Wikipedia talk:Did you know/Statistics/Archive 2

Archive 1

Archive 2

variation in statistics

The calculation of DYK Stats is kind of complex. Thankfully, checking the day/time recorded at Wikipedia:Recent_additions and entering it into this dykstats tool makes it pretty easy. However, many people don't appear to use this so come up with wildly varying numbers. Just looking at those for August 2016 so far:

• Shawshank Tree - number is the raw views for the day
• Scaly-foot gastropod - number is the raw views for the day
• Trump Vodka - number is the raw views for the day
• Manika Batra - for some reason adds two days' pageviews and subtracts half of the views on the two days on either side
• Khizr and Ghazala Khan - I just added this one based on the dykstats tool (seems to be one day minus half of the previous/next day's -- though I'll note that single day's hits are slightly different from the number reported in the pageviews tool).
• Dean Smith (pilot) - raw views for the day
• Flag of Southern Rhodesia - raw views for the day
• New York Yankees appearance policy - just adds two days' pageviews together?

In other words, it looks like the one I just added is the only one that follows the "[pageviews for the day] - [previous day + next day]/2" formula. Am I going by outdated information? It sure seems like it would be a lot more straightforward to just stick with the single day pageviews without worrying about calculations. I presume that the people adding two days together are doing so because something ran from e.g. 12:00 to 2:00 or 23:00 to 12:00 or something? Understandable, but I don't think there's any way to break down by hour, so it seems it would be better to just go with the day for which it's listed at Recent Additions rather than count 48 hours worth of pageviews for a 12-hour DYK. — Rhododendrites ^talk \\ 20:57, 13 August 2016 (UTC)

• Pinging editors who added the above: @The C of E, Snek01, EvergreenFir, Doug Coldwell, and Dee03: — Rhododendrites ^talk \\ 21:00, 13 August 2016 (UTC)
  • Generally DYK articles don't garner many views on their own as by their very nature they are new articles that don't have time to get lots of links to them in the early days.Generally it's around 60 max without DYK clicks. So I don't think that is nessecarily an issue when hooks overrun. I have to disagree with the notion that it should be only on the day it runs that counts simply because although in theory it works, in practice hooks do go past 12-12 from time to time. For example, the Yankees hook only had about 3 hours on the day it ran compared to Trump Vodka which had 12. Seems only fair to give all hooks the opportunity to gain the views for inclusion the same time period. The C of E God Save the Queen! (talk) 22:17, 13 August 2016 (UTC)

I think hooks posted outside the 12/12 guideline should just be treated as outliers, or the day that "counts" should just be the day a hook appears on for the longest. No mechanism is going to be perfectly fair, but couting 48 hours worth of pageviews to m aske up for slippage on a 12 hour cycle is nonsensical (I.e. "my hook didn't get credit for all 12 hours, so it's only fair to act like it ran for twice as long as all the others"). — Rhododendrites ^talk \\ 03:54, 14 August 2016 (UTC)

• I used the Day1+Day2-(Day0+Day3)/2 formula because that set ran from 12:35 6 August 2016 (UTC) to 07:11 7 August 2016 (UTC) which is 11+ hours on Day1 and 7+ hours on Day2. I agree the set overran its course on Day2, but even then Day1 alone got 5,781 views and therefore the hook qualifies for the statistics page. Using raw pageviews data does not give us a clear picture on popular topics like Tiffany Trump which generate high traffic on a daily basis, and IMO the formula (perhaps a better one) should be used for all hooks consistently. "the day that "counts" should just be the day a hook appears on for the longest" seems unfair on the hooks that get say 5 PROPRIOhours on the first day and 6 hours on the second. Dee03 05:49, 14 August 2016 (UTC)

• Just to be clear, I'm not trying to remove any from the list. I just want it to be consistent. It's sort of meaningless otherwise. — Rhododendrites ^talk \\ 13:44, 16 August 2016 (UTC)

I'm brand new to all of this and will defer to others. For Shawshank tree, there were a number of views the first few days but it tapered off. The DYK say produced a huge boost in views that quickly tapered if as well. Seemed logical to me at the time to just include the one day of views. But honestly not sure how long it was up. EvergreenFir (talk) 06:07, 14 August 2016 (UTC)

If we're going to require calculation like we technically do now (despite almost none of the current entries using that calculation), we might as well make it more precise. This could, of course, easily be automated, and I'm sure there's a simpler way to express it than I am here:

• H1=Hours (rounded to the nearest hour) it ran on the day it appears under at Wikipedia:Recent additions
• H2=Hours (rounded to the nearest hour) it ran on either the day before or the day after the day it's listed under at Recent additions, if applicable. (presuming it won't run for 3 days)
• V1=Pageviews for the day it appears at Recent additions.
• V2=Pageviews for the day it ran for H2 hours.
• X1=Pageviews the day before it ran (will be either two or three days from X2)
• X2=Pageviews the day after it was removed (will be either two or three days from X1)
• DYKSTATS=V1(H1/24)+V2(H2/24)-((X1+X2)/2)

I know that looks obnoxious spelled out like that, but the only difference between that and what it's supposed to be now is that it factors in hours it ran per day. — Rhododendrites ^talk \\ 13:44, 16 August 2016 (UTC)

The dykstats tool is something I wrote - it follows the formula on WP:DYKSTATS (which is imperfect, I think somewhere on this page there is a discussion about it). My intention was always to look to automate it (I mean; the numbers and the article dates/times is all accessible data) rather than leave it to editors to remember to list. However, time got scarce and then one thing led to another... Numbers have always been rather fudgy on this page (one reason I wrote the dykstats tool was so that I didn't keep calculating them wrong :)) --Errant ^(chat!) 14:14, 16 August 2016 (UTC)

Between September and October

Elin Rombo appeared on 30 Sep, archived 1 Oct. For which month should she be added? --Gerda Arendt (talk) 08:32, 4 October 2016 (UTC)

I'd say put it in the month that it ran in. The C of E God Save the Queen! (talk) 09:16, 4 October 2016 (UTC)

Hooks added by WP admin staff split pageviews

As a result of the DYK placement on Oct 18 '16, my Alan Hale page expansion got over 12,000 additional views. Unfortunately, the Wiki admins added a link to the Heaven's Gate page to my hook (as well as one to the Hale-Bopp comet page). This change (which I briefly considered doing myself - but purposely avoided after realizing the consequences) gave the Heaven's Gate page approx 21,000 extra views. The 9000 difference would all (or at lease mostly) have gone to the Hale page if my hook had been unaltered. From there it was an easy navigation to the Heaven's Gate page (and comet page) if a reader was interested. This change actually drove most traffic generated from the hook to the Heaven's gate page (some through a visit to the Hale page, but 9000 circumventing that path.) My suggestion would be to not add links to a hook for anything beyond the spotlighted page. In my opinion DYK should drive all traffic through the page which is supposed to be the one being highlighted here!

Due to the edit to my original hook, I am not sure which if any of the Wikipedia:Did you know/Statistics categories applies to the Hale page's DYK appearance. I welcome any feedback. RobP (talk) 17:46, 19 October 2016 (UTC)

Excellent observation, RobP. I've raised a similar question here. LavaBaron (talk) 21:30, 3 December 2016 (UTC)

Trouble with dykstat tool?

On 4 January 2017, a hook based on a fact from the article Eriksen M/25 ran. Now, according to the pageview analysis the article got 18,789 views that day (seems the hook ran for 24 hours), but the dykstats tool doesn't seem to work for the article. Does the dykstats tool not work for 24 hour hooks? Manxruler (talk) 18:42, 8 January 2017 (UTC)

It's probably the slash in the name of the article. I'll look at it. --Errant ^(chat!) 08:36, 9 January 2017 (UTC)

That sounds logical. That slash has been problematic. Hope you can help. Manxruler (talk) 19:40, 9 January 2017 (UTC)

Too early for tool?

Coon hunting ran on May 9 and according to my own calculation got approximately 10k views. When I plugged it into the stats tool though, I got an error message saying that it was too recent and wait till May 11. Isn't that today? I wanted to add it to the statistics list for the month. White Arabian Filly ^Neigh 22:28, 11 May 2017 (UTC)

Not sure about that, I'm getting it through OK. According to the stats, it got 9,897 views. The C of E God Save the Queen! (talk) 22:47, 11 May 2017 (UTC)

I was talking about the ErrantX tool, which still isn't working. (?) I added it to the statistics list for this month, though. White Arabian Filly ^Neigh 21:38, 12 May 2017 (UTC)

Archiving

Could someone archive the April 2017 list? Thanks, Yoninah (talk) 13:45, 4 July 2017 (UTC)

Odd view statistics

One of my past DYKs, Hawaii and the American Civil War had a very odd jump in views three or four days after it was promoted on the main pag which was not reflected in any increase linkage of the article or promotion of a related DYK. Wondering why this happened?--KAVEBEAR (talk) 07:17, 24 January 2017 (UTC)

@KAVEBEAR: It most likely was linked from some other website after someone found it interesting. Daniel Case (talk) 04:49, 14 August 2017 (UTC)

Threshold for views for monthly stats

Since DYK recently moved back to 2 updates a day after ~5 months or so of one queue a day, should the thresholds for inclusions be reconsidered? (I have no idea what would make for a short snappy list, just it does seem weird if we effectively catalog articles from that 5 month period much more heavily than ones from other times, due to all of them presumably getting higher page views from longer time on the main page.) SnowFire (talk) 06:30, 23 September 2017 (UTC)

I'm still confused by how this works...

Hi, I think I have two hooks which qualify, but I don't know how to find out the number

I seem to get different results depending on which site I use, and I'm not sure which is the right way to do it. Say I want to check Schizomyia viticola on 2015-10-25. If I just look at the graph where it says "check views" and go see how many views an article got that day, I get 8,160

https://tools.wmflabs.org/pageviews/?project=en.wikipedia.org&platform=all-access&agent=user&start=2017-10-15&end=2017-11-04&pages=

If I use https://tools.wmflabs.org/dykstats/, I need to pick a time slot which confuses me a bit since hooks run for 24 hours not 12. And if I select the one starting at midnight, I get a negative number

https://tools.wmflabs.org/dykstats/count.php?article=Schizomyia+viticola&date=20171025&timeslot=AM

Calculation: - ((14 + 8305)/2) = -4160 Views

If I select the one starting at noon I get a positive number

https://tools.wmflabs.org/dykstats/count.php?article=Schizomyia+viticola&date=20171025&timeslot=PM

Calculation: 8390 - ((14 + 45)/2) = 8361 Views

And 8361 is close to 8160, but they're not the same, which also confuses me.

I took a guess and used the wmflabs tool for the number but linked the pageviews, as can be seen in https://en.wikipedia.org/wiki/Wikipedia:Did_you_know/Statistics#October_2017

But I don't think this is right.

