Talk:Usage share of web browsers

What about stats from w3schools.com

Hello, I am wondering why you do not include statistics form "w3schools.com". According to Alexa it has a much higher traffic ranking (presently #2000 worldwide) than any of the other services mentioned here. (see their statistics page) There must be a reason. :-) Best regards --Marbot (talk) 10:31, 8 December 2008 (UTC)

Further down on their results page is the following:

• W3Schools is a website for people with an interest for web technologies. These people are more interested in using alternative browsers than the average user. The average user tends to use Internet Explorer, since it comes preinstalled with Windows. Most do not seek out other browsers. These facts indicate that the browser figures above are not 100% realistic. Other web sites have statistics showing that Internet Explorer is used by at least 80% of the users.

So, W3Schools' data doesn't seem to be as accurate as the other websites we have listed here. For example, IE and Firefox have about 50% each. Xenon54 11:08, 8 December 2008 (UTC)

Fair enough. You are right. However, I find it amazing, that more progressive users may generate enough traffic for a page to make it rank at about 2000. --Marbot (talk) 13:12, 8 December 2008 (UTC)

There is no relation between the traffic to the site reporting the stats, and the total traffic measured by the service the reports are based on. A single site's usage (w3schools.com) is not a representative sample of browser usage share, where the other sites are reporting on the usage share measured at tens of thousands of different websites.

BE CAREFUL! the data that this article show differs from the w3schools.com one. It is imperative to fix it.

While the disclaimers concerning W3Schools are very valid, I note with interest that their numbers match those for my own website much better than those currently given in the article, in particular by having Firefox clearly ahead of Explorer instead of the other way around. (And, no, unlike W3schools, I do not serve a niche of web experts.) We should bear the risk in mind that the systematic errors mentioned in the article may over-favour Explorer.94.220.255.11 (talk) 22:44, 19 November 2009 (UTC)

Hourly data

A few weeks ago I added a link to http://safalra.com/website/web-browser-market-share/ after it was on ComputerWorld because I though it was useful, but I lost the bokmark and came back here to find it and it was gone. The history says Wikiolap deleted it because he said it „provides information how to break into Net Application system without paying“ but that's not right. It doesn't say how to hack into the system, and you can't get their other data you have to pay for like data for US states, it just says how to change the options for their public graphs so they show diffrent browsers, so I dom't see what's wrong with it. --82.33.205.102 (talk) 17:18, 15 December 2008 (UTC)

I've just received a message through my website about this (I created the referenced page), presumably from the anonymous editor above (although it came from a different IP address, and they seem to be on a fixed IP). Actually, I'd have to disagree with one of their minor points — I don't think it's hugely useful, as I don't think anyone really needs hourly data (although daily data can be obtained using a similar technique, and is probably more useful). I'm mildly offended by the 'breaking into' comment, but that's Wikipedia for you — unnecessary external links are a constant issue, so I can understand over-reacting. To '82.33.205.102': add it back in if you want, but I'm not particularly bothered — just don't start an edit war over this. —Safalra (talk) 18:10, 15 December 2008 (UTC)

As both sets of graphs have now stopped working, I think the point is moot. —Safalra (talk) 09:08, 10 January 2009 (UTC)

The counter .com

I think it's not fair to average the data from The counter .com to others, as what they measure is quite different and will show a much greater share for older browsers and browsers which used to be more popular, and a much lower share for newer browsers and browsers which used to be less popular. It defeats the purpose of having reliable data. Furthermore, The counter .com looks and feels outdated as it doesn't even recognize Google Chrome/Chromium, so I have big doubts about the reliability of it. 12:04, 18 January 2010 (UTC)

Why is "The Counter" getting included in this list? By not specifically recognizing the third most popular and fastest growing broswer, it is less useful than the other sources. At least put it at the bottom of the list. —Preceding unsigned comment added by 99.247.48.243 (talk) 14:50, 19 January 2010 (UTC)

The counter dot com data, from the links on the right of the table, is not monthly as the links imply, but it starts on feb 01 for 2008 and ends on the month listed in the link such as september 08 which is actually a 243 days survey. The same seems to be true for all the other links.

