User:EdChem/DYK clicks

I've been thinking a lot about the criticisms of WP:DYK relating to poor view / click rates associated with hooks. In particular, the claim that hooks on average result in 1000 views strikes me as an underestimate. However, Physchim62's statistic is an empirical result based on his analysis, whereas my intuition is really just an opinion - and as a scientist, I am well aware which carries more weight. So, I am creating this page as a record of my investigation of the views that DYK hooks receive.

Preliminary Investigation
As a preliminary investigation, I have examined a few recent hooks at random, and found that single-article hooks with under 1000 views seem rare. I also found that the non-bolded links in the hooks appear to also attract hits. So, step one will examine a few of my DYK noms.

Hook for Hans Freeman

 * ... that Hans Freeman introduced protein crystallography to Australia and determined the structure of plastocyanin, the first protein to be structurally characterised in the southern hemisphere?

On 13 August 2010, the day of the hook's main page appearance, the Hans Freeman article received 1158 views. Average views for the other 30 days of August was (1907 &minus; 1158) &divide; 30 = 25 views.

The protein crystallography article received 988 views on 13 August. Average views for the other 30 days of August was (1440 &minus; 988) &divide; 30 = 15 views.

The plastocyanin article received 1063 views on 13 August. Average views for the other 30 days of August was (2015 &minus; 1063) &divide; 30 = 32 views.

The Australia article receives so many views that attributing any views on 13 August to the hook is unjustified. The protein article shows no variation in views when the hook was on the main page, and in fact the views spiked to more than triple the daily average on 27 August (the reason for this spike is not clear). The southern hemisphere article does show a spike in views on 13 August, one of four that month. Taking the average of the remaining 27 days, the average views for that article run at (38663 - 9400) / 27 = 1084 per day, so the views in excess of that (2000 &minus; 1084 = 916) on 13 August might reasonably be attributed to the DYK hook.

So, for this hook, views of all articles linked in the hook over and above their average daily views (and leaving out articles where no extra views can reasonably be inferred) were (1158 &minus; 25) + (988 &minus; 15) + (1063 &minus; 32) + (2000 &minus; 1084) = 4053. Put another way, the bolded article views represented (1158 &minus; 25) / 4053 = 28% of all views attracted by the hook.

Hook for rhodocene

 * ... that the rhodocene monomer exists in solid state at or below liquid nitrogen temperatures (&minus;196 &deg;C) and in gas phase above 150 &deg;C but not at room temperature?

On 14 August 2010, the day of the hook's main page appearance, the rhodocene article received 2147 views. Average views for the other 30 days of August was (3072 &minus; 2147) &divide; 30 = 31 views.

The monomer article received 1298 views on 14 August. Average views for the other 30 days of August was (16316 &minus; 1298) &divide; 30 = 501 views.

The gas phase article received 893 views on 14 August. Average views for the other 30 days of August was (1326 &minus; 893) &divide; 30 = 14 views.

The liquid nitrogen article receives views that follow a predictable and repeating pattern with a period of seven days. Based on that pattern, the article received approximnately 1000 extra views above that pattern on 14 August. The room temperature article follows a similar (seven-day pattern and approximnately 850 extra views above that pattern is a reasonable estimate of the effect of the hook.

So, for this hook, views of all articles linked in the hook over and above their average daily views were (2147 &minus; 31) + (1298 &minus; 501) + (893 &minus; 14) + 1000 + 850 = 5642. Put another way, the bolded article views represented (2147 &minus; 31) / 5642 = 38% of all views attracted by the hook.

Multi-Article Hook

 * Ophiophagus hannah (2).jpg ... that ophanin, piscivorin, ablomin, latisemin and triflin, found in the venom of the King Cobra (pictured), the water moccasin, the Mamushi snake, the Erabu sea snake and the Habu snake, respectively, are all cysteine-rich secretory proteins that can reduce muscle contractions?

On 30 October 2010, the day of the hook's main page appearance, the ophanin article received 871 views. Average views for the other 30 days of August was (1177 &minus; 871) &divide; 30 = 10 views.

