[Warning: Contains non-ideally formatted graphs and a lackadaisical attitude towards statistical assumptions. I was using my NaNoWriMo output to learn how to use new statistics software, so the graphs were formatted and the tests picked to let me use as many beginner-friendly R functions as possible, not necessarily to give the most appropriate tests. Sorry.

Also warning: This is just unremitting navel-gazing about my writing habits. That’s all it is. Not sorry about that.]

Background for the NaNo Plan

This year seemed like my year for NaNo. I’ve spent the last two years stressed out of my brain, and couldn’t face it, but this year has actually gone well for me. I had my Candidacy exam (giant oral exam of doom) scheduled October the 5th so I came up the following scheme.

I’d finish my exam, spend a week or so recuperating and then have the rest of October to clear things up at work so I would have free time, and to outline and prep things for November.

The Original NaNo Plan

I’ve never finished a novel draft before, and I’m a slow writer, so I was always planning to do a modified NaNo to ease into it.

I settled up on the following rules:

  • – Write as many words as possible during November, aiming for 50,000
  • – Do not edit, just move onto the next chapter or the next project.
  • – You can start new projects if you want to (I’ve been trying not start new projects until I finish some of my open ones… it hasn’t been going well)
  • – Any and all writing projects count, journaling doesn’t count, internet comments don’t count and nothing written for work counts.

NaNo Goals

I didn’t come into the project with especially clearly defined goals.

Realistically, I knew I wasn’t going to manage 50,000 words. I figured 25,000-30,000 would probably be pretty good optimistic count to aim for, but I basically just came into it aiming to write as much as possible, however much that was. I actually wrote just bit under 22,000 words, so I wasn’t far off.

I didn’t have any specific ideas about writing every day or not going in, but by a few weeks in I decided that I really wanted to try and write every day. This I did manage. I have no daily word counts of 0.

My long standing, number one writing goal, independent of NaNo is to finish a long work of fiction. I didn’t expect to do this during NaNo. I did use NaNo to polish off a number of shorter drafts that have been hanging in partially completed limbo though. I count this as a partial success. I didn’t complete a long work, but I did complete things, which, I hope, is good practice in finishing stuff.

Actual NaNoWriMo Proceedings

I did basically stick to the NaNo rules I laid out above. Some of my NaNo word count ended up coming from letters, which I did edit so I could send them, but to fit that into my no-editing rule, I wrote the whole first draft without stopping, counted all of that in my NaNo word count, and then excluded any editing.

What I didn’t do was any of the outlining I’d planned. At all. I totally underestimated how much post-exam downtime I would need and so, by the time November 1st rolled around, I hadn’t touched a single writing project in about 2.5 weeks for any reason, and so I started NaNo fairly cold.

And, because I apparently enjoy making my life difficult I simultaneously decided to make and send Christmas cards to my friends and extended family which also has an effective December 1st deadline (this is what happens when you need to mail cards to Australia).

Quatitative NaNo Outcomes

I collected 4 pieces of data about what I wrote: word count per work per day, which is summed into word count per day if, for each work, I’d used any existing notes, scene fragments or outlining, if I’d completed the first draft of the work that day, and if, subjectively, I felt busy that day.

I toyed with the idea of trying to collect a less subjective measure of busyness, but to start off with busyness is inherently subjective. Tracking actual writing minutes would have been distracting and any other measure would have involved some judgment about how my day compared to my baseline level of “busy” and how that in turn compared to the “average” level of “busy” which I just wasn’t up for.

So how did I do?

Basic Statistics:

Total Word Count: 21,922

Mean Words Per Day: 731

Median Words Per Day: 595

Total “Works”: 24

  • 3 essays
  • 3 letters
  • 6 drafted science blog posts
  • 12 fanfic chapters split across 7 fanfics
    • Of the fanfics 4 chapters were additions to ongoing works, 2 were totally new and 1 was a rewrite from scratch of an older fic.

Works Completed: 19

So, breaking that down a little:

overallwordcountgraph

Here’s my overall word count trajectory. The blue line in the middle is my cumulative word count per day.  For comparison the green line at the top is an “ideal” NaNoWriMo performance, its what you would get if you hit the recommended word count of 1,667 words per day, every day. The red is my best estimate of my “usual” or “average” output. I haven’t previously tracked my word count, so I don’t have a true average, and I don’t typically write every day. But I decided that 200 words per day was a reasonable representation of my typical output, and that’s what’s represented by the bottom line.