60.231.195.143 (talk) 04:46, 19 December 2008 (UTC) silver_xxx

ps I reread the top posts and this has been handled correctly —Preceding unsigned comment added by 60.231.195.143 (talk) 04:49, 19 December 2008 (UTC)

I derived the monthly data for thecounter.com data for the month of November and I got a very strange result:

Internet Explorer (all versions)	42.77057471%
Firefox	21.01947469%
Safari	9.95037636%
Opera	1.025331744%
Netscape compatible	24.30565329%
Other (Netscape+Konqueror+Unknown	0.929411471%

Strangely enough the Netscape compatible category is about equal to Firefox's share plus three which is about equal to Chrome's share in other sources. Could thecounter.com be double counting Firefox as Netscape compatible? I know Netscape has been using Gecko since 6.x so I suppose it is possible and it has already been established that Chrome could be counted under WebKit as Safari as well as being technically "Netscape compatible". By the way I got these numbers by taking the difference between monthly totals as the total hits for the month in question then I took the difference between months for each browser version. I then calculated the percentage of each browser version by dividing each of their monthly differences by the total monthly difference and multiplied by 100. I added all the Internet Explorer percentages together and lumped all Netscape versions with Konqueror and Unknown for Other. In case this is confusing I'll show you how I got the Netscape compatible value: the difference between November and October's visitors is 92266913-89104911=3162002, but there was an error of 5-4=1 so its actually 3162001. Next the monthly difference was found to be 4428036-3659491=768545 and so the percentage was found to be 768545/3162001*100=24.30565328726967512027984810884.

If you don't believe these numbers then go to http://www.thecounter.com/stats/2009/November/index.php and find out for yourself. Either my whole process is flawed and this point is moot or the method thecounter.com uses is flawed and we may need to exclude it from the summary table calculations or adjust for its flaws. Thorenn (talk) 22:04, 3 December 2009 (UTC)

IE shells

shouldn't we mention the problematic of the ie shells = more "browsers / different guis" --> "same" user agent --> recognized as ie? they are different browsers, its really the same like shared gecko browser: you install 5 gecko browsers/same engine (but al with the engine built-in) and 5trident shells/browsers (layout engine already installeD) mabdul 0=* 20:58, 20 December 2008 (UTC)

Objectiveness of Summary Graph and Stats

Considering this article goes into detail about the different sources, does it not seem a bit subjective to include a graph and statistics in the Summary with only one source taken into account? (Especially since all the sources can be drastically different). Babydutka (talk) 20:13, 29 January 2009 (UTC)

... Well, there is the issue that various sources have quite different time chunks - people updating the graph tended to do it frequently which was nice. Unfortunately, Net Applications hasn't updated in a couple of weeks while they do some mysterious review. In my opinion the pie chart should switch to StatCounter - and perhaps note more prominently this is a single data source. —Preceding unsigned comment added by 162.99.35.70 (talk) 18:29, 14 July 2009 (UTC)

There should be an diagramm, which includes several stats site, so there is a possibility to get the best result. --87.78.23.122 (talk) 01:40, 29 July 2009 (UTC)

I would second basing the graph on StatCounter. Net Applications always seemed to US-centric so odds are that even with their switch in how they caluclate shares, they still overestimate IE global share. The Arkady (talk) 21:23, 13 August 2009 (UTC)

I'm open to doing the same as the Usage share of desktop operating systems page where the chart data is from the median. The addition of a summary would also be good. I'm still waiting for Net App to come out with fresh data so I can update the pages. Jdm64 (talk) 05:08, 29 July 2009 (UTC)

Doing median or any other deriviation is problematic. Different sources have different scale (by orders of magnitude), the methodology is different (apples and oranges) etc. Also, trying to derive any measure will be dangerously close to doing Original research. Having graph with clear attribution to its source is at least non-ambigious. It serves as example. Wikiolap (talk) 17:55, 29 July 2009 (UTC)

Well, the desktop os page currently uses the median, and I think it works fairly well. The percentages are all close and there was all ready a summary table, so using the median in the graph was natural. Will it work on this page? Maybe? I'd like to see at least a summary table because it would make retrieving useful information from the page much easier. Jdm64 (talk) 20:01, 29 July 2009 (UTC)

I don't particularly care *which* one you guys use, but I'd like to note that right now the graph claims to use the median, when that is clearly, wrong - the median for example for IE is 58%, not 64% - the graph should make clear which source it is using at the moment. Heck. I have no idea if the numbers even match the pie slices.

Differentiating between desktop and embedded

I think it would be useful to split off the embedded browser market share from the overwhelmingly dominant desktop share.

Opera Mini, NetFront, Pocket IE, Nokia S60, Safari for iPhone/iPod, Playstation 3 Browser and PSP Browser all compete on platforms like PDAs, phones and gaming devices. Their market share on those platforms is not evident from the 0.00 and 0.01 variously attributed to them here.