Thanks for any help. :) Umimmak (talk) 12:14, 27 November 2017 (UTC)

Also unclear re above

I have a non-top hook with >15,000 (Shapira Scroll). Onceinawhile (talk) 15:15, 21 May 2018 (UTC)

The viewership formula should change

The viewership formula is mistaken. V_DYK = V_X − (V_X−1 + V_X+1)/2. The day before V_X−1 gives an indication of the background views, that's not disputed. However, the day after count V_X+1 contains not only the background views, but it also has a post-DYK "bump". So a fairer accounting would have been V_DYK = V_X + (V_X+1-V_X−1). This equation subtracts the background views, while adding the post-DYK day "bump". Dr. K. 18:38, 9 June 2019 (UTC)

Case in point: The article Temple of Hera, Mon Repos according to the Pageviews stats in the 14 days before its DYK appearance got 1 day of 34 views, 1 day of 19 views, and the rest of the 12 days before DYK were mostly single-digit views, with 1 day at 14 views, 2 days at 11 views, 1 at 10 views and the day before DYK (V_X−1) with 13 views. During its DYK appearance it got 5031 views, and on the day immediately after DYK (V_X+1) it got 87 views, while on V_X+2 the viewership fell to 25, with the subsequent days to that falling to 16, 25, 27, and 38. It is clear that V_X+1 with 87 views is an aberration and it contains a post-DYK day bump. To add (87+13)/2 and then subtract it from the 5,031 of DYK day is simply wrong. The number 87 was never anywhere near the rest of the pageviews for this article on any other day. Dr. K. 08:10, 10 June 2019 (UTC)

By that logic it should be V_DYK = V_X + V_X+1-2V_X−1 , which subtracts the background views from the post DYK bump too. --Spacepine (talk) 13:10, 25 July 2019 (UTC)

Fair point, Spacepine. I agree. Dr. K. 21:47, 27 July 2019 (UTC)

12 vs 24 hour DYK

My article ended a 12 hour DYK yesterday morning, and I have no idea how one determines the DYK views. The pageview tool does not do less than 24 hours, so how is this data determined, yet the notice and link placed on my talk page makes it sound easy. What am I missing? Also, are 12 hr DYKs on equal footing for ranking with 24 hour DYKs? If so, that is beyond unfair.

On 11 May 2020, Did you know was updated with a fact from the article Unorthodox: The Scandalous Rejection of My Hasidic Roots, which you recently created, substantially expanded, or brought to good article status. The fact was ... that the Netflix miniseries Unorthodox was inspired by Deborah Feldman's memoir Unorthodox: The Scandalous Rejection of My Hasidic Roots? The nomination discussion and review may be seen at Template:Did you know nominations/Unorthodox: The Scandalous Rejection of My Hasidic Roots. You are welcome to check how many page hits the article got while on the front page (here's how, Unorthodox: The Scandalous Rejection of My Hasidic Roots), and it may be added to the statistics page if the total is over 5,000. Finally, if you know of an interesting fact from another recently created article, then please feel free to suggest it on the Did you know talk page.

RobP (talk) 01:23, 13 May 2020 (UTC)

How do I calculate the number of dyk hook views?

This article, Gunjan Saxena, was on the main page on 18 August 2020. How do I check the number of views via only main page clicks? DTM (talk) 06:13, 20 August 2020 (UTC)

If you go into the talk page DYK banner, click on where it says "check views" and you should get it @DiplomatTesterMan:. The C of E God Save the Queen! (talk) 06:39, 20 August 2020 (UTC)

More precisely, the article's talk, Talk:Gunjan Saxena. Other method: in the article (any article), you click on "View history", then on "Pageviews" - no need to search for that small print "check views". --Gerda Arendt (talk) 06:51, 20 August 2020 (UTC)

DTM, but now we get to the real question how to calculate, and for her, that's tough because she caused sensational views before DYK, she's certainly no case for the DYKSTATS, with the formula subtracting "normal" traffic. --Gerda Arendt (talk) 06:57, 20 August 2020 (UTC)

(edit conflict). The C of E thank you for the reply. What you are suggesting has always got me the result in the past. But the problem this time is that the organic views skews the final tally. The total view count for the day is 69,617 views. So I read the rules section properly. Applying the method mentioned under rules for such as situation: V_X − (V_X−1 + V_X+1) / 2

69617 - (89420 + 43487)/2 = 3163

I get 3163... So this doesn't go into the statistics table since it is clearly less than the 5000 limit. That's all I wanted to ask. Thanks.

thanks, Gerda Arendt DTM (talk) 07:18, 20 August 2020 (UTC)

Well, you found the same yourself, but somehow I feel it's unfair to not list it somehow? I don't know how we could make an exception, but her case is truly exceptional. I often have a similar case, article appearing In the news, with high numbers, and later DYK, with much smaller numbers, but - thanks to the slowness of DYK, it has so far been apart enough to not cause the problem. Today Bernard Ładysz. --Gerda Arendt (talk) 07:27, 20 August 2020 (UTC)

Sorry, bad example, he wasn't on ITN (I saw his death too leate.) Better: Hellmut Stern, 2 peaks, far enough for him to make stats. --Gerda Arendt (talk) 08:10, 20 August 2020 (UTC)

Page views upon article move

If an article was moved to a different name during the DYK period, how can I check total views? The pageviews tool is only calculating views at the current name, and it won't let me enter the old name so I can check views under the previous title. Thanks! 78.26 _{(spin me / revolutions)} 15:21, 20 November 2020 (UTC)

All-time lists -- time to cull?

Now that we've implemented a revised threshold for the monthly STATS page, let's take a look at the all-time lists, which have grown far too long. The all-time lead hook list now has 160 entries, and the all-time non-lead hook list has 237 entries. The huge size of these lists undermines their purpose (i.e., recognizing the truly exceptional all-time best hooks) and bloats the size of the page (now over 224,000 bytes). In order to ameliorate these problems, I propose the following adjustments:

• Increase the threshold for the lead hook list from 25,000 to 35,000 page views. This would cull the list from 160 entries to 52 entries -- a figure that is more in line with the spirit of an "all-time" best list.
• Increase the threshold for the non-lead hook list from 15,000 to 25,000 page views. This would narrow the list from 237 entries to 43 entries.

What do others think? Cbl62 (talk) 12:23, 3 December 2020 (UTC)

Shouldn't using views/hour apply consistently to monthly and all-time lists to account for different run times? MB 14:48, 3 December 2020 (UTC)

My first inclination was to propose just that.

Using a views/hour approach, a cutoff of 4,000 views/hour gives us 57 lead hooks. However, such an approach under-represents more recent years: Only 4 of the 57 lead hooks with at least 4,000 views/hour are from the past five years. Under that approach, 31 of the top 57 hooks are from the years 2008-2010. I don't know for sure why this is the case, but it may be driven by the fact that, when we had six-hour queues, hooks that were posted only during prime viewing hours racked up massive views/hour.

In order to avoid such a severe skewing to the years 2008-2010 (and away from 2016-2020), my conclusion is that we are better off staying with total views for the all-time lists. Cbl62 (talk) 17:34, 3 December 2020 (UTC)

I'm lukewarm on the idea of changing the threshold for the all-time lists. We've been doing this for many years now, so the "all-time" designation covers a large chunk of time. Someone whose hook exceeds 25,000 hits (like mine just did—my first success!) is going to be proud that they rank up there with previous leaders. The increase in hooks that meet the threshold is showing, IMO, the increase in interest in the DYK section by main-page readers, which is a good thing.

I'm having trouble following this second discussion about page views. Yoninah (talk) 17:42, 9 December 2020 (UTC)

A few thoughts in response:

(1) Adjusting the threshold, in the manner I proposed, would also wipe out all of my hooks from the all-time list, but the lists have simply become enormous. At 400 combined entries (and growing rapidly), the all-time lists have simply lost legitimacy as bona fide all-time lists. An all-time list, to my way of thinking, would be somewhere in the 25-50 entries range -- pushing it to 100 might be conceivable, but 400 combined entries is simply not an all-time list.

(2) Aside from the bona fides of list with 400+ entries, there is an issue of page size. At 224,000 bytes, this page is far bigger than it should be.

(3) What we could do, and an idea I like a lot, is to create annual lists for the hooks that get the most page views. I've already cobbled one together at Wikipedia:Did you know/Statistics/Archive 2020#Top hooks of 2020. When time permits, I'll create an alternative version of the all-time lists with yearly best lists so that you and others can evaluate it.

(4) Is the part of the discussion you had trouble with the part dealing with views / hour in the all-time lists? The problem in a nutshell is that if we change all-time lists to rank them based on views /hour, a huge percent (roughly 70%) of the all-time hooks based on views / hour would date from the 2008-2010 time period.

Best, Cbl62 (talk) 00:20, 10 December 2020 (UTC)

Thank you for the clearer explanation. As I was reading it, I was already thinking, "Why not do an annual list?" I've seen editors get excited when their hook makes the monthly pageviews list (minimum 5,000 views), so an annual list makes perfect sense.

I'm loathe to slough off hooks that are already on the all-time list. I think you should leave them alone, include the relevant hooks on an annual list, and then make any threshold adjustments going forward from 2021. Best, Yoninah (talk) 00:26, 10 December 2020 (UTC)

• Oppose Why should that which was already there before these discussions be chucked out because we arbitrarily decided to cut the entrace criteria? I would say at the very least, grandfather in those already there and say from next year, only if it meets these new rules (once discussed and consensus made of course) can new articles get in. The C of E God Save the Queen! (talk) 16:49, 11 December 2020 (UTC)

Low performing hooks

We have long tracked our best performing hooks but have not done the same with our lowest performing hooks. On the theory that we learn as much from the extremes on both ends of the spectrum, I've begun tracking our lowest performing hooks in my user space. Anyone interested can find the list at User:Cbl62/DYK low performing hooks. Cbl62 (talk) 11:36, 9 February 2021 (UTC)

WP:DYKSTATS

Hi, is there some reason why so many hooks are out of order? I'm wondering if you're trying to order them by per-hour pageviews rather than total views. Since the total views column comes first, it makes for a very disorienting presentation. Yoninah (talk) 15:58, 15 December 2020 (UTC)

• @Yoninah: I had them in order. Since we switched to views per hour, that's how I've been ordering them. If our qualifying principle is the quantum of views a hook receives per hour on the Main Page, it seems wholly logical to order them according to their performance based on that qualifying principle. I thought there was agreement on that. Cbl62 (talk) 16:44, 15 December 2020 (UTC)

I never saw any consensus for that. We have always ordered them by total views, no matter how many hours they appear on the main page. All those "all-time" statistics do not take into account how many hours they ran. I think you need to open an RfC before you implement the changes you have in mind. Yoninah (talk) 16:53, 15 December 2020 (UTC)

We had also always previously used total views as the controlling factor for whether a hook qualified for DYKSTATS. We then decided by consensus to switch that qualifying factor to views per hour. Since that consensus was reached, I've been ordering them, logically, according to that principle. Cbl62 (talk) 17:07, 15 December 2020 (UTC)

I got that. But the first column is total views. And it looks really odd putting a 6,000-views hook before a 27,000-views hook. Yoninah (talk) 18:13, 15 December 2020 (UTC)

I don't find it disorienting. The first column is the article name, and there's no disorientation resulting from the fact that we don't sort everything based on that column. Total views is the third column, and views per hours is the fourth column. But would it help from your point of view if we were to switch those two columns so that views per hour is third and total views is fourth? Cbl62 (talk) 18:21, 15 December 2020 (UTC)

An example of why we should sort by views per hour: If we sort by total views, we place the New Zealand MPs hook (898 views per hour divided among 19 articles) ahead of the City Investing Building hook (1,588 views per hour). Yet, the performance of the latter hook is FAR more remarkable. By sorting according to views per hour, we reward the truly high-performing hooks. Cbl62 (talk) 18:41, 15 December 2020 (UTC)

I hear you. But I think we need more input. At least let me ping Gatoclass for comment. Yoninah (talk) 23:53, 16 December 2020 (UTC)

Hi there Yoninah, and thanks for the ping.

