Analysing 5 Years of YouTube Stats (Patreon)

Hey everyone!

So, with GMTK turning five years old, this seemed like the perfect time to do a big overview of the channel’s numbers to see what worked, what didn’t, and what can be learned for the next five years of GMTK.

This is going to be a big, detailed, nerd-tastic article and so I won’t be offended if you skip this one. But I personally find this stuff fascinating and it’s really important sometimes to take a step back, and figure out where I’m going with all of this. 

A couple quick disclaimers before I start:

- All traffic is relative. When I say a video is successful or disappointing, I mean in terms of GMTK’s traffic, not YouTube as a whole.

- View counts alone don’t drive my decisions on GMTK. A show like Designing for Disability might not receive much traffic, but I’m going to make it anyway because it’s important.

So, since releasing my first video (Adaptive Soundtracks) on November 6th, 2014, I’ve released 124 videos. Here’s the lot, in order of release

Here’s how many videos I made each year:

• 2014 - 2
• 2015 - 22
• 2016 - 27
• 2017 - 28
• 2018 - 24
• 2019 - 21 (so far)

2017 was my most productive year because I turned GMTK into my full-time job. But I also worked way too hard that year and threw my work-life balance into the garbage. 

2018 and 2019 saw me try to fix that (I got engaged, I moved twice, I learned to drive, I stopped working weekends (mostly)). But I’ll always try to hit 24 videos a year: a two-a-month average. I plan to hit that in 2019, with three more videos before the end of the year.

Which videos got the most views

Okay, so let’s shuffle that image up.

Here’s every GMTK video released in the last 5 years, ranked from most-viewed (Protecting Players from Themselves) to least-viewed (Pac-Man).

The top 25 are:

1. How Game Designers Protect Players From Themselves (3.2 million)
2. What Capcom Didn’t Tell You About Resident Evil 4 (1.6 million)
3. What Makes Good AI? (1.5 million)
4. How to Keep Players Engaged (Without Being Evil) (950k)
5. How Cuphead’s Bosses (Try to) Kill You (900k)
6. Half-Life 2’s Invisible Tutorial (850k)
7. How Games Do Health (800k)
8. What Makes a Good Puzzle? (800k)
9. Super Mario 3D World’s 4 Step Level Design (750k)
10. Do We Need a Soulslike Genre? (750k)
11. The Rise of the Systemic Game (750k)
12. The Design Behind Super Mario Odyssey (700k)
13. Legend of Zelda: Breath of the Wild - An Open World Adventure (700k)
14. Morality in the Mechanics (700k)
15. The Magic of the First Legend of Zelda Game (700k)
16. What Makes a Good Combat System? (650k)
17. Nintendo - Putting Play First (650k)
18. The World Design of Dark Souls (650k)
19. What Makes a Good Detective Game? (650k)
20. Following the Little Dotted Line (650k)
21. Why Metro Exodus is 2019’s Most Immersive Open World Game (600k)
22. Analysing Mario to Master Super Mario Maker (600k)
23. What Makes Celeste’s Assist Mode Special (600k)
24. Redesigning Death (600k)
25. Anatomy of a Side Quest: Beyond the Beef (600k)

I’m not too surprised about any of these videos. 

They feature big, popular, well-loved games (Cuphead, Dark Souls, Zelda, Fallout, Half-Life). They are some of my most highly-regarded videos (Mario 4-step Level design, Invisible Tutorial, Putting Play First). And a number have, arguably, clickbait titles (“How Game Designers Protect Players From Themselves”, “What Capcom Didn’t Tell You About Resident Evil 4”).

Some of these videos also do something really important, and that’s to reveal something genuinely new and interesting about game design - like Mario’s “secret” level design and Resi 4’s “secret” adaptive difficult. Not that either were completely unknown, but I believe my videos were the first time such facts were made widely known.

That’s very easy to make viral (“OMG, did you know THIS about Resi 4…”), but coming across that info is hard. It’s something I’ll continue to try and do, but it’s not something that can be achieved with every video!

Also, it's interesting that in these top videos, there's more "Topic Videos" (a game design topic, through the lens of lots of games) than "Analysis Videos" (videos on a single game).

There's about 14 topics (protecting players, AI, engagement, health, puzzles, genres, systemic, morality, combat, Nintendo, detective games, waypoints, assist modes, and death), and about 11 games (Resident Evil 4, Cuphead, Half Life 2, Mario 3D World, Mario Odyssey, Breath of the Wild, Zelda 1, Dark Souls, Metro Exodus, Mario Maker, Fallout New Vegas).