— Nicholas (reply) @ 13:17, 4 February 2009 (UTC)

Make Colors in Pie Chart Match Prominent Colors of each Browser

I like the pie chart, but I thought of one way to potentially improve it. Use prominent colors of each browser as the color for that browser's pie slice. For example the blue pie slice already closely (exactly?) matches the color of the "e" in the Internet Explorer desktop icon. However, some of the other pie slices don't match their browser that well. For example, I think instead of yellow for the Mozilla Firefox pie slice, a better choice would be the orange-ish color of the fox's head in the Firefox desktop icon. What do you think? I don't use the other browsers. Perhaps other people could choose more appropriate colors for the other browsers as well. —Preceding unsigned comment added by 70.157.123.148 (talk) 15:47, 11 March 2009 (UTC)

Previously, the graphic did use orange for Firefox (yellow was Chrome), but when Jdm64 replaced that graphic with the current SVG version, he changed the colors. I'd prefer an effort to "match" browsers and colors, as well. --Groggy Dice T | C 09:05, 6 June 2009 (UTC)

Another counter

Here's another counter: http://gs.statcounter.com/ . Rwxrwxrwx (talk) 12:32, 15 March 2009 (UTC)

confused about w3c

http://www.w3schools.com/browsers/browsers_stats.asp doesnt seem to match up with the page. Am i reading it wrong? John.n-IRL 15:09, 19 May 2009 (UTC)

I think you are confusing W3Schools and W3 Counter. -- Schapel (talk) 17:24, 19 May 2009 (UTC)

Notability vs reliability. NPOV

Net Application (Hitlinks) are possibly the most well known and most used stat counter but have often been criticised for being extremely unreliable and generally inaccurate despite popularity. They are given prominence in this article (mainly by the pie-chart at the top). Given that it's certainly the most NOTABLE company, giving it prominence is possibly in line with the fact that Wikipedia places a value on notability, but given that it is a commercial company should it be given prominence in a WP article over other such companies? _ɹəəpıɔnı 06:39, 3 June 2009 (UTC)

Who says NetApplications is "extremely unreliable and generally inaccurate"? Is there evidence that this is so? -- Schapel (talk) 11:23, 3 June 2009 (UTC)

"tracking firm admits data is skewed", "not the first or even second time the numbers published by Net Applications have "mysteriously" changed from one day to the next" _ɹəəpıɔnı 05:48, 7 June 2009 (UTC)

I'm sure all the browser stats are skewed one way or another. I can't imagine how to get a truly representative sample of the usage of web browsers. Net Applications data do not change "mysteriously." The data are run through quality assurance. You're not showing that Net Applications data are "extremely unreliable and generally inaccurate" at all. -- Schapel (talk) 12:52, 7 June 2009 (UTC)

That is precisely my point. I wasn't implying that Net Applications are inaccurate relative to the others, I was arguing against giving them prominence in the article relative to other sources. All stats sources probably have some inaccuracies, Net Appications and their competitors. Does Wikipedia have reason to believe Net Applications is a better source than others? If not, why does the pie-chart show their stats?

Every single browser article on Wikipedia takes it's stats from this page, and giving Net Applications prominence ensures that their stats are always used in every one of these articles.

I cannot say they are significantly less reliable than any of their competitors, but I have given two references to articles commenting on anomalous statistics from Net Applications suggesting that they can occasionally be off. If there's no references to say that they are significantly more reliable, or encounter less anomalies in their statistics, than competitors, I don't see the benefit to Wikipedia to use them as an exclusive source in other articles. Giving them prominence in this one ensures that they will be used as such. _ɹəəpıɔnı 19:31, 7 June 2009 (UTC)

Having changed the marketshare table in Microsoft Windows from being NA-only to including NA, Xiti, W3C, and OneStat, (need to update it!) I am sympathetic to your concerns. However, I think you are mistaken in believing that the prevalence of Net Applications in other articles is due to its pride of place in this article. Rather, both this article and those articles are reflecting their prevalence in the media.

As for why Net Applications is so dominant in the media, I've thought about writing something up on a userpage about why the media prefers NA. But a big part of it, I believe, is this: regular monthly updates. It gives more chances for Net Applications to be written up than those who release their figures more infrequently. It allows reports to plan ahead for the release of their latest figures, as opposed to outfits who release reports more irregularly. Even in stories that aren't directly about browser or OS usage stats, this gives NA an edge. If a journalist is writing up a story on Steve Jobs' health, and wants to look up-to-date, would he rather say, "according to Net Applications, OS X's marketshare last month was...", or "according to OneStat, OS X's marketshare eight months ago was..."? --Groggy Dice T | C 21:01, 7 June 2009 (UTC)

How about maintaining a table and pie-chart showing and averaging the latest results from all the major counters, and use those as the main features of the page? It would cancel out any biases. Rwxrwxrwx (talk) 10:49, 8 June 2009 (UTC)

@Groggy Dice, I take your point about NA's notability in media contributing to their use as a source here, it's a good one. I was speaking from experience of people making points on talk pages referencing this article, but you're probably right that it's not as oft referenced as bloggers/etc. in article references are.

