The Winter is Coming guys recently carried an article about the possibility that the Game of Thrones series will catch up and possibly overtake George RR Martin’s Song of Ice and Fire books. The authors of the original article assume a seemingly optimistic date of late 2015 for the release of the putative sixth book – The Winds of Winter. Since the sixth book is supposed to be huge, my initial reaction is that there is absolutely no chance that GRRM would finish it in only five years. However, looking around the web for publication dates, I noticed that the third book, The Storm of Swords, came out only a year after the second. This got me thinking – how often did these books actually come out? And more naively (and geekily), can we learn anything about when we might expect the next book? If you have a look at the release dates of the books, you can plot a book over time graph, like so:

The numbers on the x axis correspond to the number of the book in the series, except for ‘0’ which I set at 1994 – a date cited several times as the time GRRM started writing the series in earnest. By turning all the data points into numbers, we can plot a trendline and try to predict when the next books might realistically be expected. The simplest way to describe a relationship between two numerical data sets is a linear regression: effectively a mathematical model which assumes that the two data sets change (increase or decrease) at a constant rate. This can then be used to predict unknown data points – in this case the years that the next SOIAF books will come out. Plotting a linear regression gives you this:

Using the equation above, we get release dates of ~**2012**, **2015** and **2018** for books 6, 7 and a possible book 8*. Now, since we are in 2013 and book 6 had not been published yet, clearly this model is not very useful. Specifically, it accounts poorly for book 3 (which came out much sooner than the trend predicts) and book 5 (which came out much later than predicted).

**Dodgy Math**

Looking at the original graph once again, we can see that the data could be (somewhat arbitrarily, but bear with me) divided into two groups: data points 0-2 lie on a straight line as do data points 3-5. If we look back at interviews with GRRM, it’s clear that there was supposed to be a time gap of a few years in the plot between book 3 and 4. This plan was subsequently changed and GRRM decided to continue with the plot of book 4 continuing the timeline of the end of book 3. Book 4 then blossomed into two books: Feast for Crows and Dance with Dragons, which subsequently became books 4 and 5 on our graph. Using that rewriting gap between books 3 and 4 as justification for breaking up the data, we can plot two separate linear regressions**:

The red line models what might have happened if the books continued at their initial rate (2.5 years per book and the series would be close to finished by now). The black line models the rate new books will come out if GRRM maintains the rate of writing of the last two books of the series (a book every 5.5 years or so), giving us publication dates of books 6, 7 and 8 as: **2016**, (late) **2021** and **2027**! While the last two dates might be somewhat shocking, the estimate of 2016 is not that far off from the Wertzone’s prediction of late 2015.

**Ever the optimist**

However, there is more optimistic estimate we could get from the above models. Let’s assume that the structural and chronological changes between book 3 and 4 caused an unexpected dip in GRRM’s rate of writing. Let’s also assume that these problems have now been circumvented and that an emboldened GRRM might start churning them out at the same rate as the early ones. In addition, as various plot lines get wrapped up, the books may get easier to write once again. I don’t have a graph for this, but the values of 6, 7 and 8 would be: 2013.5, 2016 and 2018.5, which is a book every 2.5 years. If we adjust for the fact that GRRM stated he wouldn’t start writing book 6 until January 2012, we get: **2014.5**, **2017** and **2019.5**. As far as I can see this is the most optimistic estimate possible – and I find it unlikely, as the manuscript would have to be more or less complete by the end of 2013 to hope for a mid-2014 publication date.

**Winter is not coming just yet**

Contrastingly, we could also mathematically model a much more pessimistic situation: let’s assume that writing these books gets more difficult with time. Or mathematically: that the decrease in writing rate has to do with the increasing complexity of the overall ever-branching plot, and that, as books go on (and GRRM gets older), it would only decrease further. We could then model a quadratic equation on all our data:

This model would give us estimated publication dates for books 6, 7 and 8 as: **2016.8**, **2024** and **2032**! In 2032 GRRM would be 84 and I would be in my fifties, which would be quite galling for a book series I started reading as a teenager. I really hope that this won’t be the case…

**Conclusions**

Before we get too serious, I have to emphasise that the above was just a fun*** way to spend a couple of hours on a Saturday morning, and I am not seriously suggesting that I am able to predict the future using excel. In fact, my predictions about book 6 are quite vague and range from it getting published last year (!) to sometime in late 2016. Nevertheless, once book 6 is eventually published, I look forward to (reading it!) adding another data point to my graphs to see which model is looking best. Until then we can just say: write, write like the wind!

You’re a geek, but I love it!