PLANET are we supposed to quantify that? How many awesomepoints is a diamond worth, for example? – user Dec 14 '17 at 18:43 • @kingledion that's fair. I haven't worked with the query editor in a while. Thanks for the data you've been able to share! Dec 14 '17 at 18:59 • @Green I'll just put it here: this is the base query for most of the things I've done on this page data.stackexchange.com/worldbuilding/query/770555/… Dec 14 '17 at 19:05 • Can you do something on the best closed-unclosed ratio, the worst closed-unclosed ratio (for users with a reasonable number of questions, say at least ten), and perhaps something on the most posts? Dec 15 '17 at 1:05 • Read through all this, now am sad. My 2017 contribution has been lacking. Dec 15 '17 at 13:45 • The last thing I can think of is highest voted questions and answers of the year Dec 18 '17 at 8:27 • @Separatrix The highest views is in the last answer at the bottom of the page. Dec 18 '17 at 12:15 • I was looking at those and it seems that views and votes are less closely tied than I initially expected. I'm really looking for the one off medal winner here rather than average over all questions Dec 18 '17 at 12:19 • @Separatrix Done and done. And that wraps it up! Dec 18 '17 at 16:55 ## 7 Answers # Which users are the most awesome in units of James We have a few possible definitions of awesome that we can investigate ### Q & A: Users who post good questions and answers (1 James = 302) Sum of all upvotes 1. Will: 19.6 James' 1. L.Dutch: 12.5 James' 1. Separatrix: 10.7 James' 1. kingledion: 9.6 James' 1. Cort Ammon: 8.8 James' 1. James: 1 James ### Site Maintenance: Users who edit and review a lot (1 James = 114) Sum of all edits plus review tasks completed where the reviewer left a comment 1. Secespitus: 9.8 James' 1. JDLugosz: 4.5 James' 1. a4android: 4.4 James' 1. Brythan: 3.9 James' 1. L.Dutch: 3.4 James' 1. James: 1 James ### Number of moderator elections won (1 James = 1) • T-1. James: 1 James • T-1. JDLugosz: 1 James • T-1. Michael Kjorling: 1 James • T-2. Everyone else: 0 James' Thank god for ties being broken alphabetically, or James would never be on top :) • I am feeling sadly left out of all of this. Surely there's something I do that can be counted in units of Jamesness? :-) Dec 14 '17 at 20:23 • @MonicaCellio In editing/reviewing you are at 0.3 James'; but on the bright side, in moderator elections won before 2017 you are at$\infty\$ James'! Data science, where everyone can be special! Dec 14 '17 at 20:26
• I lament that I have but one upvote to give. Dec 14 '17 at 20:38
• I can't even qualify against that....oaf? Dec 14 '17 at 20:52
• @dot_Sp0T When you get hit by lightning you shall know it was ME! Bwahahaha (I hope you don't get hit by lightning, I will fee bad.) Dec 14 '17 at 20:57

# What is the relationship between views and upvotes?

### For a question

X-axis is views, y-axis is net votes.

The linear relationship in the graph is actually kind of amazing. See that outlier to the right? If you take the outlier out, the slope changes by less than 0.03%. That is pretty unusual.

Overall, you get one upvote (on the question) per every 250 views.

Here instead of plotting question score on the y-axis, I use the sum of the answer score.

Still a linear relationship, but with more heteroscedasticity; that is, variance increases more strongly with number of views for answers than it does for the questions.

Also, you can see our outlier question doesn't get nearly the answer votes that it could.

• So on average, you get 1 upvote for every 250 page views (which is cool). Do you have a way of knowing if those votes are clustered in the first day or two after the question was asked? My experience is that upvotes are not received in a linear fashion, but happen in clumps. Dec 18 '17 at 17:49
• @Green Upvotes have their timestamps stripped in public data, in order to keep SE's policy of no vote attribution intact. Otherwise you could potentially use a machine learning algorithm to match votes to votes. In any case, we don't have the data to track votes over time. Dec 18 '17 at 18:34
• And all data is "as of three days" ago? Conceivably, we could track individual questions and upvotes by sampling at daily intervals. Not useful for questions that get only a handful of upvotes but may be useful for the tens or hundreds of upvotes questions. Hmmm.... Dec 18 '17 at 18:55
• @Green Nah, the data dumps are once a week. At least so it appears, for two weeks in a row the data was published about 3am on Sunday morning. So I suppose we could get them at weekly intervals if we really wanted. Dec 18 '17 at 19:03
• That's probably not enough resolution to answer my question. Most questions stop getting attention within the first couple of days. Two weeks to get a single sample isn't good enough. Ah, well. It was a fun thought. Dec 18 '17 at 19:18
• It's very tight to the line at the low end, then suddenly breaks down somewhere past the 50 vote/10,000 view mark Dec 19 '17 at 12:28

# What users generated the most discussion?

I summed the total number of comments and answers and divided by the number of posts. Obviously, comments can get cleaned up by the mods, so perhaps not a representative sample.

### All users (sitewide average: 3.5)

• Lysero: 65
• XenoDwarf: 53
• JNW: 47
• Sebastian Araneda: 47
• Myself: 44

### Occasional users (>10 posts in 2017)

• Gstestso: 16.2
• OneSurvivor: 16.0
• Terry: 16.0
• Shard martin: 14.5
• Unhappymarshmellow: 14.3

### Frequent users (>100 posts in 2017)

• Vylix: 7.1
• Sum of all users whose accounts have been deleted: 6.2
• Separatrix: 4.5
• kingledion: 3.8
• MichaelK: 3.6

# What users had the highest score per post?

Summed upvotes divided by number of posts.

### All users (sitewide average: 5.6)

• Reid Rankin: 138
• bjmc: 109
• jncraton: 107
• rluks: 82
• frodoskywalker: 82

### Occasional users (>10 posts in 2017)

• Pavel Janicek: 27.3
• Thriggle: 21.9
• T. Sar: 20.5
• Joe Bloggs: 19.5
• AndreiROM: 17.6

### Frequent users (>100 posts in 2017)

• Separatrix: 15.6
• MichaelK: 12.5
• kingledion: 11.6
• sphennings: 11.2
• AlexP: 9.9

NOTE: It is interesting that number of votes and number of comments tracks each other closely

# What users liked to type the most (characters per post)?

Summed number of characters in the 'Body' field divided by number of posts. Only questions and answers, not comments.

### All users (sitewide average: 1523)

• brendan: 14989
• Chris S: 14120
• Rob Stening: 13459
• Akmedrah: 12650
• qGold: 12071

### Occasional users (>10 posts in 2017)

• Flater: 6340
• Schwern: 5445
• dsollen: 5343
• Theraot: 5267
• MolbOrg: 4422

### Frequent users (>100 posts in 2017)

• AlexP: 2852
• LSerni: 2701
• JBH: 2692
• Erin Thursby: 2419
• In the score per post section can you eliminate anyone with fewer than say 5 or 10 posts? And average of 138 leads me to think that is small sampling anomaly. Dec 14 '17 at 17:10
• @James Yeah, occasional users are 10+ posts, frequent users are 100+ posts Dec 14 '17 at 17:13

# Users who made the best use of their words?

I summed the number of net votes and divided by characters, then multiplied by 100. The result is net votes per 100 characters of question or answer.

# Who is generating the most pageviews?

### Most viewed single question

Views stopped counting on 12-10 at about 2 am. Sorry to Green's T.Rex which will probably make this list by tomorrow.

The number of accepted answers divided by total answers; for questions posted in 2017, not answers. Query here.