As for the regular updates, I'm not sure how true that is (that this sets them apart). Take the link provided by someone two sections above this - seems a fairly new statistics source as it has yet to be added to the article that I can tell, but an example of a regular source.

@Rwxrwxrwx, not a bad idea though I wonder how easy it would be to maintain such. I was merely suggesting toning down the prominence rather than bringing other sources up to the same level, but that's an interesting idea all the same. _ɹəəpıɔnı 15:28, 8 June 2009 (UTC)

This has been suggested before (cannot find thread right now), but there were several problems with this suggestion. For one, different providers use different methodologies and track different number of web sites, so averaging across them would not be very meaningful. Second, deriving any non-trivial statistical measure is dangerously close to original research. So either we don't have any graphs at all, or keep one for specific provider, but not derived ones. Wikiolap (talk) 15:31, 8 June 2009 (UTC)

The most important problem of NetApplication is that it's not global. They does not take in account good half of the world. It becomes very obvious after you look at StatCounter data, which provide quite accurate country-by-country statistic where you can see IE has extremely large popularity in China and Asia. Safari has almost no popularity in Europe and Asia, while Opera makes over 40% in CIS countries that results in 10% for Europe and about 4% wordwide, while Firefox isn't really choice of Eastern Europeans. Elk Salmon (talk) 19:27, 18 June 2009 (UTC)

If NetApplications were "not global", the number would not match closely with the other global sources. However, NatApplications' data do, but StatCounter's do not. If anything, it looks like StatCounter and W3Counter are "not global". The statistical term for this is that the data are skewed, likely due to an unrepresentative sample. The article explains why W3Counter's data are skewed towards less popular browsers. We should figure out why StatCounter's are also. -- Schapel (talk) 14:54, 18 June 2009 (UTC)

Still more like NetApplications makes their stats only from several western counties leaded by USA and Canada, where Apple products popularity is relatively high. Just like OneStat. I just gave an example. Taking in account active internet population of Eastern Europe and Asia Safari cannot reach so high position, as Apple products are completely unpopular in Europe. As Safari has 8% in North America it cannot be 8% worldwide. Large internet population of Europe and Asia will reduce this number significantly. So as IE has dominant position in Asia while Opera has dominant position in CIS. World is very different. It's like CIS is all about Yandex, China is all about Baidu, USA has large Live/Bingo popularity with no dominant position of Google while rest of the world is all about Google. Same as USA is about Linux, Blackberry, iPhone and Windows Mobile on phones with not, while CIS and Europe is all about Symbian with somehow large popularity of Windows Mobile and with zero popularity of Linux and Blackberry. NetApplications is absolutely obviously onesided. Elk Salmon (talk) 19:27, 18 June 2009 (UTC)

That doesn't make a whole lot of sense. NetApplications matches OneStat and TheCounter fairly closely, and are show global share. StatCounter is even more heavily skewed away from IE and towards Firefox and Opera than Adtech, which shows European usage only. If anything, it looks like StatCounter is "not global". Of course, they are all skewed to some degree one way or another, but W3Counter, Adtech, and StatCounter are among the most highly skewed stats on the page. NetApplications is relatively balanced, as are TheCounter and OneStat. -- Schapel (talk) 20:27, 18 June 2009 (UTC)

There's two issues. Firstly, NA are more English-language oriented and more oriented towards western websites. So when they say they are reporting global stats, they are not lying. They are in fact reporting stats of vistors FROM all countries, but TO websites primarily located in th US, because Hitslinks is primarily marketed in the US. So while visitors shown in NA's stats are from all over the world, they are visitors to websites more oriented towards the English-speaking world, rather than smaller local sites in national languages. Instead of comparing NA/StatCounter to OneStat and TheCounter, try comparing them to the individual stats of smaller local sites in national languages of any given countries. Yandex.ru is probably the biggest example, it matches StatCounter quite closely.

The second issue is the type of websites Hitlinks and such services are marketed at - NA tends to market at corporate customers which will likely lead to higher stats for IE, while marketing at blogs may lead to lower IE stats for StatCounter (as the more technically proficient users will be blog readers and not use IE, while those on corporate networks will probably be stuck with IE). Both lead to inaccuracies, my point here is all are inaccurate so none should be given prominence. _ɹəəpıɔnı 02:01, 20 June 2009 (UTC)

Statistics from Wikipedia

Is it possible to include browser stats from Wikipedia-hits?