Given that we've switched from sorting by total number of views to most views per hour, my view is that the views per hour column should take precedence, ie it should be the third column and total views should be the fourth. It's not clear to me who is advocating what in the above discussion, but that's my view. Gatoclass (talk) 10:00, 17 December 2020 (UTC)

@Gatoclass: thank you. OK, I defer to consensus. Please switch the columns starting in December, Cbl62. And please set up something for a yearly chart to reduce the byte count on what we have now. Thanks, Yoninah (talk) 12:27, 17 December 2020 (UTC)

• Personally I don't think this conversation should be happening here, it should be happening on the DYK stats page where everyone can see it. Personally, I reiterate that such a dramatic change such as this should have gone through a formal RFC. The C of E God Save the Queen! (talk) 12:48, 17 December 2020 (UTC)

• Thank you, The C of E. I agree. I mentioned opening an RfC above, but I felt pressured to agree with Cbl62's changes. Yoninah (talk) 12:55, 17 December 2020 (UTC)

• @Yoninah: I do feel that such a big shift in the way things are done to a much more complicated system needs a proper RFC rather than just being done on a whim. I would fully support the opening of an RFC (and temporary restoration of the old system until such time as there is formal consensus to change would be a bonus, but I'd be satisfied with just the RFC being opened so we can have a decent discussion on it). The C of E God Save the Queen! (talk) 13:11, 17 December 2020 (UTC)

Yoninah, by all means start a discussion or an RFC at WT:DYK if you disagree with Cbl and myself, but I very much doubt it will result in a different outcome. Gatoclass (talk) 14:22, 17 December 2020 (UTC)

OK. But I don't think the RfC should be about which column goes first. The bigger issue is the 500 pages per view. We now have hooks getting 11,900 hits and not qualifying for the monthly list because they ran in a 24-hour cycle. This is starting to bother regular contributors who are used to 5,000 pageviews earning a place on the list. I'll try to put together an RfC, but not today. Yoninah (talk) 17:55, 17 December 2020 (UTC)

@Yoninah: Are you now opposed to (a) sorting by views per hour, or also (b) using 500 views per hour as the threshold (a decision made after notice at the DYK talk page and a decision that you fully supported)? Cbl62 (talk) 18:33, 17 December 2020 (UTC)

Here's a better idea then. Instead of making it complicated with the views per hour thing, you just keep it simple and increase the number of views required on 24 hours cycles. Personally I think this is a case of WP:AINT because I don't see a reason to change anything from what we originally had. The C of E God Save the Queen! (talk) 18:51, 17 December 2020 (UTC)

We already calculate views per hour for every hook posted at DYKSTAT so it's not adding any layer of complication. Cbl62 (talk) 18:59, 17 December 2020 (UTC)

Most people I've noted don't include them and they get added later. I don't add them myself because I consider them surplusage from the main rationale which is the total number of views. The C of E God Save the Queen! (talk) 19:04, 17 December 2020 (UTC)

The real issue is this. When I created DYKSTATS more than a decade ago, we were running six-hour queues. I selected 5,000 views because that was a really extraordinary level of achievement for a hook. We then switched to 8-hour queues and then 12-hour queues and now 24-hour queues. By keeping the same threshold of 5,000 views, we have dramatically depreciated the significance of DYKSTATS. As a result, in November, we had queues where the 'majority of the hooks qualified for DYKSTATS. Do you really believe that the threshold should be the same today for a 24-hour queue as for a 6-, 8-, or 12-hour queue? Cbl62 (talk) 19:13, 17 December 2020 (UTC)

I'm not saying that. I am saying I am open to increasing the total views needed on 24 hour sets but I do not think that it should be using a complicated per hour thing. Stick to the old WP:KISS principle. The C of E God Save the Queen! (talk) 19:43, 17 December 2020 (UTC)

There's nothing complicated about views per hour. If you look at the DYKSTATS page, we've done the calculation for every hook in November and December. But let's consider your proposal: If 5,000 total views was the right threshold for a 6-hour queue, what do you suggest as the threshold for a 24-hour queue? Cbl62 (talk) 20:06, 17 December 2020 (UTC)

I'd say 10,000 for 24 would be about right and keep 5,000 for 12. Only a brief notion as it's a bit late tonight but its a start. The C of E God Save the Queen! (talk) 23:23, 17 December 2020 (UTC)

In the end, that's not so different than what we've recently implemented which equates to 6,000 for 12 hours and 12,000 for 24 hours. Cbl62 (talk) 23:30, 17 December 2020 (UTC)

If that's the case, why not just make it clear and simple by just saying the total views have increased rather than saying it is now "views per hour"? But there is something I have discovered that makes the original point about the 5,000 being only good for 6 hours moot in my view. People are using Wikipedia less. The fact is there are less views to go round so maybe the 5,000 isn't that far fetched in being retained at least for the 12 hour sets but I would be happy for an increase in 24 hours (even though not every 24 hour hook gets 5,000). The C of E God Save the Queen! (talk) 06:41, 18 December 2020 (UTC)

• So the record is clear, the decision to change the threshold from 5,000 total views to 500 views per hour was not implemented on a "whim". It was proposed here (Wikipedia talk:Did you know/Statistics#Adjust threshold?) and noticed to the wider community at Wikipedia talk:Did you know/Archive 173#Adjust Statistics threshold? Of course, consensus can change, but let's not have people asserting that the change was made on a whim. Cbl62 (talk) 18:56, 17 December 2020 (UTC)

My understanding is that views-per-hour was adopted for the obvious reason that it's much easier to get 5000 pagehits over 24 hours than it is over 12, and that 5000 in 24 hours is too low a bar, resulting in too many so-so hooks getting into the DYKSTATS table. Also, that the tables were getting too long with too many entries.

Now with regard to TheCofE's suggestion that we just made it a 10k threshold for 24 hours instead, the problem with that is that it still doesn't address the issue of hooks which remain on display for non-standard lengths of time. The most obvious example is when the bot goes down and hooks remain on display for 14, 17 or however many hours - with the following sets being correspondingly reduced in the number of hours they are on display. The same problem arises when manual updates are made late, and also when we run hooks for longer or shorter periods - like 8 hours, for example, or (rarely) 48 hours.

In short, views per hour is the only way to ensure that everybody's hooks are rated by the same yardstick. So I think it's clear that the move to views per hour is appropriate, and that correspondingly, hooks should be listed in order of highest to lowest views per hour and the VPH column should take priority.

The real question then, is what should the views per hour threshold be? It's awkward, because 5000 pagehits in 12 hours, which has long been our yardstick, calculates to approximately 416.7 views per hour, which is an oddball and counterintuitive number. So it was recently suggested that that threshold should be raised to 500 views per hour, which corresponds to 6000 pagehits in 12 hours.

Now in looking at the tables for some previous months, a very substantial chunk of entries are for hooks that got way, way below 416.7 VPH - entries for little more than half that are common, which of course are all the hooks which made the 5000 pageviews threshold over 24 hours. So just going to 416.7 VPH would eliminate all those entries. On the other hand, if you set the threshold at 500 VPH, while you will still be getting 30 or more hooks in the table for some months, others will have as few as 8 or 10 entries for the entire month, which isn't many and would disenfranchise many contributors. So I have my doubts that the threshold should be set that high.

Another alternative would be to decide on a set number of hooks for each table, and then just promote the top x number of hooks, for example, have a top 30 for the month and just include the top 30. Or a top 10, or 20, or 25, or whatever. But in the meantime, all things considered I'm leaning toward just leaving the threshold at 416.7 for the time being unless or until a broader consensus is reached, counterintuitive though such a number is. Gatoclass (talk) 15:18, 19 December 2020 (UTC)

Those irregular lengths we can decide. If it's up to 1 hours extra then leave as is but for longer, increase exponentially in relation to what the additional time beyond the standard was. But again, that is getting into more complex calculations which I want to avoid.

If you had said, "up it 6,000 total" rather than "we're now 500 vph", I don't think people would have been too irritated by the change (though probably that too should have been a bit more formal with the motion). As I said, I'm open to increasing the total views required because that is still the straightforward "click check views and you have your answer" that we're accustomed to rather than "click check views then get the calculator out". It's much easier for everyone if its a simple yes/no to "has it hit the target"?. I agree with you we should leave it at 5,000 for the moment until such time as there is formal conesnsus to change. The C of E God Save the Queen! (talk) 15:54, 19 December 2020 (UTC)

I don't see any indication here that there is more than one editor irritated by using VPH instead of a absolute number. This method is quite simple because it requires one straightforward calculation (views/hours) that anyone can do for any hook, and without having to know the threshold for different run times. We just need to settle on the rate, perhaps 450 VPH. MB 17:11, 19 December 2020 (UTC)

Not really, because the stats tool only give you the total and that's all you need to make a decision at a quick glance. We should be using totals, not vph. And for overruns, we are talking about a tiny about on the rare occasion the bot fails so honestly I wouldn't have special rules for it because we don't want WP:RULECREEP. The C of E God Save the Queen! (talk) 18:31, 19 December 2020 (UTC)

The stat tool in the link gives you page views for a two-week period. You have to hover over the correct day to see the number, sometimes is it spread across two days requiring you to add them together. Then you still have to know the run time (usually 12 or 24 hours), and the corresponding threshold. Only then can you know if the hook should be listed here. I fail to see how a trivial calculation significantly complicates this process. MB 00:38, 20 December 2020 (UTC)

And that isn't complicated compared with a simple hover over the day it ran and see what the results are? The C of E God Save the Queen! (talk) 15:20, 20 December 2020 (UTC)

There are two parts to this.

(1) Using views per hour as a threshold for inclusion. The current treshhold for inclusion per the prior properly-noticed discussion is 500 views per hour. Indeed, this proposal was passed without objection during its one-week run. As a result, a hook needs to receive 6,000 total views for a 12-hour queue and 12,000 views for a 24-hour queue. This is a negligible difference (16.6%) from the 5,000/10,000 thresholds that User:The C of E suggested. With respect to the claim of burden, nobody is required to do the views per hour calculation if they don't want to. The calculation is really not complicated, as I did it for hundreds of hooks in November and December and did not find it difficult at all. And it can be applied smoothly should we decide at some point to revert to six- or eight-hour queues, as we've done in the past.

(2) Sorting by views per hour. There is a secondary issue as to whether the monthly list should be sorted by views per hour or by total views. As noted in the discussion above, I believed that by switching to views per hours as the threshold for inclusion that we would likewise use this qualifying criterion as the basis for sorting of hooks and began doing so. User:Yoninah objected to this, and after her objection, I promptly reverted to sorting by total views. I continue to believe that sorting by views per hours gives us a truer read on which hooks had the best performance, i.e., a hook that runs for 12 hours and receives 15,000 views has performed much more impressively than a hook that ran for 24 hours and received 18,000 views. But that change has been tabled. I will not change to a views per hour sorting methodology without a discussion brought to the attention of the wider DYK community. Cbl62 (talk) 16:45, 20 December 2020 (UTC)

Cbl62, There is no consensus at all for 500 VPH in the discussions you listed above. In fact, there is only one person in that discussion who recommends 500, and that was Yoninah - although she now appears to be having second thoughts herself - and two who oppose it (The C of E and Gog the Mild). All the other contributors - and there are only a couple - express no preference.

