Determining the greatest Science Fiction & Fantasy novels since 1970

The one habit I actually stuck with throughout quarantine is reading. In particular, reading all of the science fiction and fantasy novels I can get my hands on. For an hour or two a day, I escape COVID to a distant world.

My bookshelf; Yes, I do own two copies of Hyperion

Like any good obsession, this one is full-on. Since the early days of the pandemic, I’ve read 70 or so science fiction and fantasy novels. I frequent a myriad of SFF communities online. I even started collecting rare and limited editions of my favorites: an expensive hobby that thoroughly confuses most of my family and friends, but does mean I spent an inordinate amount of time picking out first edition covers for all the graphics in this article (you’re welcome).

However, as someone trying to fully engross themselves in a new hobby, I was surprised at how challenging it is to determine which novels are critical to my education.

Which works must I read first? Which authors have I missed out on?

To be fair, I do know the classics. I know Asimov, Tolkien, Dune, Heinlein, Bradbury, 1984, and A Canticle for Leibowitz. But, for many, our education stops there, which only gets you to the 1960s at best. What about the past 50 years? Are none of these works and authors worthy of canonizing?

To try to figure this out, I first turned to Goodreads, where unfortunately the answer is typically the latest Young Adult vampire novel. The question posed to my favorite hive mind on Reddit often elicits Ian Banks’ ‘Culture’ series; although the order to read them in is up for much debate. On Facebook, the 50K strong fantasy group will unequivocally tell you to read the 10-book epic that is ‘Malazan Book of the Fallen’. No matter the question, in fact, the answer appears to be “MALAZAN!!”

One of many Malazan memes shared online

While each of these communities and suggestions has its merit (reading is subjective after all), it got me wondering… could there be a more objective answer? Is there a dataset that could yield a more robust list?

Now, one of my favorite past-times is using my amateur programming and data science skills to make the world a bit easier to understand. My only other article on Medium runs through a similar exercise to summarize how major cities in the United States will be effected by climate change. But after nearly 2 years of a horrific pandemic, it was time to switch gears to something a bit more... diverting.

And the winner goes to…

Thankfully, each year groups of fans and people in the industry ask this very same question, perform their own mini-algorithms, and broadcast the results. Of course, I’m talking about awards; and in particular, the award for best novel. It turns out there’s a lot of them, each with their own eligibility rules, committees, number of nominees, and more. Plenty to sink our teeth into.

But before we get to methodology, no discussion of awards is without its own controversy and debate. (If you haven’t gone down the rabbit hole of puppygate and the Hugos, good on you). Still, awards do have a few pleasing properties that cannot be found in datasets elsewhere.

First, they are awarded at the same time as the work was written. There’s no recency bias, or rose colored glasses to tint people’s judgement. In contrast, consider this list by Bookriot which has some great novels on it and even goes all the way back to Frankenstein (1818). Yet, by the end, nearly half of the “most influential sci-fi books of all time” were written in just the past 5 years. Or, take the Goodreads ratings for Fahrenheit 451, a bona fide classic by Ray Bradbury. While a 3.98 is respectable, all of these reviews came decades after the book was first published — and typically, let’s face it, from a classroom full of high school students that had to read it for an assignment. When the Goodreads reviews suggest that Vampire Academy (4.11 rating) is better than Fahrenheit 451, it’s time to look elsewhere for our source of truth.

Second, awards are granted by groups of people immersed in the industry. Oftentimes by editors, writers, and publishers themselves. People who read lots of sci-fi/fantasy (not students tasked with an assignment). Now, each tight-knit communities have their quirks, so we are going to run into issues of bias and prejudice (especially for older awards). While there is no algorithm that can erase this bias, we can do a bit of crowdsourcing — pulling in lots of awards, from lots of communities — to try and provide a slightly more robust view than that of each award alone.

Lastly, before we go any further, one final note on what awards are and aren’t. Book awards do not typically signal the “most popular”or easiest reads. Similar to the Oscars for best movie, the winner is rarely the latest blockbuster (as far as I know, The Fast and Furious franchise has yet to win). Many of my favorite vacation reads have not won an award. Instead, in theory, awards attempt to identify work that gives us a unique glimpse into the depth of humanity. Award winners, hopefully, help us truly see the world (and ourselves) more clearly than before. For my sci-fi/fantasy education, that will do.

Designing the algorithm

Alrighty, so we have a dozen or so awards with varying tenures and goals across 50+ years. The challenge now is, how do we combine this treasure trove of knowledge into a single, easy-to-digest ranking?