I was also looking for this data, would be interesting. --193.166.137.75 (talk) 07:22, 3 July 2009 (UTC)

I agree with you, would be very interesting. Anyone know how to get this data?????? --Marceloml (talk) 23:17, 18 August 2009 (UTC)

Mail.ru stats

[1]

• IE: 48.76
• Opera: 26.39
• Firefox: 20.67
• Safari: 0.56

These stats are quite surprising. Less than half use IE!? Opera more popular than Firefox!? And does no one have Macs in Russia? Even SeaMonkey's beating Safari! Anyone know how reliable this site is? 71.155.236.174 (talk) 08:39, 4 July 2009 (UTC)

My issue with mail.ru stats is that they don't keep history - the link always points to today's data, so it is impossible to verify how the stats looked last month or last year. Wikiolap (talk) 03:20, 6 July 2009 (UTC)

I don't think there's any question of reliability when it comes to a site providing its own statistics of its own visitors, although mail.ru's stats are a little different than StatCounter's (who have IE at around 33%) which I'd usually consider the most reliable global stats source. Opera's usage in that part of the world is the main reason YUI have Opera in the "A-Grade" category of their graded browser support guidelines.

The question of mail.ru stats history is one I've been trying to figure out myself, I wasn't aware they didn't keep a history at all, I just thought their stats history was quite difficult to find as I don't speak Russian. Are you sure they delete them completely? _ɹəəpıɔnı 03:38, 6 July 2009 (UTC)

I wasn't able to locate historic stats from mail.ru (I do speak Russian). Perhaps they exist somewhere, but well hidden. Unless someone can find a link to them, we are risking breaking WP:V. Wikiolap (talk) 04:21, 6 July 2009 (UTC)

I found this which seems to be almost the same. If I'm interpreting it all wrong, apologies - as I said, no russian - but it seems to show stats for mail.ru, whereas the stat.mail.ru site would presumably show stats for all mail.ru domains, including the likes of blogs.mail.ru and video.mail.ru. _ɹəəpıɔnı 06:28, 6 July 2009 (UTC)

Look at real browser stats in Russia, Ukraine or Belarus, before saying stupid things: http://gs.statcounter.com/#browser-RU-monthly-200910-201001-bar http://gs.statcounter.com/#browser-UA-monthly-200910-201001-bar http://gs.statcounter.com/#browser-BY-monthly-200910-201001-bar 95.133.48.57 (talk) 19:29, 9 January 2010 (UTC)

Ordering of stat providers

Shouldn't these be put in alphabetical order? (ADTECH at the top, WebSideStory at the bottom) 82.33.205.162 (talk) 14:29, 8 July 2009 (UTC)

Summary of known sample/bias for stat providers

Would it not be useful to include a brief summary of known sampling issues or bias? e.g. Net Applications is mostly for English sites, AT Institute mostly French, whether they are for one country or many, etc? 82.33.205.162 (talk) 14:29, 8 July 2009 (UTC)

IE8 compatibility mode

IE8 renders (almost exactly) the same as IE7, and declares itself as such in the User-Agent (but with an identifier that it's actually IE8 IIRC) - reference article for discussion here by Net Applications which gives a breakdown of IE8 mode vs IE8 in IE7 mode.

Given that IE8 is effectively being IE7 in this mode (link above says about 10% of the time in that study), how should this be handled? I see options as:

• Lump in with IE7
• Lump in with IE8
• Break out separately

82.33.205.162 (talk) 14:29, 8 July 2009 (UTC)

Usage share of web browsers: June/July 2009

We need an update of the usage share of web browsers for June/July 2009; including a updated pie chart. There is no doubt that these statics have dramatically changed taking into consideration with the fact that Firefox 3.5 has been recently released. —Preceding unsigned comment added by A9l8e7n (talk • contribs) 00:04, 24 July 2009 (UTC)

Unfortunately, the source we have been using for these statistics is Net Applications, and they are still conducting their "review" of the June numbers, and the July numbers won't be out for another week.

However, as far as the pie chart is concerned, the release of FF3.5 probably won't affect it much, since 3.5 will mostly gobble share from earlier Firefoxes, and the pie chart doesn't break down by versions. --Groggy Dice T | C 18:39, 25 July 2009 (UTC)

Exactly, this is getting in the way of showing how many users are actually using firefox 3.5.--A9l8e7n (talk) 19:26, 25 July 2009 (UTC)alen

Perhaps using a source with more regular updates in the top image would suit your purposes better? I personally believe having a single image from a single source at the top of a page representing multiple (often conflicting) sources is grossly subjective, but cycling the sources of that image might make it less so at least. _ɹəəpıɔnı 05:44, 31 July 2009 (UTC)

Summary table

I don't think the summary table should be showing the mean of all stats. The main problem is that some are global stats, and some are for specific geographical locations. If the purpose is to give the latest stats from each source, then let's do just that, and leave the median out. We should also probably consider leaving out non-global stats, because they are listed in the table without any explanation of that they are. -- Schapel (talk) 04:31, 4 August 2009 (UTC)