I fell pretty short of the 50,000 word goal, as you can see in the gap between the top and middle lines, but you can also see that compared to my ‘usual’ writing, my output increased tremendously. My average is more than three times my 200 word baseline, and my median is more than twice as high.

Speaking of the difference between the mean and median: Here’s how that breaks down per day, you can see there’s a lot of variability and it might be contributing to the gap between my mean and median:

countbyworkbyday-updated

Here the height of each bar is word count on that day, and the word count per work on each day is represented by the coloured blocks inside them. The per/work details are more or less unreadable on this graph, but it gives you a sense of how many works per day were contributing to the word count, and you can see the variations in word count over time. For all of these analysis I counted each individual fanfic chapter as a separate work, just to make them more comparable to the letters and blog posts. Otherwise I’d be comparing adding a few thousand words to a 10,000-30,000 word fanfic to completing a 1800 word letter.

To give a better sense of how my overall word count was distributed across the month I created this histogram:

histogram

So the height of each bar here, represents how many days I had a word count in the range on the X axis. As you can see it’s a big range and kind of a mess. Overall its left-skewed but then there’s that spike of high-output days on the right, but not with enough variation to make a true U-curve. Gross.

I also created a more detailed break down of not only word count for each day, but word count for each work on each day.

goodbiggraph

This shows the cumulative word count for each work on each day. So you can see where individual works were picked up and stopped, and the dots indicate which days they were actually written on. The enlarged dots indicate a completed first draft. Triangular points indicate that I used some kind of existing notes or outline. Also, a note on how I calculated the word counts. This represents word count added on each NaNo day. I didn’t count the existing word count in on-going projects, and I didn’t add a ‘day 0’, which would have shown them all starting at 0 before NaNo started. So a lot of the works that seem to ‘complete’ at very low word counts, represent me finishing an existing work, not writing something very short from scratch, and the distance from y-value of the first point represents the number of words I added to it the first day.

The numbering on the fanfic is (work identifier)-(chapter number), for reference.

From there, since I had time and needed to learn the code anyway I decided to probe some of the factors that affected my word count with some (terrible) inferential statistics.

First I ran an independent two-tailed t-test to look at the difference between busy and not-busy days (remembering that that is a purely subjective rating I made on each day and that my group size for “busy days” was only 7.

busynessttest

The graph depicts the mean plus or minus standard deviation for busy and non-busy days, with the overall mean on the left for comparison (it wasn’t included in the inferential stats, I just couldn’t figure out how to make it a different colour to indicate that). Given the small group sizes and obviously unequal variance I don’t think these are really valid, but for what its worth the difference, the Welch’s Two-Sample T-Test, assuming unequal variances, reports a T=-4.8786 and p<0.001. So feeling busier than usual, whether or not that is actually reflected in how much time I had available to write (which I don’t actually know) did seem to have an impact on my daily word count.

Next I tested how what I was working on affected how much I wrote.

I’ve been attempting, for a while now, to restrict how many ‘active’ projects I worked on at any one time, because I’ve heard from many sources that doing that will keep you from getting distracted and help you complete projects, which is my overarching goal.

So, I performed a correlation between my daily word count, and the number of different things I’d worked on that day that contributed to that word count to test that out.

countnumbercorrelationplot

Here the individual dots and their colour, just to test my dot-coloring function, show the daily word count for a single day, with how many works contributed to that count on the x-axis.

I performed a Pearson’s correlation, which is probably inadvisable for data this poor quality but with 30 data points I’ve just barely got enough data to probably get away with it. The correlation had r=0.73, r2 = 0.54, p<0.001 though which seems frankly unrealistic, so perhaps not.

Using the non-parametric Kendall’s tau instead, which should, at least in theory, be a bit more robust to my garbage data, I got tau=0.63 p<0.001. Kendall’s tau is also a better test for small datasets than the other non-parametric correlation, that I was actually taught, Spearman’s rho, which I learned while investigating why attempting to use Spearman’s was giving me error messages.. Although I do get a warning that my p value might be imprecise with Kendall’s tau as well. This is well beyond my actual understanding, because I don’t use these tests regularly, so who knows.

