Can we predict what games will be hits on Twitch?

Here's the plan: Figure out what Twitch's most popular games have in common, and then predict with perfect accuracy what games are going to be popular among livestreamers in the future. Never again will we be caught off guard by the sudden popularity of a game like Among Us. The whims of PC gamers will become obvious to us, as if translated into Matrix code.

For reasons I'll explore at the end of this article, this plan is doomed to failure. But it'll be fun to try anyway. We know which games we wish would get big on Twitch, but which games will it actually happen for? Here are our best guesses.

What games are already popular on Twitch?

We obviously need to start by taking stock of what's already popular on Twitch. Some of the biggest games by concurrent viewers are: 

• League of Legends
• Rust (a streamer server is popular right now)
• Minecraft
• Fortnite
• Valorant
• GTA Online (especially roleplaying)
• Escape from Tarkov
• Call of Duty: Warzone
• Hitman 3 (just released)
• Counter-Strike: Global Offensive
• Among Us
• World of Warcraft
• Rocket League
• Dota 2
• FIFA 21
• Virtual slot machines and other gambling
• Apex: Legends
• Path of Exile
• Sea of Thieves
• Hearthstone
• Mario Kart 8
• Dead by Daylight 

That's not a comprehensive list of Twitch staples, but it's a good sampling. Other games, such as Genshin Impact, Call of Duty: Black Ops – Cold War, Rainbow Six Siege, and Dead by Daylight, move in and out of the top concurrent viewers list. Recently, I saw Sea of Thieves sitting up near Rocket League, but it was primarily because xQc was streaming it, so it matters who's playing. (I ignored games that were only near the top because one person with a huge audience was streaming them.) There are also games that have only recently settled into lower numbers, such as Phasmophobia, which was streamed heavily in 2020. 

What obvious things do these games have in common?

• Most are multiplayer games.
• Most are competitive and cooperative, with small teams.
• High tension moments (eg, the final circle in a battle royale game). 

There are exceptions, of course. Hearthstone isn't cooperative, and Minecraft can be quite peaceful. Roleplaying seems to fill in for structured competition in GTA Online and Rust. For the most part, though, competition, cooperation, and tension are key ingredients in many of Twitch's most popular games. Games which lack those qualities, such as Civilization 6 and Euro Truck Simulator, don't get big ratings on the platform even if they're popular games. Even Cyberpunk 2077, one of the most anticipated games in recent years, only has around 5,000 viewers. Hitman 3 is clearly up near the top because it's new, not because it's a special exception (although it is good).

What else do these games have in common?

The preference for multiplayer games on Twitch makes obvious sense (they're already shared experiences, so streaming them is less of a leap than it is with singleplayer games), and it isn't all that useful to point out. So, what else seems to be the case about games that are popular on Twitch? Here are a few observations:

• Most of them are backed by big publishers.
• Many are free-to-play. 
• Many include cosmetic items.
• There doesn't seem to be any preference for fantasy, science fiction, or modern day settings. However, 10 out of the 22 games I listed heavily feature guns.
• Building is important to three of the top games: Rust, Minecraft, and Fortnite. 
• Several of these games take place on large, primarily outdoor maps. I'll also add that many of these maps contain green grass. Everspace 2, which just released in early access, is nowhere near the top of the charts. No space games are, in fact.
• Rocket League is very popular, and World of Warships is somewhat popular, but otherwise there's a clear preference for human or human-like player characters. Games in which we're looking at the back of a spaceship, car, truck, tank, or warship are not often found among the top Twitch games. 
• They're most commonly played with a mouse and keyboard, with some exceptions, such as Rocket League (which is turning out to be quite an exceptional game).
• Where applicable, victory results from a clear binary condition, and ties are impossible. Either you're the last team standing in a battle royale game, or you're not. This is in contrast to older deathmatch games, where winners were determined by a ranking based on kill count, and other games where players accrue points over time. (CS:GO and Valorant offer a sort of mix, with multiple objective-based rounds that add up to a final team score.) 

Predicting this year's Twitch hit: Chivalry 2

Based on what we know right now about the games releasing this year, I expect Chivalry 2 to be one of the most popular 2021 games on Twitch. It has big outdoor maps, cooperation, competition, and takes some skill. If a mode with smallish teams can generate the tense moments characteristic of Twitch's most popular games, and it includes a simple, binary win condition, we'll may see it settle into the top 20 for many months. Mode design will be the key.

Confidence in this prediction is helped by Mordhau's former popularity, but I think we could've gotten to it using only commonalities with other games.

What games probably won't blow up on Twitch?

Star Wars Squadrons: Like Everspace 2, Squadrons may be a good game, but Twitch has a clear preference for environments with atmospheres. We also observed a preference for seeing people, not vehicles like spaceships. The 5v5 dogfights also have the wrong kind of win condition: Reaching a certain kill count doesn't seem to be what grabs viewers. Squadrons only has a couple hundred concurrent viewers right now, and I wouldn't expect to see that needle move much.