Well, except for AT Internet Institute, all the sources from the current table are global. I still think that the median is valuable. Removing "AT-II" from the table -- ok. But, the median could be useful in creating a new chart. The current one uses Net Apps, but a more reasonable data source would be the median of all global sources (because how would we determine what 'one' source should represent the chart). This would be similar to the OS market share page. Also, I think that some of the smaller sources should even be removed from the page entirely. The page has become excessively disorderly with the amount of tables. Having more sources adds little value to the page, but an easy to read summary table and carefully picked sources would improve the page. Jdm64 (talk) 07:58, 4 August 2009 (UTC)

I agree that some sources should be removed. mail.ru, StatCounter, and W3 Counter would be my picks for removal. -- Schapel (talk) 01:03, 21 August 2009 (UTC)

As you mentioned, one is non-global. Another is not a one-month stat like the others, and thus skews to older data. It would seem to me that all of these can stay in the summary table, and simply not be included in the "median" calculation. Then again, neither should data older than the month expressly stated in the "median" table row, and yet we are. That seems like an error. And, of course, that's not the median that's being calculated -- it's the mean. (And rightly so, IMHO, but the label should be corrected.) Gnassar (talk) 11:49, 18 October 2009 (UTC)

I agree stats from different months should not be included in the median or even the mean for that matter because these stats often change from month to month by whole percentage points. I think we should only include stats from the current month in the median and mean and include the latest stats from other sources in a separate summary table. 4 of the 7 sources update their stats monthly, the rest update sporadically if at all. If we have 2 tables we can show a meaningful median and mean on one with the last month and have the other table just summarize the other sources' latest stats. Whether or not the sources are globally even is not very important IMHO as long as they have a relatively large sample size; sources covering mostly Europe balance out sources covering mostly North America. In any case the Europe only sources are old and therefore would not factor in anyway. But that's just my idea. I would appreciate feedback on it. Thorenn (talk) 15:52, 1 December 2009 (UTC)

Wikipedia Stats of browser usage share

How can we find stats of wikipedia browser usage share??? Would be good to put these stats in this article --Marceloml (talk) 01:42, 19 October 2009 (UTC)

Prefetching, is Gecko the only one?

"Gecko-based browsers (such as Firefox) can prefetch linked web pages, potentially increasing hits. Link prefetching in Gecko-based browsers is used on pages with enhanced markup, including Google search results."

Does IE 8 do this now as well? 68.13.126.138 (talk) 00:12, 28 October 2009 (UTC)

These sites differentiate between visits and hits. They are reporting browser share by percentage of visitors, not by percentage of hits. —Preceding unsigned comment added by 76.124.117.20 (talk) 00:36, 7 November 2009 (UTC)

is ADTECH used for average?

In the summary section under the table it says: "Note that The Counter value is an average of the past 18 months, and the AT Internet Institute value applies only to Europe. Neither is therefore included in the monthly summary calculations."

But ADTECH is also Europe specific, so it shouldn't be used in the mean and median calculations either. If this is already being dome, it should say so here right? Just wanted to see what's going on with it. - The Talking Sock talk 23:59, 22 November 2009 (UTC)

TheCounter math

TheCounter reports cumulative numbers from Feb, 2008. This makes its reports incompatible with everybody else. Especially for the purpose of the Summary table, where it is excluded from the median. I took cumulative numbers for October 2009, and subtracted cumulative numbers for September 2009 - this results in pure October numbers. Note, that IE8 falls under "Netscape compatible" there, because they didn't bother to include IE8 in the parsing algorithm (and user agent does say "Mozilla 4/0 (compatible; MSIE 8.0 ...)" ). I did not do it for all of their stats, just for summary table, so now the numbers can be included in the mean and median calculations. Wikiolap (talk) 02:28, 25 November 2009 (UTC)

TheCounter is substandart site. No point to include it at all. 95.133.48.57 (talk) 19:25, 9 January 2010 (UTC)

Feb 2009?

Can we get a more recent pie chart at the top of the page? Mathiastck (talk) 01:28, 2 December 2009 (UTC)

Removing TheCounter

I think TheCounter data should be removed from this article. They exclude Google Chrome, and apparently also Internet Explorer 8. Furthermore, their way of reporting statistics is incompatible with other websites. We have enough sources, so if there are no objections I will remove TheCounter. Mushroom (Talk) 12:22, 18 January 2010 (UTC)

I don't see either of the above points as grounds to remove TheCounter. If we were to only include statistics from companies deemed (in the SUBJECTIVE opinion of an arbitrary Wikipedia editor) "reliable", we may as well remove every table from the page. It is not Wikipedia's place as an encyclopaedia to judge the performance of companies included in its articles - it just documents objectively. TheCounter seems to meet inclusion requirements for Wikipedia (e.g. WP:Notability) - beyond that, I don't see any reason not to include it. _ɹəəpıɔnı 17:35, 18 January 2010 (UTC)