Choosing the awards

I pulled as many major sci-fi/fantasy awards as I could: Hugo (est. 1953), Nebula (1966), British SF Association (1970), Locus (1971), British Fantasy (1972), Campbell (1973), World Fantasy (1975), Philip K. Dick (1983), Arthur C. Clarke (1987), and Dragon awards (2016). Further, I included the Goodreads Choice Awards (est. 2011) for best science fiction novel and best fantasy novel. The Goodreads Choice Awards fairly elegantly compromise between an editor curated list of nominees (15 nominees via editors) and leveraging the masses (additional 5 nominees, plus voting on the winner). A ton of thanks to The Science Fiction Awards Database and Locus Foundation for their work to catalogue a lot of this data.

Defining our weights

We have to start here, but to be honest, this is (perhaps surprisingly) the least interesting part of the problem. Weighting these awards was less about subjective preference (although I will admit that the relatively new “Dragon awards” are discounted until they demonstrate a bit more tenure), and more about balancing genres and regions. Meaning, a fantasy novel published in the US should have an equal opportunity for points as a science fiction novel published in the UK. Further, it’s important that we don’t reward a committee for dialing up their number of nominations or winners. For example, have a gander at the nominees and winners for the Locus Awards in 1985. You’ll need to scroll a few times to see the full list, and that’s the point. Being nominated for a Locus Award this year meant that you were 1 of the top 57 books (!) published.

The 57 nominations for a Locus Award in 1985

One must wonder what poor books were not nominated? Compare that to the Hugos from the same year:

Hugo Awards, 1985

It would be criminal to weight a Locus nomination as highly as a Hugo this year. Further, perhaps more subtly, the Locus award typically grants two winners — one for each genre. Again, we need to control for this.

The greatest novels of each year (1970–2021)

A ton of data mining and coalescing, sprinkled with the simple algorithm described above, and we can already start to sort books within a given year (we still can’t compare books across years however). Let’s take a look: The following graphics show the top ranked book from every year in our awards data. Note that all years quoted in this article are the year of the award, which is almost always 1 year after the publishing date.

1970–1980: No surprise, Ursula K. Le Guin, Isaac Asimov, and Arthur C. Clarke wrote 6 of the top 10 books in this first decade of awards. These titans of classic science fiction end their reign soon after, but The Dispossessed, The Gods Themselves, and Rendezvous with Rama will certainly show up in our final rankings. Ringworld by Larry Niven, The Forever War by Joe Haldeman, Dreamsnake by Vonda McIntyre, and Gateway by Frederik Pohl all swept the big three awards (Hugo, Locus, and Nebula); a feat that has rarely happened since the 70’s. Where Late the Sweet Birds Sang by Kate Wilhelm won the Hugo and Locus, necking out Man Plus by Pohl in 1977.

1981–1994: The next 15 years appear to capture the birth (and recognition) of great sci-fi/fantasy series. Gene Wolfe wins 1982 with his second installment in his ‘Book of the New Sun’ series: The Claw of the Conciliator (The Sword of the Lictor just barely being eek’ed out in the following year by Asimov’s continuation of the Foundation series). Similarly, the second novel in David Brin’s ‘Uplift Saga’, Startide Rising, wins the Hugo, Nebula, and Locus. Orson Scott Card’s ‘Ender’s Saga’ takes back-to-back years. Dan Simmons takes ’90 and ’91 with the first two installments of his ‘Hyperion Cantos’. Lois McMaster Bujold, Connie Willis, and Kim Stanley Robinson make their first (but far from last) appearances with early novels in their soon to be epic series. Finally, the cyber-punk classic, Neuromancer by Gibson easily secures top marks for 1985, which kicks of his ‘Sprawl’ trilogy.

As alluded to, 1995–2005 contains more of Bujold’s ‘Vorkosigan Saga’ with Mirror Dance, Robinson’s ‘Mars trilogy’ with Blue Mars, and Willis’ ‘Oxford Time Travel’ set with To Say Nothing of the Dog. Neal Stephenson is best in class for the first time with The Diamond Age. The first two novels in China Miéville’s ‘New Crobuzon’ trilogy, Perdido Street Station and The Scar, earn the top slot for 2001 and 2003 respectively. Neil Gaiman’s American Gods, his most awarded book ever, runs away with 2002.

2006–2019: Vernor Vinge shows up for a second time, thanks to Rainbows End. There’s not a clear frontrunner for 2006, but StrossAccelerando takes it. On the other-hand, Michael Chabon’s first (and only?) science-fiction novel, The Yiddish Policemen’s Union, is the clear winner in 2008. Miéville does it again in 2010 with the single most awarded novel in our dataset, The City & The City. Ann Leckie’s Ancillary books win back to back years, just nudging ahead of Cixin Liu’s The Three-Body Problem in 2015. Despite sweeping the Hugo’s from 2016–2018 with her ‘Broken Earth’ trilogy, N.K. Jemisin’s The Stone Sky is her only novel with top marks in our list.