In terms of the lesser-viewed videos: well some are very new and haven’t had a chance to get rolling yet. But the others are not too surprising: smaller indie games (Downwell, Toki Tori 2, Nova-111, Her Story), or topics that don’t interest as many people (Accessibility, game history).

Some surprising ones, to me, are the videos on God of War and Spider-Man, which were released around the time of their launch. Doing a video on a new game is quite hit or miss, it turns out: you can either ride the hype train (as I did with Metro Exodus, Mario Odyssey, Cuphead, and Breath of the Wild), or you can have people skip that video because they haven’t played the game in question yet.

The Celeste controls video (31st most viewed video) goes to show that I can be a year late and still get views. There’s no need to rush things to make a video “current”: when it comes to design, people don’t really care if I’m talking about the most up-to-date games. 

When did these views happen?

I get most of my views when a video goes live. 

This graph from the video “What Makes Good AI?” shows that the video got a big whack of traffic (180,000 views) upon release, but then will get a few hundred views every day (with some nice spikes here and there), which can really add up in the long run (to 1.5 million).

Other videos might get seen a little later in their life. This Resident Evil 4 video from 2015 did okay at launch (30k views), but then suddenly got a huge spike (230k) much later in 2016. If you’re wondering why, the answer is almost always: the algorithm.

This video randomly got chosen by the algorithm and was blasted out to millions of YouTube channels on their home page. They clicked, and the rest is history.

Here’s another chunky video: Protecting Players from Themselves. It did really well at launch, and then had small spikes and residual views ever since. 

To really visualise how this works, here’s a nifty bar chart race showing how traffic has grown and changed since 2014. 

Which videos led to subscribers

It’s all well and good having a massively popular video, but it’s not worth much in the long run if those viewers don’t hit the subscribe button and stick around for your future content!

So I’m going to shuffle the list again. 

Here’s the videos that got the most subscribers, from most (Protecting Players from Themselves) to least (Metroid Prime 3).

In terms of numbers, outside of the outliers (Protecting Players and Good AI delivered 20,000 subscribers each), a typical well-performing video will generate about 5,000 new subs, while a less successful video might only give me a few hundred (Metroid Prime 3 only generated 261 new subscribers).

This image does show a few interesting things, like how the Celeste controls, DOOM enemies, and Shovel Knight nostalgia videos all lead to lots of subscribers, even if they didn’t have the most views.

These numbers are going to be biased by the number of views, of course. The more views a video gets, the more opportunities there are to get new subscribers. So maybe I can make something more useful by figuring out the percentage of views a video got, in relation to its view count. 

Here’s that:

Ah wow! Data! So, the top 25 videos now are:

1. What Makes Good AI?
2. Shovel Knight and Nailing Nostalgia
3. How to Keep Players Engaged (Without Being Evil) 
4. Why Does Celeste Feel So Good to Play?
5. What We Can Learn From DOOM
6. What Makes a Good Puzzle? 
7. What Makes a Good Combat System?
8. How Game Designers Protect Players From Themselves
9. Why Nathan Drake Doesn't Need a Compass 
10. The Design Behind Super Mario Odyssey
11. How (and Why) Spelunky Makes its Own Levels 
12. Hitman, and the Art of Repetition 
13. How Games Do Health
14. Theme and Mechanics in Far Cry 2 and Far Cry 4 
15. Nintendo - Putting Play First
16. The Rise of the Systemic Game
17. The Legend of Zelda: Link to the Past’s Dungeon Design | Boss Keys 
18. Following the Little Dotted Line
19. The Magic of the First Legend of Zelda Game
20. Morality in the Mechanics
21. Super Mario 3D World’s 4 Step Level Design
22. What Makes Celeste’s Assist Mode Special
23. The World Design of Super Metroid | Boss Keys
24. Half-Life 2’s Invisible Tutorial
25. How Jonathan Blow Designs a Puzzle 

Now this is really interesting. 

For example: look how far down the Resident Evil 4 video is. It’s my second most popular video (1.6 million views), but while it generated a bunch of subscribers (5,400), that’s only a 0.3% attach rate. Compare that to Good AI, which also got a lot of views (1.5 million), but a lot more subscribers (20k) - meaning 1.3% of people who watched hit the subscribe button. 

So this is a real display, I think, of video quality. Clickbait (“What Capcom didn’t tell you about Resident Evil 4”) can get people in the door, but it doesn’t guarantee they’ll stick around. Not that that video sucks, but the numbers show that I’ve clearly made better videos than that. 

So real high quality videos with insight (What make a good puzzle?), high production values (Shovel Knight), developer interviews (Celeste), and so on is what drives people to go from viewers to subscribers (and, ultimately, Patrons). Which is good, because that’s what I’m trying to make anyway. But it’s nice to know that it’s a valid business model!