Actually I don't think it is notable, and that would be another reason to remove it. The same is true for most other sources, by the way, except maybe Net Applications which seems to be the only one notable enough to have its own article. Anyway, since you disagree with my proposal I will leave it there. Mushroom (Talk) 18:01, 18 January 2010 (UTC)

I am not opposed for removing TheCounter from the article. Editors of the Usage share of operating systems also decided to remove this source.Wikiolap (talk) 18:26, 19 January 2010 (UTC)

I agree we should remove it. It has multiple issues that make it difficult to work with and which make it incompatible with the rest of the stats. The only reason I did not remove it when I changed the layout of the summary table was that it updates monthly although even that has issues. The main problem is that it has a historical usage share table, the removal of which would probably be disruptive if it were done wholesale; perhaps we should move it to its own section since it isn't outdated as far as having a lack of new data goes. The summary table can be changed to exclude it from the calculations of mean and median like the non-monthly reports. On that note does any one object to those reports being moved to the older reports section since they are no longer relevant to the present year? Thorenn (talk) 19:39, 19 January 2010 (UTC)

I concur with this proposal, except TheCounter shouldn't be included in the summary table at all because of its bad sampling method, it would only be misleading. I also think the calculating method should be included in each section and those sources that does not publish the methodology - if any - should be removed. At this point it's only reported in w3counter. --Sapeli (talk) 05:52, 20 January 2010 (UTC)

I agree TheCounter has a bad sampling method but it would be problematic to remove it from the summary table; at least in its current form. Perhaps we should label the main sources as something along the lines of "Monthly updating sources with data from only the relevant month" or place an end-note to that effect. As for your second comment, I am not sure what you mean by "methodology". All the sources give information on how their data is generated: StatCounter Net Applications StatOwl and of course W3 Counter. —Preceding unsigned comment added by Thorenn (talk • contribs) 15:36, 20 January 2010 (UTC)

Any counter included here should be reliable and up-to-date. If The Counter doesn't meet those standards, it should go. Regarding the summary table, it was originally introduced simply to provide average figures for feeding to the pie-chart. Any counter which does not contribute to the average figures should not appear there; all it does is clutter up the table. Rwxrwxrwx (talk) 15:52, 20 January 2010 (UTC)

There are many sources in the article that do not have up-to-date information. It has been this way for years. We should retain older sources of information so we can see the change over time. TheCounter is an important source in this respect because it was one of the few sources that had information in the first several years of the millennium. -- Schapel (talk) 16:18, 20 January 2010 (UTC)

I phrased that badly; what I meant was that any counter which is not reliable and up-to-date should not be included in the summary table or be regarded in the same light as those which are reliable and up-to-date. You're right that if it's historically significant it should be kept, in some form. Rwxrwxrwx (talk) 16:28, 20 January 2010 (UTC)

Since we already have an "Older reports" section, we could just move TheCounter and other unreliable or outdated sources there, and remove them from the summary table. This way no information would be lost (except from the table, but it wouldn't be very significant), and we would solve the problem. By keeping TheCounter at the top of the list we are implying that it is one of the most reliable sources, which it clearly is not. Mushroom (Talk) 16:55, 20 January 2010 (UTC)

If we move TheCounter's historical usage share table to the Older Reports section we would have to rename it to include unreliable sources because TheCounter's data is technically up to date while still being unusable for calculations. If we reserve the Summary Table for sources used to derive mean and median then the Other sources part of the table will need to be removed. This is not problematic for most of the sources listed because they are listed again in their historical usage share tables but the recently added Wikimedia stats are not listed elsewhere on the page and I don't know where else it could be put should we choose to retain it. Thorenn (talk) 18:10, 20 January 2010 (UTC)

TheCounter's data is not up-to-date, because they have no data for January 2010. How about we wait until February 1, 2010, and if there still isn't any TheCounter data for 2010, we move it to the Older Reports section, which is exactly where it will belong. -- Schapel (talk) 20:00, 20 January 2010 (UTC)

Of course they have no data for January 2010; the month is not over yet! None of the sources give preliminary data (with the exception of StatCounter) unless you are a paying member. Therefore, as far as having data from the previous month TheCounter's data is up-to-date. Had it not been it would be a simple matter to move it to the Other sources part of the Summary Table and move its historical usage share table to the bottom like the rest of the old sources that no longer update. Since it technically qualifies as up-to-date, moving or removing it is a non-trivial matter since we have to redefine what sources are suitable for inclusion in the article and where they should be in the page. Thorenn (talk) 20:31, 20 January 2010 (UTC)