Lastly, we have our two most recent top rated (2020–2021): A Memory Called Empire, Arkady Martine’s debut novel, and Network Effect by Martha Wells. The latter being the first full length novel following a span of award winning novellas in the smash hit ‘Murderbot Diaries’ series.

Journeying across space and time!

Here’s where things get interesting. Coalescing data, choosing awards, determining weights… none of that is all that novel or controversial. The real challenge is time travel! Meaning, how do we look at a book written in the 70’s and compare it to a recent novel? To keep it simple, consider the example of the following two novels published about a decade apart:

In our current algorithm, Gateway is the clear winner, having won all the same awards as Speaker for the Dead, plus pulling out a victory in the John W. Campbell awards (Speaker was nominated, but lost). Further, there were less awards to win (7 in 1978 vs 9 in 1987), so Gateway won a greater proportion.

To truly answer this question though, we need to confront two glaring asymmetries in our dataset:

Asymmetry #1: An ever increasing number of awards

The number of awards has increased over time. In 1972, there were four major awards: Hugo, Nebula, British SF, and British Fantasy Awards. However, a book published in 2019 is eligible for more than a dozen awards in our dataset. The naive solution for handling this would be to suggest that it’s the proportion of awards you win that matters. Meaning, winning 2 of 4 awards (50%) is equivalent to 3 of 6 (still 50%). However, as the number of awards increases, this heuristic seems to break down. This is likely because the probability of winning multiple awards is exponential, not linear. The probability of winning n awards is p(w)^n, not n*p(w). Thus, we will use a logarithmic transform in our algorithm to account for this.

In our Gateway vs. Speaker example above we can see the difference a log makes. Assuming a linear (weighted sum) of awards we’d expect this to result in a 13% boost for Gateway. However, applying the proper transformation only nets us a 6% boost in favor of Frederick Pohl’s novel. Still edging it ahead of Speaker, but not by as much as we might intuit.

Asymmetry #2: Varying levels of competition

Similarly, just like the number of awards has changed over time, so has the competition. While it’s difficult to fully understand how many “good” sci-fi books were being published in a given year, we can look at some examples to get a sense for what competition indicators we might already have in our data.

Consider the Locus Awards in the mid-1970s. Joe Haldeman’s Forever War is nominated for best novel in 1975, and then again in 1976 when he wins the award. Or, look at the 1972 and 1973 Hugo Awards. The talented Robert Silverberg is nominated FOUR (!?) times in two years for the same award. He’s quite literally competing with himself.

Two Silverberg novels nominated for the Hugo in 1972
1973 Hugo Award for best novel (2 more Silverberg’s!)

So, while there’s no disputing Forever War and Silverberg are greats, we do want to try and control for how difficult it is to edge out the competition in a given period. Using a more scaled version of this exercise we can tease out a competition factor, which we use to further refine our algorithm. Our method suggests it has not been perfectly linear, but certainly has increased over-time (as we’d expect).

Graphing the level of competition for sci-fi/fantasy awards since 1970

We can now account for competition when comparing Gateway to Speaker for the Dead. As we saw in the Haldeman and Silverberg examples above, the early 70’s was a relatively low point for competition. This is not to say some of the greatest novels of all time didn’t arise from the 1970’s (they did, and our data will show that), but instead is a commentary on the number of distinct authors writing in the genre at the time. The 80’s, however, is a different story — enter Gene Wolfe, Lois McMaster Bujold, Kim Stanley Robinson, Connie Willis, David Brin, and of course, Orson Scott Card. This additional competition results in a material boost for Speaker.

Where does that leave our two novels? Check out the final rankings below, but (spoiler), Speaker for the Dead leapfrogs Gateway after the above is accounted for.

One algorithm to rule them all

And voila! Here we have it: A function that takes in a bunch of award data, cleans it up, weights it reasonably, controls for the number of awards over time, and properly accounts for the level of competition. The notation is pseudo-math of course, but the final algo looks something like the following:

Now, onto what you have been waiting for…

The top ranked SFF novels of the past 50 years

For those that read all of the above and are still here, cheers. For the rest of you that just skipped straight to the results… as one of my favorite fantasy characters would say, “You have to be realistic about these things.” So without further ado, a ranking of the best sci-fi/fantasy novels since 1970:

Ranks 1–10: Time to knight Sir Arthur C. Clarke again, as he takes the top slot of the past 50 years with Rendezvous with Rama — cementing the “Big Dumb Object (BDO)” construct into science-fiction. It won 4 awards (including a rare singular Locus win) and was nominated for a 5th. I’ve mentioned most of these already, but happy to see that our algorithm returned all distinct authors, heavy hitters from across different genres, and a good mix of decades.

