Day of the 'Stache
you liked best. And there were all kinds of opinions. But once I
nerded up and crunched the numbers*,
it became clear. You like the 'stache.
Here are the covers in
question. If you're not seeing a graph,
try this. If you are,
and you enjoy playing with graphs, you can click site names in
the legend to add and remove them. I mention that because
it's awesome fun.
I separated votes by domain because there were interesting
differences depending on whether you responded on
my site,
Reddit,
tumblr, or
Facebook.
Observations:
In all cases,
Cover #5
(Victorian-era dude with enormous 'stache) was
most popular. This was a surprise because I'd thought it was just
too weird. In retrospect, I was probably headed for that trap of trying to
imagine what other people might like, which is always a sure path to
something conservative and uninteresting. So this was a handy
reminder to not do that. Many
people responded very positively to the originality of this
design and were turned off by the same-ness of some others.
Cover #3
(Millions o' Parts) was least popular. This was lucky,
because it was the design that started this whole debate with
my publisher,
and if it turned out that people actually liked it best,
I would have been an asshole. The votes also seemed to back up
my thesis that it appealed more to arty types than geeks, with it being
quite popular on tumblr but abhorred on Reddit (where there were
actually more negative comments than positive ones).
Covers #4
(Smoking Capacitor) and
#6
(Smoking Processor), which were
deliberately similar to the style of my previous covers in
Jennifer Government and Company,
were a lot more popular with people who knew that
(i.e. people on maxbarry.com and my Facebook page).
Reddit liked
Cover #2
(the Robot) a great deal, practically as much as the 'Stache.
I suspect this is due to an affection for retro robots (something I share).
A few people observed that it was less true to the story than #5, though.
Cover #1
(Pixelated Guy) I think suffered from a general feeling
that this kind of thing had been done before. It was seen
as pleasant but not particularly arresting.
If you were wondering, covers #3 and #6 were designed by Vintage,
cover #4 by me, and covers #1, #2, and #5 by up-and-coming
design superstar
Matt Roeser. I
didn't mention that earlier to avoid prejudicing votes.
Comment of the week, from G Lainagier:
In numerical order:
Couplandesque cubicle farce,
Rankinesque steampunk,
Kathy Lette tries something new,
Tom Clancy for the kids of today,
what you wrote,
what you might write but not really this.
I also enjoyed seeing Caleb's battle against indecision,
as he transitioned over the course of
three comments and several
hours from saying #6 was terrible to liking it the best.
I forgot to mention earlier that most of these covers were
concept sketches, not finalized designs. With #5, for example,
a few people criticized the machine legs, which were only supposed
to be placeholders. I'm now working with Matt and
the publisher to refine that. I promise you, those legs will
be awesome. Also: the 'stache stays.
Thank you again to everyone who helped out with this. You
are the burning propulsive mass beneath my rocket boots.
* Nerd details:
I assigned a weighting to expressed preferences:
3 points for most preferred, 2 points for any second preference,
down to -2 for last preference, if one was mentioned. When people said
they liked multiple things equally, I alternated entering them in
the order listed or in reverse. To allow opinions on different
sites to be compared, despite very different numbers of respondents
(about 260 on maxbarry.com, 390 on
Reddit plus a thousand-odd votes, 70 on tumblr, and 50 on Facebook),
I scaled the results: the most
popular choice is scored as 1,000 and other covers based on their
relative popularity on that site. A cover exactly half as popular as the top choice,
for example, on whichever site, has a column exactly half as tall.
Note that this exaggerates a single person's vote on
Facebook and tumblr:
the Reddit and maxbarry.com columns represent many more people's opinions.
On Reddit, where users can endorse another person's comment by
upvoting it, I multiplied the score of each comment by the number
of upvotes. But since users can upvote multiple comments,
even comments saying the same thing, I took the square root of each result in
order to minimize the exaggeration that would have otherwise occurred.
(Without this, the more popular covers on Reddit appeared
wildly more popular.) When highly upvoted comments expressed
equal preferences for multiple covers, I assigned equal scores,
rather than relying on the averaging nature of the alternating system
mentioned earlier.