Even assuming both of these tests are overestimating the effect by quite a lot, I think I’ve been getting bad advice. Apparently, by having a lot of projects open and moving between them in a writing session, I really do write more, at least in terms of raw word count. I don’t have statistics to test whether it affected completion rate which is, at least in theory, the other reason to try and focus on one project at a time.

I have some data on what was completed, but given that it’s mix of different project types, some of which were started and some of which were not, and with wildly different completion rates, I don’t have enough data points (I only have 5 non-completed works) to be worth analyzing, even by the very lax standards I’m using here.

The last test I used to examine what affected my word count, was a one way ANOVA to test how what kind of material I was writing affected my word count. My work conveniently falls neatly into 4 categories, fanfic, fan meta, science blog posts, and letters. I would have liked to have original fiction in there to, but it didn’t happen. And this led me to more adventures in learning about how statistics works.

anovaboxplot1

This boxplot shows the word count per day in quartiles. The median is the black line in the middle of the box plot. The second and third quartiles are the top and bottom of the boxes, the edges of the whiskers are the first and fourth quartiles. The two black circles are outlier greater than 1.5x the 4th quartile range.

The ANOVA was non-significant p=0.07.

I was taught not to do post-hoc tests on non-signficant ANOVAs, but I ran it anyway, just to see if I could get the analysis to complete, expecting a pile of nothing. But my post-hoc (Tukey) test was significant. Science posts have higher per-day word count than fanfic  (p<0.001), and letters (p=0.02) and fan meta have a higher per-day word count than fanfic (p=0.007), all the other differences were nonsignificant. Google reveals some debate as to whether these are meaningful or not and even if it is, while the R documentation I was able to find says that the Tukey test function I used is set up to compensate for a ‘reasonable’ amount of group size variability, I have no idea if my degree of variability is ‘reasonable’ or not.

It does confirm my overall intuitions about my writing speed… but given the small N, the varying group sizes and uneven variance, I think this analysis is probably fairly meaningless. But hey, I’ve learned a lot about post-hoc tests now. So there’s that.

I opted not to graph the results that I had about the effect of using existing notes, because the kind of notes I was using was different for each type of writing and I just didn’t think that made for a meaningful comparison. Fiction mostly involved existing scene fragments, I write detailed outlines for my meta, and the science blog posts, most problematically often come with a built in “outline” in the form of whatever I’m summarizing, even when I don’t write a formal outline for them.

Qualitative NaNo Results

In addition to the data that I collected, I also have feelings.

Even though I don’t have stats to back it up, I am extremely sure that the lack of outlining affected both my output and my completion rate.

By about half-way through I could really tell when I ran out of outlines, in the sense that I was stopping things or slowing down because either I hadn’t planned the next set of events, or because I needed to work out world-building or timeline details. To the point that for the 7 unique fanfics I worked on, 6 of them I stopped working on (either I didn’t complete my chapter, or I didn’t move onto the next one once I did) because I needed to stop and work out outline details.

The reason I didn’t start any original fiction, even though I have 2 or 3 ideas I’d really like to take a shot at, was also that I didn’t have a chance to set up any world building or general outlines.

This is interesting to me because I think of myself as more of a pantser than an outliner. I write detailed outlines for most essays, but for fiction I usually write a general premise and then outline a chapter or so at a time.  I guess I needed those short-term outlines more than I thought.

I also feel like just the act of tracking my word count, and the projects I was working on had an affect on how I worked, and how much I wrote. Having to refer back to a central list of works meant I never had to spend time or energy figuring out what to write, and because I had reference for when I’d last worked on each thing, nothing got lost.

Conclusions

There’s a couple of aspects of NaNo that I want to keep doing and a few that I don’t.

Firstly, it was a  great way to break out of a writing slump. A+ writing slump cure.

Daily Writing – No

I’ve been periodically trying to get myself to write every day for a while, but this is the first time I’ve actually managed it for more than 2-3 day streaks. Having now written every single day, for a month, I feel confident saying that its not for me. I enjoyed it as a challenge, I’ll probably try NaNo again, but as a day to day practice, it got stressful towards the end. This is partly because some of my writing days need to go to editing and outlining  and I don’t alway have time to do both in a day, but its also because I have other hobbies.

