 |
| Naked Thinker |
At this time of year it is my custom to review the posts for the year just ended to see which attracted the most visitors. This year I have changed my approach to try and produce a more representative selection. For this year's list I have included some posts from the previous year (2018) which were published too late in that year to impact on the list. I have evened out the distortions caused by different publication dates by ranking them all based on a time- averaged hit rate*. The full result is shown and discussed in Most Popular Posts of 2019, this article explains the new methodology
Rationale for change of method
I have agonised in recent years about the arbitrary nature of this annual survey. As it is simply based on accumulated hit counts, the timing of publication significantly affects the opportunity to score and hence the position in the table. Thoroughly deserving posts often don't get chosen at all. Typically posts after mid year don't have enough time to register anywhere in the top 20. Wrestling Arsenal (December 2018) was a startling and unique exception last time, but notable recent anomalies were 'BuiltByTallSteve' and Chirenon's 'A New Life For Our Airforce Boys' both published in late 2017. Neither of these got a mention in their year of publication despite turning in a top 10 performance in their first 12 months.
This year (2019) nothing after July made the cut based on total hits. At the other end of the scale several posts from December 2018 registered counts during 2019 that would have got them into the top 20 in any year since records began (as they say) had their publication date been more favourable.
To address these issues I have cast my net a little wider this year, looking right back to mid 2018 to see if any of those posts ought to be recognised. Based on my usual criteria (i.e. considering 2019 posts only) the cut-off point for inclusion in the 2019 top 20 would have been 475 total hits or if I based the selection on an average hit rate* that irons out duration differences, the cut would have been 617 hits per year (pa).
I compiled a short list of posts from 2019 and the second half of 2018 which met both these criteria but excluding those that had previously made it into the 2018 chart. This gave me a list of 23 posts in total which left me some scope for 'tweaking' anomalies in the final top 20.
The cut-off figures of the top 20 in this selection were lifted by about 10% by this process, but for the top 23 they were exactly the same as they would have been under the old method. This is because the 2018 additions in the new list were offset by the loss of 2019 posts which met the total count criteria but had taken much longer to reach it.
The cut-off figures of the top 20 in this selection were lifted by about 10% by this process, but for the top 23 they were exactly the same as they would have been under the old method. This is because the 2018 additions in the new list were offset by the loss of 2019 posts which met the total count criteria but had taken much longer to reach it.
Under this new method, late posts from 2019 still get excluded, but they have a second chance of inclusion next year when they have proven more durable appeal.
The table below shows all the affected articles listed in order of total hits since publication
2018 articles brought in under the new method are shown in red italics.
2019 articles excluded under the new method are shown in green
Column 2 shows how the table would have looked under the old system.
I discuss the results after the table.
The table below shows all the affected articles listed in order of total hits since publication
2018 articles brought in under the new method are shown in red italics.
2019 articles excluded under the new method are shown in green
Column 2 shows how the table would have looked under the old system.
I discuss the results after the table.
Old Style Table Based on total hits since publication
| ||||
New
Rank
|
Old Rank
|
Title
|
Hits
|
Publication
|
1
|
1
|
2820
|
Jan
| |
2
|
2288
|
Dec '18
| ||
3
|
1697
|
Dec '18
| ||
4
|
2
|
1302
|
Jan
| |
5
|
3
|
1269
|
Mar
| |
6
|
1176
|
Dec '18
| ||
7
|
4
|
1166
|
May
| |
8
|
1051
|
Sep '18
| ||
9
|
1031
|
Aug '18
| ||
10
|
1002
|
Nov '18
| ||
11
|
5
|
938
|
May
| |
12
|
701
|
Dec '18
| ||
13
|
6
|
690
|
Jun
| |
14
|
7
|
639
|
May
| |
15
|
8
|
605
|
Jan
| |
16
|
9
|
602 est
|
Jun
| |
17
|
10
|
544
|
Feb
| |
18
|
11
|
539
|
July
| |
19
|
12
|
537
|
Apr
| |
13
|
528
|
Jan
| ||
20
|
14
|
520
|
May
| |
15
|
518
|
Jan
| ||
16
|
505
|
Jan
| ||
21
|
17
|
504
|
Mar
| |
18
|
489
|
Jan
| ||
22
|
19
|
486
|
Apr
| |
23
|
20
|
475
|
Apr
| |
Stats collected on 31st December 2019
This table vividly demonstrates the effect of publication timing once again, with nothing after July 2019 making the cut and the first seven posts published in 2019 all getting in. Unsurprisingly the older, newly-included 2018 posts do well in this basic hit count table, but the top 2 posts are actually the same under the new system too with the same winner romping away with the title
(see final MPP ranking article for 2019).
(see final MPP ranking article for 2019).
The three 2019 articles excluded from the table by the new criteria are shown in green above. All these articles had a full 12 months to make their score so replacing them with three far better performers from December 2018 is not an earth-shattering injustice. I scored a bit of an own goal by excluding No 16, but the 'War Comic' article that's linked to it gets into the new ranking at No 9 so I'm happy with that.
The winning total of 2820 comfortably exceeds last years record by IFNB of 2519 by about 12%, but this primarily reflects a longer scoring period (12 months for Milking Factories vs 9 for IFNB). Also, total visits to the site increased by 24%. this year so this performance probably doesn't count as a meaningful 'record'.
The full final results produced by the new system come out slightly different and are discussed in Most Popular Posts of 2019
*The Average (Annualised) hit rate was calculated as follows
(Total Hits) divided by (Days since Publication) multiplied by 365 (days in a year)
This does not produce a 'real' number, the formula reduces the counts of articles that have been around more than a year and inflates those published less than 12 months ago but they end up on a level playing field for comparison purposes.
Remember, these comparisons are only made for entertainment only, since I don't really know how accurate Google's stats are and I suspect they don't even count people who just browse through the blog roll.
Thank you to all my readers for your interest in my blog,
I will publish the updated all-time list of favourites shortly.
Happy New Year!
I will publish the updated all-time list of favourites shortly.
Happy New Year!
No comments:
Post a Comment