So again, I think you need to set it back to 416.7 per hour - that is, 5000 per 12 hours, our longstanding yardstick - unless or until a new consensus emerges. Gatoclass (talk) 18:22, 20 December 2020 (UTC)

@Gatoclass: Are you suggesting we go to 416.7 per hour (i.e., 5,000 views for a 12-hour set and 10,000 for a 24-hour set) or 5,000 total hits (regardless of queue length? Cbl62 (talk) 18:41, 20 December 2020 (UTC)

416.7 per hour - I thought I'd made that clear in my longer post the other day. Gatoclass (talk) 18:48, 20 December 2020 (UTC)

Gato -- I had not seen your earlier post when I logged on this morning. I think your suggestion is a good compromise. My goal here has been to deal with the fact that 24-hour queues often result in the majority of hooks garnering 5,000 hits, such that DYKSTATS had become bloated, burdensome to maintain, and unexceptional. The compromise largely alleviates that problem, and I am satisfied with it. Someone will need to adjust the standard notice left on users' talk pages to reflect this change. Cbl62 (talk) 19:11, 20 December 2020 (UTC)

 Done, thanks.[1] Gatoclass (talk) 19:19, 20 December 2020 (UTC)

416.7 per hour is near enough to what I wanted. I am happy to support this as the requirement to get on the statistics list. Gog the Mild (talk) 19:28, 20 December 2020 (UTC)

Hi, sorry I wasn't able to participate in this discussion. I am happy with the outcome because I was worried that telling people they have to get 416.7 pageviews per hour to qualify is a very strange and unmemorable figure. But telling them they need 5,000 pageviews in a 12-hour slot or 10,000 pageviews in a 24-hour run, which averages out to 416.7 pageviews per hour, makes sense.

I would like to know, now, if we're still promoting hooks to the "Over 25,000 pageviews for a lead hook" and "Over 15,000 pageviews for a non-lead hook"? I notice that in December, Lewis Strauss got over 18,000 pageviews. But that was in a 24-hour period. Will we be starting new rules for inclusion on the all-time pageviews charts on January 1? Yoninah (talk) 01:04, 25 December 2020 (UTC)

I would support a proposal to tighten the all-time lists, but no changes have been made at this point. Cbl62 (talk) 17:15, 25 December 2020 (UTC)

Adjust all-time lists

Cbl62, Yoninah, the all-time lists still haven't been adjusted appropriately. I would suggest changing the all-time leaders criterion from the current 25k, to 2000 views per hour. This equates to 24k views in 12 hours, which is very similar to the current 25k cut-off.

Correspondingly, the all-time non-lead hooks list criterion should be altered to 1,250 views per hour, which just happens to equate exactly with the current threshold of 15k views total. Gatoclass (talk) 14:02, 18 January 2021 (UTC)

No need for that. They met the criteria at the time so no sense in removing them ex post facto. Just put a note saying "Before December 2020, the rule was....". The C of E God Save the Queen! (talk) 14:14, 18 January 2021 (UTC)

If they don't meet the criteria, they should be removed. Both lists are ridiculously long anyway and contain many hooks which clearly were not exceptional. We don't need to clutter up the lists with "legacy" entries that just happened to meet the very sloppy criteria at an earlier time; that would just make for a confusing and inequitable mess. Gatoclass (talk) 16:43, 18 January 2021 (UTC)

Except they did meet the criteria as it was at the time. We should not be engaging in revisionism. Those were the rules up until very recently so they should remain but with a note explaining historical context if needs be. The C of E God Save the Queen! (talk) 17:03, 18 January 2021 (UTC)

The original rules were poorly thought out as they didn't take account of varying exposure times; when they were first proposed, having the same hookset on the main page for a full 24 hours was virtually unheard of, as in the early days we usually ran three or sometimes even four sets per day. The system simply doesn't work now that we have 24-hour sets; that's been the entire point of the above discussion, that the original system just isn't stringent enough for 24-hour sets, and never has been. Many of these hooks simply should never have been added to all-time best lists, they only got there because of outdated criteria, and it's well past time to fix the issue. Gatoclass (talk) 17:39, 18 January 2021 (UTC)

Equally some on there got on there back on the even older 8 hour set and yet they'd be binned if we went by what you propose. You can't just start cherry picking or start stripping out with a new arbitrary number. These are here as historical reference and we should not be changed based on our new standards. I'd say the ones that are there are to be grandfathered in and future ones be on the new standards but put a simple note (as I said) saying pre-December 2020 use the old standard. The C of E God Save the Queen! (talk) 17:46, 18 January 2021 (UTC)

The lists were developed to recognize the "best of the best". It is entirely natural and appropriate that the threshold for all-time best lists be adjusted over time. When we started these lists, we had thresholds of 11,000 and 20,000 views. When the lists became overcrowded, we adjusted the thresholds upward and pruned (or "binned") the hooks that did not qualify. It is long past time to do this once again. I support the proposal to once again adjust the threshold. Cbl62 (talk) 18:28, 18 January 2021 (UTC)

For editors who may be unhappy about their prior hooks being "pruned", we could preserve the full, old lists in an archive, but display only the pruned list on the main DYKSTATS page. Cbl62 (talk) 18:47, 18 January 2021 (UTC)

Well we do have a version of that for WP:DYK/HoF. If push comes to shove, yes, I'd be ok with it (though I'd still prefer the retention of them on the main page) providing what is left on the main page is done on total views not the silly complicated vph. I just think because we had Lord's Honours Boards for example which got over 25,000 back in the 8 hour days (which frankly is much more impressive). The C of E God Save the Queen! (talk) 19:16, 18 January 2021 (UTC)

I agree that Lord's honours boards performed remarkably and far more impressively than a hook that garnered 30,000 views (or even 40,000 views) during a 24-hour queue. For this reason, Lord's honours boards has rightfully earned a spot on the proposed all-time list referenced below and shows why the vph approach is the most effective method to use in ranking the performance of hooks over time. There is nothing "silly" or "complicated" about this approach. It involves simple arithmetic and allows us to compare performance of hooks that were on the Main Page for drastically different periods of time. Cbl62 (talk) 00:51, 19 January 2021 (UTC)

• I prepared a mock-up of a list of hooks that have received at least 2,500 views per hour (vph). It can be viewed here: User:Cbl62/All-time DYK. Gato had suggested a 2,000-view threshold (which could be done). I continue to believe that vph is the most effective way to depict our most successful hooks. It is the only effective way we can calibrate and meaningfully rank performance between and among hooks that were on the Main Page for drastically differing period of time ranging from 24 to 6 hours. For those interested in total views, the information is included and can be sorted in that manner. I tried a number of different ways to sort and IMO this is the best way to depict our all-time most successful hooks. Cbl62 (talk) 00:46, 19 January 2021 (UTC)
  • I don't think its a good idea to merge the pic and non-pic hooks. We already have different criteria currently for the reason that the picture hook tends to get more clicks. IF (and I must say if) we are going to do this (and I still don't really see a need for it), keep the picture and non-picture hooks with separate targets with a lower amount needed for the non-picture hooks. The C of E God Save the Queen! (talk) 08:05, 19 January 2021 (UTC)

Would be good to get input from Gatoclass, Yoninah, and others. I'm leaning about 55-45 in favor of merging the lead and non-lead lists. The rationale for merging is simplicity and greater ease in viewing the best hooks in one place. When you look at it, and to my surprise, non-lead hooks compare pretty favorably, holding six of the top ten and nine of the top 20 spots (including Nos. 1, 3, 6, 7, 8, 10, 12, 13 and 18).Cbl62 (talk) 08:12, 19 January 2021 (UTC)

Another approach would be to have separate lists of the top 50 lead hooks and the top 50 non-lead hooks. If done that way, the lists would automatically cull as new hook were added to the lists. Thoughts? Cbl62 (talk) 08:32, 19 January 2021 (UTC)

Not a big fan of cut offs (as you probably guessed) because that can just disincentivise people to write interesting hooks, especially as the bar would get higher and higher each time and it will get to a point where it's almost nigh on impossible to climb. Keep a set minimum please. The C of E God Save the Queen! (talk) 08:38, 19 January 2021 (UTC)

A top 100 of all-time most popular hooks, both lead and non-lead, sounds like a good idea to me. You could then still maintain separate lists of lead and non-lead hooks which met a certain criteria, like more than 20k or 25k for leads and more than 1250 for non-leads and so on. It would mean a fair bit of redundancy though, as hooks in the all-time best list would also appear in the other lists. But I guess you could avoid that by making them lists of "Other lead hooks which got more than 25k views per hour" and "Other non-lead hooks which got more than 1,250 views per hour" or somesuch. Gatoclass (talk) 20:35, 20 January 2021 (UTC)

I also like the idea of a top 100 all-time most popular hooks list, and User:Cbl62/All-time DYK gives us the data we need to get it started. The question remains as to what should then be done with the very long existing lists. Options include: (a) move them to an archive where they can be preserved, (b) keep them but modify the thresholds to 2500 vph for lead and 1500 vph for non-lead hooks, or (c) keep them as is. My preference is option (a), but option (b) is reasonable as well. Thoughts? Cbl62 (talk) 16:16, 4 February 2021 (UTC)

• I'm late to this discussion, but I agree that we ought to have the lists be based on a per hour threshold, not a total threshold, as that disadvantages 12h hooks. I don't mind if the adjustment results in some trimming, as this page has gotten quite long. {{u|Sdkb}} ^talk 08:41, 23 August 2021 (UTC)

Adjust threshold?

When DYK STATISTICS began back in 2008, I selected the threshold of 5,000 hits because it seemingly represented a very high level of performance by a hook. At that time, however, we were running four queues a day, and 5,000 hits represented 833 hits per hour. Now that we are running one or two queues per day, the relative value of 5,000 hits has been depreciated. For a 24-hour cue, a hook qualifies for this page even when it garners only 208 hooks per hour.

As a result, we now see queues where the majority of the hooks qualify for inclusion here. On November 19, for example, five of the eight hooks (62.5%) garnered > 5,000 hits. It seems that an upward adjustment in the threshold might be appropriate. Or has the "5,000" mark has become so entrenched that there might be resistance to change?