Back 4 Blood: Even if this spiritual Left 4 Dead 2 sequel is excellent (I really hope it is), I'll be surprised if it tops the Twitch charts for long. Like Deep Rock Galactic, it's dark, and it's cooperative but not competitive.

The next Battlefield game: Based on our observations, you'd think Battlefield 5 would be popular on Twitch. It comes from a big publisher, and it's got guns, large maps, teamwork, and competition. But it's got big teams, not small ones, and the win condition in the popular Conquest mode is based on respawn tickets, not a simple yes or no question, like whether or not the bomb goes off in CS:GO. (Perhaps these things will prove to be problems for Chivalry 2, as well.)

Games with potential

Fallout 76: Could we be in for a surprise spike in Fallout 76 viewership at some point? It's a few years old now, but Rust is even older, and that just shot to the top of Twitch on account of a streamer server. I think it could still happen for Bethesda's online Fallout spin-off, even though it's not highly competitive, because there are fun opportunities for roleplaying. Interesting question: Would Fallout 76 have been more popular on Twitch if, rather than its post-apocalyptic drabness, it featured a lusher environment?

Ark: Survival Evolved: Likewise, if it can happen for Rust, it seems like a Twitch spike can happen for any popular multiplayer survival game.

Icarus: If it releases, Dean Hall's upcoming survival game could be a winner. Icarus doesn't appear to have a competitive aspect, but its structure—you drop to a planet surface with a limited amount of time to complete your mission—could help it generate tension. It's got building, too, and the planet it takes place on looks quite large and lush in places. It'll need opportunities for roleplaying if it's to stick like GTA Online has.

Tribes: Ascend: This game had it all: teamwork, competition, open spaces, guns, green fields, building (in a small way), tense moments. And yet Tribes: Ascend never quite took off, and today it's more or less dead. There were problems with Ascend that had nothing to do with its format, but it does bring team size and win condition  into consideration. Could tweaking those things—say, by lowering the team size and resetting the game after each flag capture—have helped Ascend find a bigger audience, even though Tribes fans would've been furious?

Project Winter: Based on these simple observations, it's hard to say why Project Winter wasn't a huge Twitch hit instead of Among Us. They're similar games, but Project Winter has a much bigger, outdoor map, and involves more roleplaying. It's also more complex, though maybe that worked against it. (Or maybe Project Winter is just waiting to blow up in popularity itself?)

Halo Infinite: We'd have to see what its free-to-play multiplayer modes look like. Last we heard, a battle royale mode wasn't happening, but if it otherwise features competition between small teams, the delayed shooter could become a new Twitch mainstay.

Red Dead Online: There's a standalone version of Red Dead Online now, but the game lacks something that's keeping it from finding the same streaming success as GTA Online. If it finds that roleplaying potential—and maybe it really has to come from modders—it's easy to see RDO becoming a sensation, too.

In the end, it'll probably be something we don't see coming

The impossibility of making strong predictions is plainly obvious: If it were possible to accurately predict what's going to be a hit and build to that specification, then a big company like Amazon would just hire the best predictors around and never release flops. And yet Amazon's Crucible was a total bust. Tons of other duds release every year. 

Take Ubisoft's attempt at battle royale, Hyper Scape, which hasn't found a big audience like Warzone or Apex Legends. With what I've observed here, the best I could do is argue that instead of a city setting, Ubisoft should've put it in grassy fields. It's hard to believe that a variable like grassiness has anything to do with Hyper Scape's struggle to go big. The genius is in the execution of the details—all the ineffable things included in the phrase "a fun and exciting game"—not just observation of superficial features.

Additionally, what will be popular in the future can have very little in common with what's already popular, or at least not have similarities that are obvious to an observer in the present. You can draw a line between Team Fortress Classic and Fortnite, but it's unlikely that anyone in 1999 had a fully-formed vision of the battle royale genre in their head. The free-to-play business model alone would've confused a 1999 person, who would wonder if you were talking about shareware, or a kind of demo. What's popular in a year or two might be something that has nothing obviously in common with the games that are popular now, that we could only predict by making wild leaps that might actually appear to be bad predictions.

Predictions also run into knotty questions about cause and effect. Did Among Us become popular because popular streamers streamed it, or did popular streamers stream Among Us because it had all the makings of a popular game? Is Ninja popular because of Fortnite, or is Fortnite popular because of Ninja? Are both things true at the same time? How do we resolve that?

In the end, human behavior is hard to model, and we can make educated guesses about the future, at best. It's fun to play Nostradamus, though, and maybe I'll end up being right about some of this. I genuinely think Chivalry 2 will be big on Twitch, at least briefly, and that there's something to the fact that so many popular spectator games feature big open spaces, grass, human players, and a simple scoring system. The most popular spectator sport in the world, soccer, also features those things. (Maybe we're just obsessed with grass as a species.)