Total Kynosarges page hits as determined by Google Analytics shrank considerably year-over-year, going from 174,787 in 2015 (weblog 83,766, website 91,021) to 125,757 in 2016 (weblog 56,993, website 68,764). I’ll discuss likely reasons below. But first, here are the annual Top Ten pages, starting with weblog posts. I’m listing the original publishing date and total number of non-unique views for each page.
- DPI Settings in Windows 10 (September 2015, 10,309 views) was by far the most popular post. This is essentially a new addendum to the general advisory on the static website, split off as a blog post because the main article got too long.
- JavaFX DPI Scaling (August 2013, 4,435 views) lost 1,100 views, which was expected since the issues analyzed here have been fixed for a while.
- Java Method Reference Evaluation (October 2016, 4,017 views) seems to have gotten most of its hits because the original version contained a silly error, thankfully discovered by Hacker News readers…
- JavaFX DPI Scaling: Fixed! (June 2015, 3,976 views) covers the widely adopted Java SE 8 and accordingly gained 2,100 views while the earlier post declined (see above).
- WebView, the other JavaFX UI (December 2013, 3,345 views) lost another 2,500 views as continuing improvements to JavaFX make the HTML5 “nuclear option” less attractive.
- .NET Core: Hype vs Reality (November 2014, 2,634 views) lost another 600 views. This makes sense – it was only intended as a contemporary news commentary, and I no longer keep it up-to-date.
- No .NET 4.5 for XP/2003 (August 2012, 2,389 views) lost another 1,800 views as the XP/2003 era slowly keeps receding in the rear view mirror.
- JavaFX ListView Sizing (November 2013, 1,904 views) actually gained 50 views. This particular JavaFX API seems to remain problematic for developers.
- Simulating Platform.runAndWait (May 2014, 1,558 views) lost 600 views but still managed to return into the Top Ten.
- Class Diagrammer for Java (April 2016, 1,149 views) was an open-source project announcement that made it into the Top Ten, mostly thanks to the FX Experience aggregator.
Only two 2016 posts appear in the year’s Top Ten, but another four followed close behind (places 11, 14, 15, 19) and may well ascend next year. Notably, all of them were related to JavaFX. Aside from Java Method Reference Evaluation, the top non-JavaFX post in 2016 was the book review Core Java 10 for Java 8 which got all of 160 views. (And both of these posts were still related to Java!)
Next up, we have the Top Ten hits from my static website where I host my own open-source projects, as well as articles I expect to remain relevant for a long time and get a lot of traffic.
- Rise of Nations on Windows 8 (February 2013, 16,098 views) finally trended downward, losing a massive 11,800 views. Presumably, most people who want to run the original release on modern Windows have by now either succeeded or given up.
- Java for C# Programmers (August 2013, 11,931 views) remained fairly stable, losing only 500 views. I expect it to keep doing well as long as both C# and Java remain popular languages.
- High DPI Settings in Windows (May 2011, 10,471 views) lost another 3,000 views, although this was overcompensated by the rising popularity of the corresponding blog post (see above).
- DPI Scaling in Windows GUIs (May 2013, 6,737 views) lost 1,600 views. This may be natural decay as the article has been around for a while in its essentially final form.
- WPF Drawing Performance (June 2011, 4,945 views) lost 5,300 views, over half its total. I suspect the few remaining WPF developers just don’t care about drawing primitives…
- Check .NET Version with Inno Setup (September 2012, 4,515 views) lost another 700 views, likely mirroring the general decline of .NET for desktop clients.
- Class Diagrammer for Java (September 2010, 1,917 views) gained 1,850 views and returned from .NET oblivion with its Java/FX port in April 2016. A fine example for why I’ve left the .NET open source desert for the fertile Java oasis.
- Hexkit Strategy Game System (March 2000, 1,684 views) lost another 50 views but remains remarkably stable for an old frozen project.