My suggestion would be to limit this page, starting in December, to hooks that generate at least 400 (or even 500) hits per hour. This would restore things to the original intent of recording extraordinary hooks. What do folks think? Cbl62 (talk) 21:13, 20 November 2020 (UTC)

• As we switch between 12 and 24 hour sets, we should do it on views/hr rather than total hits (which biases towards the 24hr sets). 400-500 views an hour seems reasonable benchmark (for 12 hour set, this is comparable to 5k benchmark). Joseph2302 (talk) 02:52, 21 November 2020 (UTC)
• I'd support going to a per-hour measure. —valereee (talk) 13:47, 21 November 2020 (UTC)
• Thank you for bringing this up. I also notice that 5,000 hits has become a ridiculously low benchmark. I support listing pageviews per hour, and 500 seems like a good measure. Yoninah (talk) 16:46, 21 November 2020 (UTC)
• Support switching to a per hour basis. 500 views per hour seems high, 12,000 views in 24 hours; not that much less than what is required to get on the all-time list. Unless the all-time list was similarly tweaked - which I would support - say to 750 views per hour? Gog the Mild (talk) 17:16, 21 November 2020 (UTC)

I'd be fine with adjusting the all-time list, but that will require some careful analysis to figure out the right level. (My rough guess is that it would need to be in the 2,000/hr range, as 750 hits an hour was extremely common in the old days. Another possibility, in order to adjust for fluctuation in DYK views over the years, would be to feature the Top 25 hooks of each year based on hourly views.) Anyway, that's a separate discussion and shouldn't IMO sidetrack the current proposal. Cbl62 (talk) 01:34, 24 November 2020 (UTC)

I created a Top 25 list for 2008 (see Best of 2008), which I could prepare for each year. There were 25 hooks that year alone with > 2,850 views per hour -- which confirms that a threshold of 750 views/hour would be way too low for an all-time list. Cbl62 (talk) 03:02, 24 November 2020 (UTC)

• I think consensus supports the move. I've started the December table limiting it to hooks with at least 400 views. I'd be fine with going to 500 but wasn't certain whether consensus supported it. We can shift upward if this still proves to be excessive. Cbl62 (talk) 09:28, 3 December 2020 (UTC)
• Following discussion with Yoninah, I believe I erred in choosing the 400 views level rather than 500. Yoninah believes consensus supported 500 and I agree. I went with 400 to be cautious. However, the 400 views level results in an anomaly in that it actually allows hooks with 4,800 views (12 hours x 400) to qualify. The intent was to tighten, not loosen, the threshold. Accordingly, unless there is objection, I will go ahead and adjust the new hourly limit to 500 views per hours. Cbl62 (talk) 18:10, 7 December 2020 (UTC)

• I support the 500 views per hour threshold, as in keeping with what we've been doing at DYKSTATS for years. Yoninah (talk) 11:48, 9 December 2020 (UTC)

• Oppose Only just seen this. I object to use changing this without a proper full scale discussion, let alone in the middle of the month. Also who is to say that it has to be one particular hour gets it over the line or spread across evenly all the time it was there? It just complicates things unnecessarily. The C of E God Save the Queen! (talk) 16:46, 11 December 2020 (UTC)

  @The C of E: This question was bothering me to, so to anwer it (a little late, to be sure), I pulled the data from the ten days of two-a-day sets from October 2021—the difference was, on average, 600 views a hook more for the ones that ran from 12:00—00:00 UTC. It doesn't seem to be a huge difference. theleekycauldron (talk • contribs) (they/them) 22:04, 5 November 2021 (UTC)

November?

We're well into the month... is there a reason there's no November yet? Should I just make it? Maury Markowitz (talk) 20:06, 5 November 2021 (UTC)

yeah—i made most of the october list, i just never got around to starting november. I'd bet my bottom dollar the september list is incomplete, by the way theleekycauldron (talk • contribs) (they/them) 21:39, 5 November 2021 (UTC)

@Maury Markowitz: if you want to speed up the process slightly, i wrote a script that adds a "Get views!" button to every WP:Recent additions page and archive. You have to open up the console before you click it, or else it's a little laggy, but it'll spit the pageviews for every hook out to console. theleekycauldron (talk • contribs) (they/them) 21:58, 5 November 2021 (UTC)

Out of curiosity, how do you know if you were in a 12 or 24 hour queue? The notice on your talk page doesn't say. Maury Markowitz (talk) 22:17, 5 November 2021 (UTC)

You can check WP:Recent additions and do the math, I suppose—but other than that, there isn't a great way to check, yeah. Should we add that into the credit message? theleekycauldron (talk • contribs) (they/them) 22:20, 5 November 2021 (UTC)

Maybe... I did my entry manually so it's a bit of a pain, but if the calculation is going to be automated I'm not sure it's required. Maury Markowitz (talk) 14:06, 6 November 2021 (UTC)

Are these generated automatically?

I saw that a couple of mine have had quite a lot of views, so I was thinking of adding them to the table, but I don't know how this is generated (it seems like a waste to put it in if it's just going to be overwritten by a bot later). jp×g 12:26, 11 December 2021 (UTC)

Hi, JPxG! I semi-automatically generate the monthly tables with a script i wrote, so depending on when the hook ran, it'd probably be overwritten—it's possible there's an error with the script, or you might be talking about the all-time pageview leaders, which I've yet to overwrite. Which hooks specifically were you thinking of? theleekycauldron (talk • contribs) (they/she) 17:48, 11 December 2021 (UTC)

Oh, it's fine -- I was just looking at a couple of my own DYK pageview counts and it occurred to me that there might be some onus upon the page author to put it in the list (there's a list of Wikipedians by DYK count somewhere, which does work that way). If it's automatic, I won't bother myself with it. jp×g 00:34, 12 December 2021 (UTC)

Lowest performing hooks, part 2

Theleekycauldron's fabulous script allows us to track the highest and lowest performing hooks more easily. I've long thought we can learn as much from what doesn't work as we can from what does work. With that in mind, listed below are the lowest performing hooks of the last 12 months (each drawing fewer than 50 hits per hour while being featured on the Main Page). They are listed not for purposes of criticism, but to hopefully draw lessons to help in constructing better hooks. Feel free to add your comments about what went right or wrong with any of these hooks. Cbl62 (talk) 20:13, 23 August 2022 (UTC)

1 ... that Gian Marco's Días Nuevos, the best-selling album of 2011 in Perú, was the first quintuple platinum awarded by Unión Peruana de Productores Fonográficos? (25.0 hits per hour while on the Main Page)

What went wrong?

My theory: The hook is not sufficiently focused on the subject of the article. Instead of offering some interesting tidbit of information about the Unión Peruana (what it is, what they do), the focus of the hook is on a particular artist's album. It doesn't make the reader want to click and learn more about the Unión Peruana. Lesson: Hooks should make us want to learn more about the main subject. Cbl62 (talk) 23:35, 23 August 2022 (UTC)

2 ... that in a choral tour program titled Salmo!, Bach's 18th successor first conducted Salmo 150, an a cappella setting of Psalm 150 by the Brazilian composer Ernani Aguiar? (29.2/hr)

What went wrong?

The Easter egg link to Andreas Reize was more enticing and stole the thunder with 1,429 hits. Cbl62 (talk) 19:15, 22 August 2022 (UTC)

This went right. The bold article was a vehicle to mention the concert, conductor and choir. Reviewers didn't want the choir, so the conductor. Worked as designed. Would have been better with the image. --Gerda Arendt (talk) 19:22, 22 August 2022 (UTC)

"Would have been better with an image" isn't useful advice though, because every hook does better if placed first and has an image. We can't have every hook be the first one, so that means optimizing the pure text case, too. SnowFire (talk) 00:01, 24 August 2022 (UTC)

The thing is, this hook is about Salmo 150 and not Reize. So the goal was to bring in more views to Salmo 150. If Reize got more hits than the subject, then that's actually a failure not "working as designed". Narutolovehinata5 (talk · contributions) 14:14, 24 August 2022 (UTC)

As for the hook, I feel that this would have done better if the proposal about the work supposedly being the most performed Brazilian chorale work in the US was used instead. Narutolovehinata5 (talk · contributions) 14:14, 24 August 2022 (UTC)

It worked as I designed it. I wasn't interested in Salmo - nice 2 minutes piece, but not outstanding. I was interested in spreading the news about that whole outstanding concert! --Gerda Arendt (talk) 21:29, 24 August 2022 (UTC)

3 ... that thanks to the advocate general's opinion, one is likely to know the outcome of a court case before the European Court of Justice before it even starts drafting the ruling? (29.5/hr)

What went wrong?

My theory? The hook is confusing ... and the double use of "before" adds to that confusion. I ended up shrugging my shoulders as to what the hook means, but not in a way that enticed me to want to click and learn more. Cbl62 (talk)

4 ... that George Balanchine choreographed his ballet Sonatine for New York City Ballet's Ravel Festival, a tribute to its composer in his centenary year? (32.5/hr)

What went wrong?

I'm not sure. The hook is composed well and doesn't seem to have any glaring deficiencies. On the other hand, we've had many Balanchine hooks perform poorly. It may be that there's just not broad interest in Balanchine's ballets. Or perhaps readers have become weary from so many hooks on Balanchine's works? 19:15, 22 August 2022 (UTC)

I imagine if a reader is unfamiliar with Balanchine, they're not really going to understand the significance of the centenary. Narutolovehinata5 (talk · contributions) 14:14, 24 August 2022 (UTC)

5 ... that professional golfer Cary Middlecoff co-owned Memphis, Tennessee, radio station WCBR in the mid-1950s? (33.5/hr)

What went wrong?

The Middlecoff link diverted some hits, but I suspect the real problem was that the hook was just not that interesting. Likely reactions include, "So what?" or "Who's Cary Midddlecoff?" Cbl62 (talk) 19:15, 22 August 2022 (UTC)

@Cbl62: I'll also point out that Sammi Brie's radio hooks tend to suffer a surprising amount when she includes the station callsign. Maybe a tighter/punchier hook wouldn't've been so far down, but you're right that this particular hook wasn't very interesting to begin with. theleekycauldron (talk • contribs) (she/they) 23:20, 23 August 2022 (UTC)

This was one of the last call sign hooks to go through the system, Cbl62. I now only use call sign hooks when absolutely necessary to avoid country confusion (for instance, I have a radio station in Georgia in a queue and don't want people thinking it's the country) or when the call sign lends itself to a quirky (KOOL, KECH). I really only did this article to "finish" WLOK—its history intersects at a pivotal point. Sammi Brie (she/her • t • c) 23:35, 23 August 2022 (UTC)

Yeah, I noticed that after Sammi started to hide the call sign, her hooks started to perform a bit better (for example, WBKO). Epicgenius (talk) 15:25, 24 August 2022 (UTC)

6 ... that when Josef Rheinberger conducted the first performance of his Mass in A major for three women's voices on Christmas Eve 1881, he added a flute and a string quintet to the organ accompaniment? (35.6/hr)

What went wrong?

Unclear. The Rheinberger link diverted some hits, but the reality is that neither link attracted a lot of traffic. It was featured on Christmas Eve, so that may be part of the issue. Cbl62 (talk) 19:15, 22 August 2022 (UTC)

The hook was probably too long. A much snappier hook that omitted the "three women's voices" and "on Christmas Eve 1881" part may have attracted more attention. I'm not sure how much it would have helped, but in this case I think a shorter hook could have helped retain interest more than this relatively long and winded one. Perhaps "... that Mass in A major by Josef Rheinberger has a flute and a string quintet accompanying an organ?" would have been a snappier option here? Narutolovehinata5 (talk · contributions) 14:21, 24 August 2022 (UTC)

7 ... that George Balanchine choreographed his ballet Scherzo à la Russe for the New York City Ballet's Stravinsky Festival, a tribute to the composer after his death? (38.4/hr)

What went wrong?