The piscivorin article received 675 views on 30 October. Average views for the other days in October when the article existed was (1071 &minus; 675) &divide; 12 = 33 views.

The ablomin article received 603 views on 30 October. Average views for the other days in October when the article existed was (1010 &minus; 603) &divide; 16 = 25 views.

The latisemin article received 517 views on 30 October. Average views for the other days in October when the article existed was (738 &minus; 517) &divide; 12 = 18 views.

The triflin article received 619 views on 30 October. Average views for the other days in October when the article existed was (983 &minus; 619) &divide; 12 = 30 views.

The Cysteine-rich secretory protein article received 1404 views on 30 October. Average views for the other days in October when the article existed was (1521 &minus; 1404) &divide; 4 = 29 views.

The venom article received 1300 views on 30 October, one of three days in October where a spike in views occurred. Each of these days corresponds to the appearance of the venom article as a link in a DYK hook. Looking at the trends, approximately 500 additional views can reasonably be attributed to the DYK hook.

The King Cobra article receives views that follow a predictable and repeating pattern with a period of seven days. Based on that pattern, the article received approximnately 3400 extra views above that pattern on 30 October.

The water moccasin article received 1901 views on 30 October. Average views for the other 30 days of October was (4580 &minus; 1901) &divide; 30 = 89 views.

The Mamushi snake article received 1217 views on 30 October. Average views for the other 30 days of October was (2290 &minus; 1217) &divide; 30 = 36 views.

The Erabu sea snake article received 1057 views on 30 October. Average views for the other 30 days of October was (1229 &minus; 1057) &divide; 30 = 66 views.

The Habu snake article received 1259 views on 30 October. Average views for the other 30 days of October was (4759 &minus; 1259) &divide; 30 = 117 views.

The pattern of viewing of the muscle contraction article follows a seven-day cycle, but estimating the extra views reasonably attributable to the DYK hook appears difficult and imprecise. So, the sensible inference is to assume that the hook produced no extra views.

Interestingly, there is one other possible link in this hook, that to the image itself. The File:Ophiophagus hannah (2).jpg received views on 30 October. Average views for the other 30 days of October was (5656 &minus; 3192) &divide; 30 = 82 views.

So, for this hook, views of all articles bolded in the hook over and above their average daily views were (871 &minus; 10) + (675 &minus; 33) + (603 &minus; 25) + (517 &minus; 18) + (619 &minus; 30) + (1404 &minus; 29) = 4544. Views of all articles not bolded in the hook over and above their average daily views were 500 + 3400 + (1901 &minus; 89) + (1217 &minus; 36) + (1057 &minus; 66) + (1259 &minus; 117) + (3192 &minus; 82) = 12136. Put another way, the bolded article views represented 4544 &divide; (4544 + 12136) = 27% of all views attracted by the hook.

Summary
Looking at just these three hooks, we find the following:

At least for this small sample, it is clear that the views for the non-bolded articles make up the majority of views for the hook - somewhere between 1.5 and 3 non-bold article clicks per bold article click. Thus, statistics based solely on the articles highlighted are likely to substantially under-estimate the interest generated by the DYK project. Unless these hooks are massively atypical, the claim of an average of 1000 views from each hook at DYK appears also to be an under-estimate. Perhaps the most important conclusion, examining an entire days worth of hooks by this method would provide a worthwhile next step in this investigation.

Complete Set of Hooks - 14 August 2010
Looking at this single set of 9 hooks, and after correcting for typical views for articles, the bold articles attracted 21,100 views. The non-bold articles attracted a further 21,204 views. Thus, a total of 42,304 views (or 4700 views per hook). Assuming each of the four sets in a day achieved similar results, this equates to almost 170,000 views from the DYK section in a single day

Looking at this single set of 9 hooks, and after correcting for typical views for articles, the bold articles attracted 14,438 views. The non-bold articles attracted a further 20,983 views. Thus, a total of 35,421 views (or 3936 views per hook).

Adding this to the previous set, we have 35,624 views of bolded articles and 42,803 views of non-bolded articles from two update sets (half a day).