Another thing of note: look at how the Space Invaders Design Icons goes from being one of my worst-performing videos in terms of views, but jumps up to the big leagues in terms of subscriber attachment. This shows that while not as many people watched the video, the people who did watch, did at least subscribe because they wanted to see more episodes in the series. Likewise, the first episode of Boss Keys (Link to the Past) is in the top 25.

This shows the importance of serialised content: it encourages people to hit subscribe so they’ll stick around. Even designing for disability gets a big jump! Though, the Dead Space trilogy underperformed in this regard.

Speaking of series…

Which type of videos have I made

I’ve made 124 videos. That breaks down to:

• 91 episodes of Game Maker’s Toolkit
• 21 episodes of Boss Keys (13 Zelda, 5 Metroid, 1 each of Castlevania, Dark Souls, and Hollow Knight)
• 3 bonus videos (Mario Maker, Zelda music, turn timers)
• 3 Game Jam results videos
• 4 Designing for Disability
• 2 Design Icons

What we can do, to try and get some idea of how these different series perform, is to look at the average number of views an episode gets.

So, in order of best to worse:

1. GMTK - 542,000 per episode
2. Game Jam - 479,000 per episode
3. Boss Keys - 402,000 per episode
4. Bonus videos - 337,000 per episode
5. Designing for Disability - 222,000 per episode
6. Design Icons - 167,000 per episode 

So, in terms of views, it’s clear that GMTK is where it’s at. I’ve had so many ultra successful episodes of GMTK, that the average is over half a million views.

Designing for Disability and Design Icons have been quite disappointing in that regard, but Boss Keys has been more successful, with Dark Souls even making it into the top 25 videos.

The Game Jam results videos are also really popular, which is a little bonus on the top of what is already a super fun experience.

As I said before, these serialised shows are good for attracting more subscribers, but maybe not as hot as I’d have assumed. Part of the reason for making Design Icons was to make a series that would encourage people to stick around and while the Space Invaders ep had a good attach rate, it still didn’t do as well as normal GMTK episodes from this year like Celeste and Metro Exodus.

Of course, it’s early days with Design Icons. These ancient arcade games are a hard sell, and things might pick up a bit when I get to slightly more modern games.

But still, I think this goes to show that while these series are important (they generate subscribers, they broaden the audience of the show, they keep my enthusiasm up), I can’t let them take over GMTK. And, lately, that is what has happened. 

Here’s the breakdown, ignoring bonus videos and the game jam. 

• 2015 - 21 GMTK
• 2016 - 21 GMTK, 7 Boss Keys
• 2017 - 21 GMTK - 4 Boss Keys
• 2018 - 15 GMTK, 5 Boss Keys, 3 Designing for Disability
• 2019 - 13 GMTK - 4 Boss Kets, 1 Designing for Disability, 2 Design Icons

I’m going to discuss this more in the 2019 stats wrap-up post, but clearly something’s not right here. And it’s the main thing I plan to address in 2020.

How much watch time did these videos get?

One final stat that’s really important is watch time. 

Now, for YouTube’s purposes, long videos are better. I think (it’s hard to know for sure) that YouTube would rather people watched 10 minutes of an hour long video, than all nine minutes of a nine minute video.

That’s not how I roll, though: a video needs to be as long as it needs to be. Hollow Knight’s Boss Keys needed to be 40 minutes long, but most episodes only need about 12 minutes to say what I need to say. I try to be precise and succinct with my scripts. 

So I’m more interested in watch percentage. How far through the video does the average person get? Happily, my average is 64%. That’s really high in YouTube land: that means people actually want to keep watching my videos, and don’t just get bored and leave.

There’s only a few outliers. 3 videos have less than 50 percent average views: 

• Hollow Knight Boss Keys. This video is really long, and is super spoilerific. Two qualities that can turn people away. 
• Why Does Celeste Feel So Good to Play? This one’s a bit disappointing. The video is long, and maybe the developer interviews run a bit long at times. But still: it’s a good video. Sad about that.
• 2019 Game Jam - Okay, this one did end up getting a bit long. But I don’t really care - I want to celebrate these games and will continue doing so!

Where did these views come from?

So, in the last 5 years I’ve had 63 million views. Luckily, we can break that big ol number down into smaller bites.

21.3 million views - Home page

The home page delivers videos to two key people: subscribers and people who YouTube’s algorithm think might enjoy my videos. I’ve been very lucky to be seen favourably by YouTube’s algorithm (my content is advertiser friendly, and has good clickthrough rates thanks to good names and thumbnails), so the bulk of my traffic comes from here. 