More Blanchine. See my thoughts on No. 4. Cbl62 (talk) 19:15, 22 August 2022 (UTC)

8 ... that although George Balanchine set Raymonda Variations to musical excerpts from the ballet Raymonda, his new ballet does not follow the plot of the original? (39.3/hr)

What went wrong?

See my thoughts on No. 4. Three of the ten lowest performing hooks are Balanchine ballets. Cbl62 (talk) 19:15, 22 August 2022 (UTC)

9 ... that Red Blanchard, the owner of Iowa radio station KSMN, commuted 800 miles (1280 km) by plane from Mason City each week to host a radio show in Chicago? (39.7/hr)

What went wrong?

My theory: The hook is about Blanchard, and the radio station is peripheral to the hook. For this reason, it is not surprising that the Blanchard link attracted twice as many hits as the target article on the radio station. Cbl62 (talk) 20:07, 22 August 2022 (UTC)

10 ... that Johann Reinhold Forster's 1778 book Observations Made During a Voyage Round the World has been described as "the beginning of modern geography"? (40.1/hr)

What went wrong?

Feels like a decent hook so why it underperformed is a bit of a mystery to me. Maybe it should have mentioned the bolded link first? Narutolovehinata5 (talk · contributions) 14:14, 24 August 2022 (UTC)

This is one of mine, part of a project to cover all of the books that came out of Cook's three voyages. None of them did brilliantly, but the only one mentioning Cook (Characteres generum plantarum) did better, maybe because the hook did actually mention Cook? Perhaps "1778" is also a turn-off. —Kusma (talk) 21:11, 24 August 2022 (UTC)

In my opinion, "1778" was likely not the problem, since there have been plenty of popular hooks that mentioned the 18th century (Economy coffin, for instance, ran in August 2021 and was viewed 25K times, despite mentioning the year 1784). Like you said, the hook didn't mention Cook, who is more well known than Forster, but I don't think that this alone should have caused such a drastic drop in the number of people reading the article. The quote "the beginning of modern geography" is also pretty interesting at least to me. I think this may have been pure bad luck. – Epicgenius (talk) 21:37, 24 August 2022 (UTC)

11 ... that although it spent only one week on the Hot Rock & Alternative Songs chart, "Visitor" was the fifth Adult Alternative Airplay number-one for Of Monsters and Men? (40.8/hour)

What went wrong?

My theory: Maybe it's just not that hooky to note that a song did ok briefly on a couple of the alternative charts. A hook referencing the song's interesting take on alienation and being able to observe the real world without any connection to it might have drawn more hits. Cbl62 (talk)

12 ... that John Wesley Shipp, who played the Flash in the 1990 television series, was cast as his father in the pilot episode of 2014's The Flash? (41.9/hour)

What went wrong?

The link to the new article was massively eclipsed by more enticing links: John Wesley Shipp (4,965 hits) and The Flash (1990 TV series) (2,642). Cbl62 (talk) 20:07, 22 August 2022 (UTC)

Do you think the hook would have worked out better if the bolded link was mentioned earlier in the hook? Narutolovehinata5 (talk · contributions) 14:14, 24 August 2022 (UTC)

Personally, I think so. The fact that Shipp was the Flash in 1990 and plays the Flash's father in 2014 would probably be more broadly interesting if the bold link were within the first few words. Maybe fewer links would've helped as well, so something like "...that in the pilot episode of 2014's The Flash, the Flash's father was portrayed by the same actor as the Flash in the 1990 television series?" may have performed much better. With so many links in this hook (5 non-bold links), I personally believe that this is why Pilot (The Flash) didn't get as many views as it should have. – Epicgenius (talk) 15:18, 24 August 2022 (UTC)

13 ... that in the television series sequel Imortal (2010), Angel Locsin portrayed the lead role as the daughter of her lycan character in the Lobo TV series? (42.3/hour)

What went wrong?

The article link was buried. Numerous other interesting links stole the show: lycan (3,135 hits), Angel Locsin (2,798 hits), Lobo (TV series) (1,863), Imortal (1,862). If we want to feed attention to the featured content, we can't bury it in a sea of links. Cbl62 (talk) 19:57, 22 August 2022 (UTC)

I agree (see my thoughts on number 12). The fewer links, the better - we don't need to get rid of all of the non-bolded links, but there were four of them in this hook, which drew readers' attention away from the main subject, Angel Locsin filmography. Epicgenius (talk) 15:25, 24 August 2022 (UTC)

14 ... that Alfred Koerppen, who taught music theory and composition at the Musikhochschule Hannover, wrote the text and music of a 1951 opera after Virgil? (42.5)

What went wrong?

My guess that in this case, the mention of him teaching music theory at a school (whose name is written in German) may have turned away readers rather than enticing them. I'm not sure how much it would have helped, but I feel that the hook would have done better if the teaching part was dropped and the hook focused solely on the Virgil connection, perhaps even mentioning exactly which Virgil work was the inspiration. Narutolovehinata5 (talk · contributions) 14:14, 24 August 2022 (UTC)

15 ... that a review by Alan Blyth of tenor Christoph Strehl's performances described him as singing "with a Wunderlich-like strength and beauty"? (42.5)

What went wrong?

My theory is that people may not have known who Blyth or Wunderlich were, even with the presence of these links. Unfortunately, I think several hooks here suffered from the same problems in that you need to click on links to understand the hook. Epicgenius (talk) 15:25, 24 August 2022 (UTC)

If you look at the nom you see that I didn't suggest Blyth, nor Wunderlich. --Gerda Arendt (talk) 20:59, 26 August 2022 (UTC)

16 ... that George Balanchine choreographed the ballet Ivesiana to music by Charles Ives, despite previously finding his works too complex for dance? (42.6/hour)

What went wrong?

A very well-composed hook IMO. Shrugging my shoulders as to why it did so poorly ... but see thoughts on No. 4. Cbl62 (talk) 19:23, 22 August 2022 (UTC)

17 ... that R. B. Schlather directed the world premiere of Hertzberg's chamber opera The Rose Elf in Brooklyn, as well as Cimarosa's L'Italiana in Londra and Puccini's Madama Butterfly at the Oper Frankfurt? (43.5/hour)

What went wrong?

My guess is that readers probably thought "so what?", and I assume maybe they weren't aware of how much of a big deal Oper Frankfurt is. Personally I think a hook about him being inspired to take up opera following a trip with his parents to a festival could have performed better. Narutolovehinata5 (talk · contributions) 14:14, 24 August 2022 (UTC)

18 ... that the street from which Mississippi radio station WMPR broadcasts was renamed in honor of the station's longtime owner and general manager, former politician Charles Evers? (43.7/hour)

What went wrong?

The hook creates more curiosity about Evers than the radio station. Evers did get more hits, but neither one drew particularly well. Cbl62 (talk) 20:07, 22 August 2022 (UTC)

Personally I think the hook would have done better if it mentioned that Evers was a former mayor. Granted, he was the mayor of a small community, but "Mayor" would have gotten at least some more attention than the admittedly vague "politician". Narutolovehinata5 (talk · contributions) 14:14, 24 August 2022 (UTC)

19 ... that "Brich an, du schönes Morgenlicht" (Break, you beautiful morning light) tells the shepherds in Bach's Christmas Oratorio not to be afraid? (44.3/hour)

What went wrong?

20 ... that in 2014 the European Theatre Convention started a programme known as Dialogue of Cultures to support exchanges with theatres in Ukraine and other Eastern European countries? (44.3/hour)

What went wrong?

Honestly I think it's a decent hook, I'm not sure why it didn't do better. Maybe main page readers just aren't into theater? Narutolovehinata5 (talk · contributions) 14:14, 24 August 2022 (UTC)

My thoughts about this one is that it explains everything and doesn't leave any questions. Had the hook been "... that in 2014 the European Theatre Convention started a programme known as Dialogue of Cultures?" I might have been inclined to click the link to find out more. It's certainly not the only hook where I've thought that, and I'm not saying the fact isn't interesting or worthwhile, it just doesn't inspire curiosity in the reader to learn more. --CSJJ104 (talk) 20:31, 26 August 2022 (UTC)

The idea of Ukraine a part of Europe in 2014 already was the motivation to write the article (not only the hook), and why not tell it those who read the Main page but will not click? --Gerda Arendt (talk) 21:04, 26 August 2022 (UTC)

I personally have no issue with the hook as it was, and if the intention was to inform readers of those facts then it works well, but this is a discussion about how to increase pageviews and I felt I should add my views, for whatever they are worth. --CSJJ104 (talk) 23:24, 26 August 2022 (UTC)

21 ... that after a career as political journalist, Peter Merseburger wrote biographies of Der Spiegel founder Rudolf Augstein and chancellor Willy Brandt? (45.0/hour)

What went wrong?

On the surface, it seems like a decent hook with no glaring issues. My theory is that it just wasn't "hooky" or surprising -- a German journalist wrote biographies of a German newspaper publisher and a former German chancellor. There's no twist or extraordinary accomplishment to elicit a "wow" reaction or otherwise draw the reader in. Cbl62 (talk) 20:13, 23 August 2022 (UTC)

Did you read the nomination? He was supposed to be described as good for controversial TV, or if not that at least the image of him and the chancellor. The controversial TV later appeared with Alice Schwarzer pictured, and faired well (Panorama, [2]), but again more clicks for the pictured woman than the TV program. Thank goodness Merseburger had gotten his views when on recent deaths weeks before. --Gerda Arendt (talk) 21:00, 23 August 2022 (UTC)

22 ... that Óscar Catacora directed Wiñaypacha (Eternity), the first Peruvian film in the Aymara language? (45.0/hour)

What went wrong?

Another decent hook that probably should have done better. My only guess is that, since it's about a Peruvian topic, maybe English readers simply weren't interested. Narutolovehinata5 (talk · contributions) 14:14, 24 August 2022 (UTC)

My theory is that the focus was on the film and not the director himself. In fact, the film got more pageviews than the director on that day. – Epicgenius (talk) 15:28, 24 August 2022 (UTC)

23 ... that even though plans to convert the studio building of station KITN in Olympia, Washington, into a courthouse were soon changed, county taxpayers still paid its moving expenses? (46.9/hour)

What went wrong?

It may have been that readers didn't understand the connection between county tax revenue and the failed plan to convert the building into a courthouse. Readers may have missed the fact that the courthouse would have been funded using county taxes or that the moving expenses of a private entity (in this case, a radio station) would normally be funded privately. In addition, as mentioned before, some radio station hooks may have performed poorly because their call signs are used. Epicgenius (talk) 14:37, 25 August 2022 (UTC)

24 ... that George Balanchine choreographed the ballet La Valse to Ravel's music of the same name, which was rejected by an impresario and had been used in unsuccessful ballets before? (47.3/hour)

What went wrong?

See my thoughts on No. 4. Five of the bottom 25 are on Balanchine ballets. Cbl62 (talk) 19:23, 22 August 2022 (UTC)

I think this hook may have been too reliant on "insider" information. I imagine most readers didn't know who Ravel is and what an impresario is, so rather than wanting to know more they just shrugged. Narutolovehinata5 (talk · contributions) 14:14, 24 August 2022 (UTC)

It's of little interest to someone not already familiar with ballet and/or the person in question. Hooks need to be understandable to a general audience, and this isn't. It uses too much jargon and relies heavily on the reader already having knowledge of the subject area. All factors that combined to set this up for especially low views. Trainsandotherthings (talk) 03:16, 1 September 2022 (UTC)

25 ... that Progress, composed by Winsford Devine and recorded by calypsonian King Austin in 1979, was declared "the song of the last millennium" by the Trinbago Unified Calypsonians' Organisation? (47.8/hour)

What went wrong?