Ranks 11–20: The phenomena that is Martha Well’s ‘Murderbot Diaries’, occupies the #11 spot in our rankings with Network Effect. Interestingly, even though they were written 22 years apart, The Forever War and Forever Peace by Haldeman sit side by side, making them the two highest ranked books that form the same series. Jo Walton’s Among Others and Mary Robinette Kowal’s The Calculating Stars appear higher than I expected going into this, but digging into the data we can see why — both won 3 major awards and were nominated for 6 and 5 (respectively). Only a few novels on this list can boast such recognition.

Ranks 21–30: Uprooted by Naomi Novik narrowly lost the Hugo award in 2016 to The Fifth Season, but a victory there and we would likely be talking about it as one of the greatest fantasy stories ever (having won the Nebula, Locus, British Fantasy, and nominated for 2 others). It’s rewarded with the 21st spot in our rankings. The Windup Girl by Bacigalupi shared the Hugo in 2010 with The City & The City, but picked up 2 more wins and 4 more nominations to land it at #26. Again, we see more greats we already touched on to round out the top 30 — Ender’s Game, The Left Hand of Darkness, Doomsday Book.

If you don’t see your favorite book, have no fear. This is just the top 30 — out of the 3,000+ I analyzed. Feel free to drop a comment and I will try to pull some numbers on how your favorite stacked up.

The top ranked series

You probably noticed that a lot of these novels are part of a series of books. So, let’s aggregate the numbers and figure out which series come out on top. To qualify for this list, the series needs to at least be a trilogy and have 2+ books with sufficiently high scores (one great novel followed by duds does not qualify as a great series… looking at you Rendezvous with Rama!). To keep things simple, I have ranked them by summing the total number of points all books in the series earned. That does, unfortunately, mean that the longer the series the more points it will naturally acquire. Thankfully, only the ‘Vorkosigan Saga’ by Bujold seems to really abuse this system (with 17 books), so while epic, do take that one with a grain of salt.

At last, we see ‘A Song of Fire and Ice’ (aka A Game of Thrones) by George R. R. Martin. Robinson’s ‘Mars Trilogy’ gets acknowledged a bit more explicitly. And of course we find Gene Wolfe’s masterwork, ‘Book of the New Sun’ at #2. A few honorable mentions that just missed the cutoff: ‘The Uplift Saga’ by David Brin, ‘New Crobuzon’ by China Miéville, and ‘Hyperion Cantos’ by Dan Simmons.

The top ranked authors of the past 50 years

Now that we’ve covered the books in nauseating detail, we can finally review where that leaves the authors. Looking at the past 50 years of data, which authors have dominated science-fiction and fantasy? I’ve included the number of wins and nominations (in our dataset), their top ranked novel within our algo, as well as the stretch of years in which they received their first and last (so far) award nominations.

‘Gene Wolfe takes the top slot with an absurd 56 nominations and 9 wins over the past 50 years. Mostly, as we saw above, for his ‘Book of the New Sun’ series. Bujold as well absorbs a ton more awards with her ‘Vorkosigan Saga’ novels, earning her #2. The most wins goes to China Miéville (albeit a handful of these are lesser valued UK-specific awards), followed closely by Ursula K. Le Guin — whom likely would be #1 if we went back a few more years. Greg Bear earns the distinction of lots of nominations (42!) with very few wins (only 2), putting him at #13. Jemisin (#6), Leckie (#16), and Stephenson (#17) will surely keep climbing these rankings with new work still being published (assuming Leckie doesn’t continue withdrawing nominations, as she did with The Raven Tower). Perhaps surprisingly, Stephen King does quite well for himself in the fantasy category as well, rounding out this list at #30 (and just barely edging out John Scalzi).

Always a bridesmaid, never a bride

Lastly, I’ll leave you with one of my favorite insights from this exercise: Here are the top 3 books that never actually won an award. Years from now these novels might be forgotten, but being nominated for nearly every award (even without winning) has got to signal something worth reading:

#94: Blood Music by Greg Bear

#95: Piranesi by Susanna Clarke

#101: The Ten Thousand Doors of January by Alix E. Harrow

That’s it folks. Follow for an update each year with the latest batch of winners included. This data set contains over 3,000 novels, so if I didn’t write about one you are interested in — just comment and let me know, and I’ll try to respond with some stats. Also, if there’s any awards I left out (that have more than a couple years of data), please let me know and I’ll look into including next time. Thanks for reading!

P.S. If you were wondering, there has been very little award-love for either Ian Banks’ ‘Culture’ novels or Steven Erickson’s epic fantasy series, ‘Malazan Book of the Fallen’. Don’t take it to heart though, one of my personal favorite series, Pierce Brown’s ‘Red Rising’ saga, didn’t make the cut either.

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Cassidy Beeve-Morris

Cassidy Beeve-Morris

143 Followers

Passionate about statistics, understanding the economy, protecting the environment, and building AR/VR products at Facebook.