18.3 million views - Suggested videos

Also an algorithmic thing here. This is when a GMTK video appears on the right side of a video. These views are almost exclusively from GMTK videos being advertised on… other GMTK videos. So having a big back catalogue of enticing videos is really helpful. Once someone watches one, they’ll be nudged to watch another.

6.3 million - GMTK YouTube channel

Lots of traffic comes from the actual channel. I presume this is from people watching a video, clicking my user name, and then finding something new to watch from there. 

3.5 million views - Subscriptions

This is from people who get to my videos from the subscriptions page on YouTube, but doesn’t cover other places subscribers might see my videos such as notifications or the home page. 

3 million - Playlists and Playlist Pages

These are mostly my own playlists, but a few viewer-made ones too.

2.5 million - External

No matter how many articles people write, Tweets people share, or Reddit posts people upload, the external links are only a tiny percentage of my overall views.

Of those, Reddit’s the big dog, delivering GMTK some 600,000 views over the last 5 years. Google Searches, Twitter, and Facebook are good. Polygon, NeoGAF, and Kotaku have also been helpful.

2.3 million - YouTube search

Self explanatory, this one. The most popular search terms are

1. Mark Brown
2. Game Maker’s Toolkit
3. Boss Keys
4. GMTK
5. GameMakers Toolkit
6. Game Maker
7. Game Jam
8. Metro Exodus
9. Game Makers Toolkit
10. Game Design

It’s nice to scroll down and see that people searched for games like Breath of the Wild, Obra Dinn, Hollow Knight, and Dark Souls, and came across my stuff.

1.5 million - “Direct or unknown”

Bookmarks, directly typing a URL into your browser, and other weirdo sources.

800,000 - Watch Later

People can send videos to a “watch later” playlist, and then view them later. Thankfully, people actually watch the videos and don’t do what I do: chuck them in there and forget about them forever. 

640,000 - Notifications

Bing! Hit that bell icon, peeps. 10.5% of my subscribers have “all notifications” on, and 6.6% of my subs have enabled YouTube notifications. 

620,000 - YouTube Remote 

I actually have no idea what this is. According to Google, it’s used for sending YouTube videos from your phone to your TV. I think I’ve done that once or twice. 

The rest:

Things like watch history, the trending page, end screens, video cards, and other weirdo things. 

Who is watching GMTK?

So, I know from feedback and tweets and whatnot that game developers, studios, universities, etc watch GMTK. But what can YouTube tell us about the show’s demographic across this half decade?

Here’s an interesting one. There’s about a 50-50 split on views of my videos, between people who are subscribed to the channel, and people who aren’t. More non-subscribers watch the videos, but subscribers are likely to watch the videos for longer. Good on you, subs.

The show’s viewership is overwhelmingly male, with 95.8% of viewers being dudes. I’m not sure I’m ever going to move that needle very much, sadly.

My videos are definitely intended for a college / young professional type demographic, so I’m hitting that. Good to be talking to people my own age, too: easier than trying to “get down with the kids”.

America is, predictably, the biggest consumer of GMTK videos. But pleased to see the motherland in second place. Other countries where GMTK is big include Canada, Germany, Australia, Brazil, France, Sweden, Netherlands, Spain, Russia, Mexico, and India. Unrelated: I’ve visited the top 4 countries where GMTK is most popular! 

We can also see that Windows is the most popular (40% of all views), followed by Android (27%), iOS (4.6%), and Mac (6.4%). People do watch my videos on PlayStation, Xbox, Switch, Linux, Wii, and even the PlayStation Vita and 3DS, but the numbers are comparatively small. 

More people watch on their computer (48%) than phones (35%), consoles (5%), tablets (5%), or TVs (4%).

The key lessons

Okay, so I've included some lessons and learnings throughout the article, but as way of a recap:

I get the most views, subscribers, and subscriber attachment from Game Maker's Toolkit videos, as oppose to other series. Especially ones on important game design topics like AI, health, and engagement, or on popular games like Cuphead, Mario, or Half-Life. Therefore, these must be the beating heart of GMTK.

Other series are important for lots of reasons, but little and often is perhaps the way to go. Lower-performing shows like Design Icons and Designing for Disability are cool, but can't take too much focus away from GMTK. More on this in my 2019 recap next month.

Going after brand new games is hit or miss, so I shouldn't feel pressure to be "current". GMTK works just fine when I'm talking about year old games, or 10 year old games for that matter. 

Video quality seems to be the key defining factor for getting subscribers, which is how things should be. So while I would never chase hits with hugely clickbaity videos anyway, it's at least a good reminder that what I'm doing works. Quality can win out on YouTube, and that's very heartening.

Thanks for reading. If you have any thoughts, or any questions, please drop them in the comments below!

Mark