I did write every day, and I balanced my writing with making cards, but that’s it. Not only did not outline, I also didn’t record podfic, or vid, or make jewelry, or learn coding or cook any new recipes, which are all hobbies I usually balance with writing. So in future I’m going to go back to splitting my free time between writing and everything else, and inevitably, that means I won’t write every day, and that’s fine.

Word Count Tracking – Yes

I used a really over-complicated way of tracking my writing across three excel sheets. I had a sheet I used to track and compile my statistics for each day. One that I used for statistics for each work on each day, and a third which I used to track what I had been working on and when I last worked on (shown below in that order).

screen-shot-2016-12-05-at-7-18-18-pm

screen-shot-2016-12-05-at-7-19-30-pm

screen-shot-2016-12-05-at-7-19-17-pm

I’m not going to keep doing all this, its too much effort, but I’ve created a cut down version where I’m just tracking word count per work per day (like my third picture above). I’ve also added 2 columns tracking on which days I finish the first and final drafts of things, so I have a form that’s easily readable that I can use to track what I’m working on, but if I ever decide to rework that statistics on a larger data set I’ll be able to extract all the details I used with 10 or so lines of code.

This is much faster, instead of calculating and then summing words/day I just need to record the value from the word counter and if I decide I want to do more statistics to follow these up, I’ll be able to extract most of the word/day measures that I used here, except for note use, which I didn’t end up analyzing, and busyness, which will probably stop being meaningful when I’m not writing every day.

I think this may be the tool I’ve been looking for to let me balance moving between projects, with not abandoning them and I’m very excited about it.

Not Stopping to Edit – Yes

Finishing the first draft without going back and fidgeting with things is something I don’t do well, and this was very good practice. I think it really does work for me in terms of progressing through writing even though its stressful and I hate it. I’m going to keep trying to do it.

Outlining – More

I thought I didn’t rely much on outlines, but I do. I will respect my outlining needs from now on.

Sticking to One Project – No

I have a month of experience and an extremely poor quality correlation suggesting that I should cease this. And trying it has been very frustrating, so I’m stopping. Multi-tasking all the way.

More seriously and more generally, I feel like, because NaNo is so much more high intensity than my usual writing behaviour,  I essentially just crammed about 3.5 months worth of writing experience into 1 month. And because all that experience occurred so fast, I could hold a lot more of it in my head than I would have been able to on my normal schedule. Having had a chance to try and stick to one work, and try moving through a lot of them in very rapid succession, not only do my shaky stats suggest I get more done, I also know that it feels better. I don’t get as frustrated or waste as much time refreshing Tumblr in case I’ve somehow left my inspiration there.

Momentum – What to do with it?

And speaking of inspiration. The piece of common writing advice I had more or less abandoned as not working for me, which suddenly worked really well during NaNo, was the idea that inspiration is a lie, and the best way to write was to sit down and make it happen. And I’ve never been able to get that to work. I’ve always had a bad case of “oh I can’t write today, my muse is not speaking to me”. Because I’ve always seen it in the context of ‘sit down and write a specific project you want to finish’.

Well, when I sat down and wrote anything I could think of words for, because I wanted to get as close to my target word count as possible, suddenly, my muse got shockingly cooperative. Once I started working on something, no matter what it was, it got easier to think of what to do with, not only that project but on the other projects I was stuck on.

And even better, this has so far, carried over now NaNo is over and I’m not writing ever day. I have a bunch more ideas about what I want to write and how I’m going to fix the issues with my current stories. Of course, I’m also bubbling with shiny new ideas right now, while I have 19 things that need editing and I’m trying to get ready for Christmas. So there’s also that that.

Graphs – Probably not

I had fun with these graphs, but I also had two days to work on them (this was technically ‘work’, because I had some time to teach myself R). I probably won’t spend a lot of time graphing my actual word count in the next year, outside of the cumulative word count graph (the first one).

Incidentally, I’d love to see what other people’s writing progress looked like through the month if anyone else wants to share.

Advertisements

2 thoughts on “NaNoWriMo After Action Report

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

w

Connecting to %s