- .NET Struct Performance (June 2011, 1,289 views) lost 1,200 views after last year’s surprise gain. This reverse is not surprising – it’s another frozen article with many caveats.
- LaTeX Typesetting with MiKTeX (June 2012, 641 views) lost another 600 views. I guess every remaining LaTeX user has already seen it by now.
Comparison to 2013–2015
While total page views in 2016 (126k) were quite a bit lower than in either 2015 (175k) or 2014 (174k), they are still higher than in 2013 (estimated 95k). Was there anything special about 2014 and 2015? It turns out there was.
- On the weblog, I wrote about several “hot topics” in 2014 that found an eager audience. This included one Hacker News hit (Light Field Photography, 12k views) and a number of widely debated JavaFX and .NET subjects. Some of these posts were written earlier but really gained traction in 2014. Then in 2015, Programming Languages in 2014 got an unparalleled 25.5k views thanks to Steven Sinofsky’s promotion, offsetting the lower popularity of older posts.
- On the website, Rise of Nations on Windows 8 (published in 2013) and High DPI Settings in Windows (updated for Windows 8/8.1 in 2013) both peaked in 2014–15, reaching total combined views of 38k and 41k respectively, now down to 27k. Several other older articles likewise peaked in 2014, had mixed results in 2015, and declined across the board in 2016.
Over the last year I had mostly blogged about programming tips (useful but uncontroversial), book recommendations, and other subjects that don’t generate huge traffic. On the static website I was busy moving my .NET projects to Java. I do expect this to increase their popularity in the long run, as evidenced by Class Diagrammer for Java, but for now much material is becoming less relevant or finding a new audience, with accordingly lower page hits. For those articles that are still reasonably current, impressions are in fact fairly stable compared to previous years. In the long run Kynosarges could well stabilize around 100k annual views if I don’t manage to create another outrage or two.
It’s really quite unlikely that 900k views are generated by human users who block tracking scripts when Google Analytics counts only 126k views on the same site. And indeed, the WLE referrer reports show much lower numbers: only 97,593 views claim to come from known search engines, plus 46,059 views from another referring URL – and both of those include some amount of referrer spam, or other bots with fake referrers. So unless some large number of human visitors carefully hides referrers in addition to blocking tracking scripts, the Google Analytics figures still seem fairly accurate.
Random Fun Statistics
Google Search continues to provide 50-60% of all my traffic. There’s also a smattering from Bing, and the rest mostly comes from Twitter, developer forums and news aggregators, my own cross-linking, or undeclared sources. The four most popular declared sources each reach just about a tenth of Google’s hits. This is actually an improvement over last year!
As usual, about a quarter of my visitors were from the USA. Most of the rest were from Europe including Russia or the Anglosphere. Notable exceptions are India with over 4,800 visitors (–2,200) and Brazil with 2,050 (–1,450), declining in line with overall views. China curiously grew slightly to 1,600 visitors (+110), putting it ahead of Indonesia and South Korea.
Regarding visitors’ operating systems, the website is completely dominated by Windows (80.5%), followed at great distance by Android (6.5%), Macintosh (5.5%), iOS (3.8%), and Linux (2.5%). These shares are largely unchanged from 2015. Weblog visitors are more varied but noticeably less so than in 2015, with Windows (70.5% from 56.5%) gaining greater dominance there as well, at the expense of Macintosh (10.2% from 13.3%), iOS (5.4% from 12.1%), Android (7.1% from 10%), and Linux (6% from 7.2%). All other platforms are sub-1% on both website and weblog, including Windows Phone and Chrome OS.
The clearly dominant browser remains Google Chrome (60%), followed by Firefox (18%, slightly down from 20%). With the increased share of Windows on the weblog, both website and weblog now get about 12% traffic from Internet Explorer and Edge combined, and nearly half those hits already come from Edge. This puts Microsoft safely ahead of Safari (desktop and mobile) which remains more popular on the weblog (8.2% from 15.9%) than on the website (5.1%, slightly up from 4.5%).