Every month for the past several years, TheCounter has had data available on the first of the month. They have never before waited until the end of the month to provide data. In the past, it has been updated every hour of the month. The fact that this has not been happening for 20 days and hasn't been fixed yet means it's unlikely for it to be fixed soon. -- Schapel (talk) 21:36, 20 January 2010 (UTC)

But it has been updated and has been since the 31 of December; what wasn't updated was Wikipedia's link to the December data. I have fixed that now so it links to this instead of this. Thorenn (talk) 22:19, 20 January 2010 (UTC)

But every month for the past several years, TheCounter data for the current month has been updated on the TheCounter site every hour. On January 20, 2010, there is still no data for January 2010 at the TheCounter site (the link gives a 404 error). Therefore, the TheCounter site is not up-to-date. If this continues until February 1, 2010, I propose we rightfully move TheCounter data to the Older Reports section. -- Schapel (talk) 01:51, 21 January 2010 (UTC)

If this is the case then yes we can take that action then. In the meanwhile I have moved the stats in the Other sources part of the table to the Older reports section so it does not clutter up the Summary Table. —Preceding unsigned comment added by Thorenn (talk • contribs) 17:36, 21 January 2010 (UTC)

Swapping Chrome and Safari

Chrome keeps gaining usage share and Safari is practically stagnant in comparison. Should there be a point at which we should swap the places of Chrome and Safari for the whole page? If so, have we already passed that point? If we do this it will require a major overhaul of the page layout because columns cannot be interchanged easily. Every month we add at least another 4 rows of data so every month we spend without a decision on this means more work if we eventually do decide to swap them. If you are interested in this page please weigh in on this issue. Personally I think we should do it but section by section and involving multiple editors if possible perhaps using this template. Thorenn (talk) 17:46, 1 February 2010 (UTC)

I'm totally for it. I might be able to to the conversion in one go using this converter. Jdm64 (talk) 22:04, 1 February 2010 (UTC)

I did one table, other's can help doing the rest Jdm64 (talk) 04:37, 2 February 2010 (UTC)

I did the rest, all the historical usage tables are now consistent in the Chrome, Safari order Dangrossman (talk) 08:40, 2 February 2010 (UTC)

Thank you Jdm64 and Dangrossman for your work but there are still the older report sections that haven't been changed. Do you think we should convert those as well to the new order too for consistency or leave them as is because they do not show Chrome's recent rise? Thorenn (talk) 14:45, 2 February 2010 (UTC)

Accuracy Notes

I feel the accuracy notes may be misleading. Correct me if I'm wrong, but all 5 services this page pulls stats from are image/javascript-based trackers. None are web log analyzers. In that light, some the sources of overestimation listed do not actually occur:

"A web browser that refreshes the webpage at a regular time interval." -- These services all report browser share by visitors, not by page views. Refreshing the page does not alter these numbers.

"A feed reader that requests the RSS or Atom feed at a regular time interval." -- A feed reader is only downloading the feed, XML content, it's not executing any JavaScript or images these trackers could use. I have never seen an RSS feed contain one of these website trackers -- it doesn't make sense, as so many people use feed readers that wouldn't parse embedded JavaScript. The top feed readers would strip it out for security reasons. People that want feed stats get them by proxying their feed through something like FeedBurner, not by pasting a web tracker into the feed bodies.

"Extra files like CSS hacks and JavaScript hacks are often sent to Internet Explorer." -- These are not log analyzers, those requests are never recorded by these services.

"Many types of software, such as Web validators or crawlers, fetch web pages, and send fictitious user-agent strings to appear more like normal traffic." -- Web crawlers do not execute the tracking JavaScript and run the code that inserts the tracking image into the page, they are not counted by these types of services.

And from the underestimate section:

"Generally, the more faithfully a browser implements HTTP's cache specifications, the more it shall be under-reported relative to browsers which implement those specifications poorly" -- These companies send appropriate headers to tell the browser not to cache tracking code, so the opposite is true -- a faithful implementation of cache specification will result in not caching and recording the visit.

"User Agents are not guaranteed to be a certain format. As an example of the inconsistency almost every User Agent pretends to be Mozilla 5.0." -- The fact that almost every user agent includes that as PART of the otherwise browser-specific string does not make it inconsistent.

Dangrossman (talk) 08:54, 2 February 2010 (UTC)

As the article says, "Measuring browser usage in the number of requests (page hits) made by each user agent can be misleading." That's why usage share is measured by tracking visitors, not page hits in the log. I'm not sure why people keep elaborating about how using page hits is misleading. -- Schapel (talk) 12:51, 2 February 2010 (UTC)