Probably another case of having too many details in a hook. I think a hook that went like ... that Winsford Devine's composition Progress was declared "the song of the last millennium" by the Trinbago Unified Calypsonians' Organisation? could have performed better. By how much more? I'm not sure, but the hook is a bit on the longer side and I'm not sure if the extra detail was necessary in this case. Narutolovehinata5 (talk · contributions) 14:14, 24 August 2022 (UTC)

More stats

Since i have access to all the fancy-dancy tools and know my way around the stats pages, I'll do my best to honour any requests for data that are placed here :) theleekycauldron (talk • contribs) (she/they) 21:25, 24 August 2022 (UTC)

Worst-performing image hooks (requested by Z1720)

This'll comprise of the 20 worst-performing image hooks of 2022.

Article	Date	Image	views	vph	DYK hook
William Savage (ornithologist)	2022-08-02		2,154	179.5	... that American ornithologist William Savage painted pictures of birds (example pictured) that he hunted or received?
Glenwood Memorial Gardens	2022-02-05		2,142	178.5	... that Glenwood Memorial Gardens in Broomall, Pennsylvania, was initially established in 1849 as Glenwood Cemetery (pictured) in North Philadelphia and contained the graves of 702 Union soldiers?
Anaea (butterfly)	2022-04-16		2,137	178.1	... that butterflies of the genus Anaea (example pictured) are said to have "commanded the admiration of even the most gold-mad conquistadores"?
James A. Merriman	2022-03-14		2,081	173.4	... that James A. Merriman (pictured) was the only Black graduate from Rush Medical College in 1902 and the first African-American physician to practice medicine in Portland?
Indian cricket team in England in 1911	2022-01-05		2,036.5	169.7	... that the 19-year-old Maharaja Bhupinder Singh of Patiala (pictured) was the captain of the first all-Indian cricket team to tour England in 1911?
5th Combat Helicopter Regiment	2022-02-15		4,063	169.3	... that a Eurocopter Tiger (example pictured) from the French Army's 5th Combat Helicopter Regiment was involved in the 2019 Ménaka helicopter crash that killed 13 people?
Celmisia major	2022-06-05		1,855	154.6	... that the two varieties of Celmisia major (one variety pictured) are "certainly not" closely related?
Thomas Resetarits	2022-06-23		3,425	142.7	... that Burgenland Croat sculptor Thomas Resetarits created Stations of the Cross (example sculpture pictured)?
Merrimac Farm Wildlife Management Area	2022-04-11		1,568	130.7	... that Merrimac Farm Wildlife Management Area has one of the largest fields of Virginia bluebell flowers (pictured) in northern Virginia?
Climate change in Uganda	2022-07-03		2,993	124.7	... that the Rwenzori glaciers (examples pictured) are being melted by climate change in Uganda?
Italian Renaissance sculpture	2022-06-06		1,495	124.6	... that new forms appearing in 15th-century Italian Renaissance sculpture (example pictured) include the medal and the plaquette?
Laura J. Crossey	2022-08-07		2,967	123.6	... that Laura J. Crossey has shown that travertines (example pictured) are more likely to form when meteoric groundwater mixes with deeper groundwater from the Earth's mantle?
Gertrude Webster	2022-01-16		1,463.5	122.0	... that a cactus (pictured) is named after Gertrude Webster, who helped found the Desert Botanical Garden in Arizona?
Kyiv Symphony Orchestra	2022-05-30		2,798	116.6	... that the Kyiv Symphony Orchestra brought music by three Ukrainian composers to concert halls in Poland and Germany in April 2022, including the Berlin Philharmonie and the Kurhaus Wiesbaden (pictured)?
Cello Sonata (Debussy)	2022-08-23		2,779	115.8	... that Claude Debussy (pictured) described his Cello Sonata, composed within a few weeks in July 1915 at a Normandy seaside town, in a letter to his publisher Durand as of "almost classical form"?
Gian Franco Bottazzo	2022-08-22		2,706	112.8	... that Gian Franco Bottazzo (pictured) and Deborah Doniach proved that type 1 diabetes is an autoimmune disease?
Panorama (German TV program)	2022-04-01		2,329.5	97.1	... that Peter Merseburger refused to host a 1974 edition of German political TV magazine Panorama after a report by Alice Schwarzer (pictured) on an abortion was cancelled by authorities?
Natacha Aguilar de Soto	2022-07-17		2,264	94.3	... that research on short-finned pilot whales (example pictured) by Natacha Aguilar de Soto is leading scientists to reassess foraging models for the behavior of marine predators?
2022 British Indoor Athletics Championships	2022-04-09		1,108.5	92.4	... that at the 2022 British Indoor Athletics Championships, Lorraine Ugen (pictured) equalled the championship long jump record?
Messe brève no. 7 (Gounod)	2022-07-25		915	76.3	... that the Messe brève no. 7 by Charles Gounod (pictured) is an 1890 Missa brevis that he derived from an earlier work for only two voices and organ?

Notes

Some comments:

• I'm a bit surprised that the Uganda climate change article didn't do so well. Personally I would have thought the idea of Uganda having glaciers would have been surprising. Maybe a case of bad luck?
• Glenwood Memorial Gardens was an okay hook but may have meandered too much. The "initially established in 1849 as Glenwood Cemetery" part may have been unnecessary and probably just distracted from the main point.
• I'm a bit surprised at how much nature hooks are represented here. There are more nature-related hooks on that list than any other topic. Admittedly some of the hooks may not have been completely clear, such as the Celmisia major hook. Merrimac Farm Wildlife Management Area was probably a hook that wouldn't work without the image, so I don't think it's much of a surprise that it underperformed. It's not really a spectacular hook especially without the image.
• The next most represented subject on the list is classical music. Again, some of the hooks may have been too detailed for their own good. For example, the Cello Sonata hook probably did not need the mentions of July 1915 or Normandy. It probably would have done better if it focused solely on the "almost classical form" aspect. Messe's hook is okay but it's admittedly niche so I understand why it didn't do well. Looking at the article, nothing really stands out for hook purposes, so this may have been a case where it may have been better to simply not nominate Messe at all. Still, it honestly shocks me that an image hook couldn't even have 1,000 views during its run. For comparison, how did the other hooks in its set do?

Narutolovehinata5 (talk · contributions) 21:52, 24 August 2022 (UTC)

• 5 of the bottom 6 image hooks are images of not very well-known people (1 of them was mine). I guess the takeaway is that posting images of well-known people will help, but of relatively obscure people, it won't help get more views. Joseph2302 (talk) 15:26, 25 August 2022 (UTC)
• I notice that in 13 of the 20 hooks the picture is not of the article subject. Could it be they are driving interest towards articles other than what was intended? --CSJJ104 (talk) 01:27, 20 September 2022 (UTC)

General comments

There's a couple of questions I have before I propose theories:

• In what position were these hooks placed in? (second in the set, last in the set, etc.) Is there a correlation between the number of clicks and their position in the set?
• How many hooks/ALTs were proposed in the nomination?
• How long was the review in the nomination? Was there only one comment (approving the review) or was there a long discussion?
• Is there a correlation between the number of characters in a hook and the pageviews?
  • I remember doing some back-of-the-napkin math and finding little to no correlation on this... I can do some more rigorous math. theleekycauldron (talk • contribs) (she/they) 21:11, 24 August 2022 (UTC)
• Is success only defined by the pageviews of the bolded hook? Is a hook successful if the accompanying image or other wikilinks in the hook get a lot of hits?
  • We generally define it by pageviews of the bolded article; my personal rationale for upholding this (rather than including all links or the mediafile) is that the bolded article is the topic we want to educate people on. We want to teach people new things. theleekycauldron (talk • contribs) (she/they) 21:11, 24 August 2022 (UTC)
• Were these hooks on the main page for 12 hours or 24 hours? Does this factor into how many pageviews a hook receives per hour?
• Hooks with images and hooks in the final slot of the set tend to have more pageviews. What are the weakest performing of these hooks, and do they have anything in common with the above hooks?

While evaluating hooks on an individual basis is good, I think a more holistic approach is also needed in order to recommend best practices. Z1720 (talk) 00:26, 24 August 2022 (UTC)

• Based on the above, we can see that most of the entries belong to one of three topics: classical music, ballet, and radio stations. Now this does not mean that the topics are unsuitable for DYK, but it's clear that these topics tend to underperform views-wise and maybe some change in direction in hook-writing is needed. One possible piece of advice I could give is to paraphrase WP:DYKSG#C1: don't assume readers know what people or terms you're talking about. Make the hooks easy to understand, try to not make them reliant on too much "insider" information. Make the hooks hooky in such a way that readers will understand the hook even if they're unfamiliar with the names and terms mentioned. Make it so that, with the context readers know, they'll become interested in learning more. Niche topics are admittedly in a bit of a disadvantage compared to other, more popular topics, but this does not mean that these disadvantages are insurmountable. Narutolovehinata5 (talk · contributions) 14:14, 24 August 2022 (UTC)

• Two other recurring things I noticed on the list is that, 1. hooks with many links underperformed, and 2. hooks that seemed to have distracting details underperformed (such as the Flash and Trinidad hooks). This suggests that, in at least some cases, having too many links in a hook should be avoided. If it's absolutely necessary, it may be worth moving the subject to being the first linked hook so that it could at least attract more clicks. In addition, if a hook is too winding or has too many details that don't add to the hook fact, that might turn away readership instead of attracting it. Narutolovehinata5 (talk · contributions) 19:37, 24 August 2022 (UTC)

  Those seem like helpful take-aways. Cbl62 (talk) 19:51, 24 August 2022 (UTC)

  As a thought, is it worth considering only having the hooks subject linked, and nothing else? It's not hard to imagine a reader becoming frustrated that the selected link, e.g. Flash rather that the pilot episode does not contain the facts written in the hook, and there is no reason it has to, but the subject of the hook absolutely should contain all the relevant facts from the hook. Removing the other links should avoid this frustration, if it exists, while hopefully directing more readers towards the article we want to showcase. --CSJJ104 (talk) 20:46, 26 August 2022 (UTC)

• @Z1720: You raise some good questions, that warrant examination. And certainly, page views aren't the exclusive measure of success -- otherwise, we'd fill the queues with Nazis, sex acts, alien abductions, fart jokes, and profanity. That said, the Main Page (day or night, Tuesday or Saturday, slot 1 or slot 4) is a powerful tool for bringing attention to an article. A good hook will generate > 400 hits per hour (a great one > 1,000 hits per hour). When a hook makes it to the Main Page and generates only 25 or 30 hits per hour, it's extraordinary ... and worth taking another look to see what went wrong. Cbl62 (talk)
• As a general comment, I agree that "too many links" is absolutely an issue (and reviewers should aggressively pare out "dicdef"y extraneous links from hooks), but I'll also add that some other article drawing reader interest isn't all bad if the DYK'd article is particularly obscure / uninteresting. Granted, arguably these articles shouldn't even be proposed at DYK at all, but if we do them, sticking in some saucy other links that readers are actually interested in may be the lesser evil. (Readers being interested in the actor & the 1990 TV series for the "Flash" example seems entirely reasonable, for example.) SnowFire (talk) 20:57, 24 August 2022 (UTC)
• Something I've been turning over in the back of my head, but I realize would probably never fly, is performing A/B testing on our hooks. It would be simple, really: We use {{if}} and {{random}} to randomly serve one of two versions of a hook – we can measure, by linking to redirects instead of directly to the article, which performs better with readers. theleekycauldron (talk • contribs) (she/they) 21:28, 24 August 2022 (UTC)
• Interesting discussion. I had a look at my own worst-performing hook (from 2020), "..that with her victory at the 2001 Women's World Snooker Championship, Lisa Quick became the first person to have won world titles in both pool and snooker?" The hook's subject, 2001 Women's World Snooker Championship, got 362 hits, whilst Lisa Quick got 993. (Both were usually getting 6 or fewer hits a day.) Regards, BennyOnTheLoose (talk) 13:46, 25 August 2022 (UTC)
• I was rather surprised none of my train hooks showed up on here, I was expecting some to have quite low views. At least for me, I have a rule that if I don't think a given article has any kind of interesting hook, I simply don't nominate it. That's why I never nominated Boston and New York Air-Line Railroad or Pawtuxet Valley Railroad for DYK - I just couldn't come up with any hooks I felt would be interesting to a general audience. It can be hard to gauge what exactly counts as "interesting to a broad audience" but I try my best, as a subject matter expert, to remove everything I know about trains and then imagine if I'd still find the hook interesting or not. Since I started this practice, I've definitely seen my views on my hooks increase. Will expand on my thoughts later. Trainsandotherthings (talk) 17:42, 25 August 2022 (UTC)

  @Trainsandotherthings, lotta railfans out there, and it's not country-specific. Plus it looks like you're averaging less than 1 a month, which may mean the topic isn't oversaturated. Valereee (talk) 14:03, 1 September 2022 (UTC)

  Fair enough on the first point, but all my train related DYKs have been in the U.S. And yes it is true I don't usually do many DYKs at once, at least I don't think anyone has said "ugh, not another TAOT train DYK" so far. Trainsandotherthings (talk) 20:49, 1 September 2022 (UTC)

  You don't you think a railfan in another country is just as enamored of US trains as they are of trains in their own country? I'd assumed trains were like knitting or baking: if you're a Japanese rail enthusiast, you don't think "Oh, it's American trains? Nah." Great British Bakeoff is wildly popular here in the US. Valereee (talk) 22:12, 1 September 2022 (UTC)

• I think my main takeaway from all of this is that, while more page views are nice, the ultimate goal is simply to have your article be on DYK. That is already by an accomplishment itself. More pageviews is just a bonus. Still, we do want more people to learn more about a topic we care about (even if not necessarily caring about exact view figures), and so regardless of our thoughts on pageviews, the ideas expressed here can still serve as advice on how we can improve our own hooks. After all, all of us probably experienced at some point having a promoted hook that, in hindsight, we either weren't satisfied with or think shouldn't have been nominated at all. There's always room for more improvement, and knowing shortcomings can help us become better hook writers. Narutolovehinata5 (talk · contributions) 00:19, 26 August 2022 (UTC)
  • Yes, and to generalize wildly, even these "low" views will still represent several months' worth of normal views in a single day. For example, Unión Peruana de Productores Fonográficos, the #1 way above, got 25 views per hour. Now it averages 2 per day. Johnbod (talk) 14:09, 1 September 2022 (UTC)

• I think there's also the actual repetition of any subject. When we're serving up classical music or radio stations or whatever very regularly, we just aren't going to pique as much interest without a really top-notch hook.

In general I also think we should minimize unneeded links. The worst-performing image hook links the word organ, its fourth link. And, yes, when we do actually need additional links, if we can put the bolded link first, it'll help direct the reader to the target article. Valereee (talk) 13:56, 1 September 2022 (UTC)

Success for all-around information

As explained further up, I believe that we should not look at the bolded link. When Alice Schwarzer was pictured, of course she was clicked, 7k+. When Debussy was pictured (on his birthday), he of course was clicked, 16k+. For me that's success. --Gerda Arendt (talk) 21:41, 24 August 2022 (UTC)

I disagree – DYK's purpose is to showcase new and expanded content, not whatever we fancy. If a bolded article isn't interesting per se – if it merely acts as a vehicle in highlighting other, more well-known or desirable articles – it should step aside for a nomination that can stand on its own two feet, that can teach viewers something new about itself. theleekycauldron (talk • contribs) (she/they) 16:39, 25 August 2022 (UTC)

Agree with tlc here. The bolded article should be the most desirable click target. Articles that we don't want to showcase and drive traffic to shouldn't be at DYK. Not every new article is a good fit for DYK. —Kusma (talk) 09:12, 1 September 2022 (UTC)

We don't talk about the same thing. When I want to promote Thomanerchor but can't upgrade it to GA, why not use a piece they sing (of course an interesting one)? - Yesterday, we had a Bernstein composition, which faired reasonably, while the pictured composer would have made the stats, and that result was wanted. --Gerda Arendt (talk) 09:27, 1 September 2022 (UTC)

If you want to mention the Thomanerchor on the Main Page, I don't think using one of the thousands of pieces they sing makes a good DYK. Salmo 150 on DYK sent more views to Thomaskantor Andreas Reize than to the bolded article or the relevant composer. The article is about "the most frequently performed [in the US] choral work from Brazil", but does read more like a vehicle to mention the Thomanerchor performance at the Rheingau-Musikfestival (and you admitted as much in the discussion above). I think this is not a good use of DYK. —Kusma (talk) 11:00, 1 September 2022 (UTC)

You said nothing that wasn't said above, sorry. (I don't want to give the public that there's a choir, but my picture of that choir, which includes the news that after two years without tour, they are back, and great.) - I like the attention for Bernstein along with his Divertimento. The hook (and the article) wasn't by me, - I had suggested something mentioning the composition first. --Gerda Arendt (talk) 11:51, 1 September 2022 (UTC)

I agree with tlc. The bolded article is the point of this project. We shouldn't encourage shoehorning in other links. Valereee (talk) 13:27, 1 September 2022 (UTC)

Did you see the Bernstein Divertimento case? The only way to avoid that the pictured Bernstein got more clicks than his work would have been to not mention him which for me would be unthinkable, - the creator of a (bolded) work of art deserves credit. The word shoehorning is not in my limited vocabulary, sorry. --Gerda Arendt (talk) 14:01, 1 September 2022 (UTC)

Sorry, shoehorning refers to inserting something unneeded or irrelevant to meet some other agenda. No, I don't think you should leave out things that are relevant. But I don't think we should be encouraging people to add links that aren't needed by rewarding them for doing so. Valereee (talk) 14:38, 1 September 2022 (UTC)

Bernstein is closely related to his works so this is a non-example. —Kusma (talk) 17:12, 1 September 2022 (UTC)

You may want to join the discussion on WT:DYK about Süskind's novel where a hook without mentioning the (living) author has been suggested, - and I don't have found the right words to object. --Gerda Arendt (talk) 17:31, 1 September 2022 (UTC)

I would oppose any general rule forcing us to mention the author. Shorter hooks are often better. I will take a look. —Kusma (talk) 18:18, 1 September 2022 (UTC)

I wasn't talking "rules", just common sense. For a book in German (but with the translated title), I believe that the author's name is helpful for placing the thing, especially in this case as he is rather (in)famous. --Gerda Arendt (talk) 19:04, 1 September 2022 (UTC)

Leeky has this correct. If you can't find anything DYK worthy in an article, just don't nominate it. DYK is not an obligation and some articles just aren't DYK-worthy because they don't lend themselves to hooks. A good example is Bighorn Divide and Wyoming Railroad (an article I created, to avoid criticizing anyone else), which doesn't have anything wrong with it policy wise, but just doesn't have anything hooky. So I'm not going to nominate it for DYK, and that's ok. DYK isn't just a participation trophy for creating articles. Trainsandotherthings (talk) 20:53, 1 September 2022 (UTC)

It wasn't the question if a new article has something interesting to say, but what to say. We could have said about Salmo 150 - as the piece's publisher said - that it is the most-performed sacred choral music from Brazil in the U.S., or could say - as we did, supported by more sources and more solid sources than the U.S. publisher - that it was chosen for a well-observed concert in Europe, even giving it the title, which is saying something stronger. --Gerda Arendt (talk) 07:22, 2 September 2022 (UTC)

I'm gonna ignore the fact that your edit summary was borderline insulting towards me, and focus on the fact that I did not criticize any specific hook or editor besides myself. I was simply agreeing with leeky that DYK's purpose is to showcase new and expanded content, not whatever we fancy. That you felt the need to defend your hook for Salmo 150 unprompted says a lot to me. Multiple experienced editors are disagreeing with you here, and you're dismissing them out of hand. Again, some articles just aren't fit for DYK, such as the one I wrote that I linked, simply because there's nothing "hooky" to be said about them. That doesn't mean there's anything wrong with the articles or their editors. Trainsandotherthings (talk) 11:50, 2 September 2022 (UTC)

thank you --Gerda Arendt (talk) 12:17, 2 September 2022 (UTC)

I have a question about the stats for Kline's Mill, Virginia

I think it did crazy-good for a little article that was resuscitated from an AfD, about a little-known historic mill in Virginia - not a world event, not a place that's notorious. It got 4500 hits in the 12 hours that it was on the Main Page...is that enough for the hook to get added to the statistics page? Thanks, Shearonink (talk) 05:02, 28 October 2022 (UTC)

Oh, and...ummm...wth does "vph" stand for in the DYK/stats-tables? It's not explained anywhere on this page or on any of the stats pages I looked at...if it's there I couldn't find it. I was thinking that if I don't know, maybe other people don't know either - I'm just the person who decided to ask the question. Shearonink (talk) 05:10, 28 October 2022 (UTC)

Hi there, @Shearonink! "vph" stands for "views per hour" – in general, "making the stats page" (as in being recognized for it; these days, all hooks get recorded here anyway) requires 5,000 views in 12 hours, or 10,000 in 24. So, 416.6666... vph. Sorry about that! I do agree that it did remarkably well, you put together a good hook :) theleekycauldron (talk • contribs) (she/her) 10:09, 28 October 2022 (UTC)

Raising the notification bar

The stats pages' current bar for listing a hook is 416 2⁄3 views per hour – but that number is out of date. First off, our hooks have become more popular – if that threshhold was once rare, it's now around 20–25% of all the hooks we run, and over half of our image hooks. In addition, that barrier was about practicality as much as it was rewarding – back then, the stats pages were updated manually, and there were only so many hooks that could be added. Now, every hook is recorded, it's just that not everyone is notified that their hook has been recorded. Also, 416.6 is just a really awkward number. Would anyone object to me raising the bar of notification to 600 views per hour for non-image hooks, and 800 views per hour for image hooks? theleekycauldron (talk • contribs) (she/her) 07:11, 27 November 2022 (UTC)

• Fine by me. I don't care about stats at all. --Gerda Arendt (talk) 07:32, 27 November 2022 (UTC)