Storytelling in Esports: It's Pro Wrestling, with Data

This was my half of a GDC 2019 talk discussing how to create narratives in eSports through analytics and pro wrestling. It was given alongside Adam Savidan, of LoadingReadyRun, who covered the pro wrestling aspect.

This talk is not actually the talk I gave, it was the talk I intended to give. Technical difficulties and nerves resulted in a different talk that covers the basics listed out here, but in a much less eloquent way. So I'm publishing the prepared speech for people to peruse at their pleasure.

Why do we watch esports?

If you ask a person this question, chances are you’ll get one of three answers:

I like the team or player and want to see them succeed
I like to find interesting outliers and see the peak of human skill and conditioning
I like the sport and enjoy learning more about how it works

All of these answers tie into an underlying principle: people who watch sports want to experience the tension of competition without actually competing themselves. That tension arises from the gulf between our expectations and reality, and by analyzing what people get out of spectating we can better understand how to generate tension.

You can see this reflected in how audiences react to “bad matches”. If a match is a complete rout, or goes the way everyone expects - like the Patriots-Rams 2019 Super Bowl - the audience loses interest because there’s no tension. Likewise, if a match is fairly even, or there is a significant turnaround for the underdog - like the infamous Evo Moment 37 - the audience becomes more invested, because that’s great dramatic tension.

If you know the conclusion before the game is played, there’s not much reason to play. As immortalized by Avery Brooks in the first episode of DS9:

“Every time I throw this ball, a hundred different things can happen in a game. He might swing and miss, he might hit it. The point is, you never know. You try to anticipate, set a strategy for all the possibilities as best you can, but in the end it comes down to throwing one pitch after another and seeing what happens. With each new consequence, the game begins to take shape.”

So how can we use numbers and statistical data - something that most people associate with the droll and tedious - and use it to make a sport’s shape more tense? How do we contextualize competition?

Hidden Information, Human Drama

For a good example of ways you can use data to make player success or failure more interesting, let’s look at the exciting world of poker.

Poker is a game of luck and bluffing. You get random cards, and if they are good enough (or you think you can trick your opponent into thinking they are good enough), you win the hand. Unfortunately, this is not terribly exciting to watch without aid. You see the players make their wagers, but there wasn’t much analysis you could do as a spectator beyond that.

However, with the invention of hand probabilities and hole-card cams, poker’s popularity with spectators exploded.

As a game of secrecy, the tension of playing poker is derived from guessing how likely your hand will turn out well, and guessing how likely the opponents’ hands turned out well. But only the most attentive in an audience understand the bets being made and the chance of success without both probabilities and hole cams.

By surfacing hidden information that players have to guess, poker broadcasters did something particularly genius: they revealed the tension between what is real, and what the players think is real.

That gulf of difference is where the excitement is in spectating poker. Now that you know a player has a bad hand and is bluffing, you can better follow their thinking. When you spot a bad play, you can understand it’s a bad play, and also get excited when it wins anyway. And when a player like Phil Hellmuth gets angry, you know how ridiculous it is. It’s like watching a movie with a horrible misunderstanding: the dramatic tension hinges upon you having more information than the actors.

This is our first takeaway: Tension is accentuated by exposing hidden or assumed information to the audience but not the players. A more informed audience is more emotionally invested in what happens to your players, because they have the spectator equivalent of Cassandra’s curse.

Tangent: As part of informing your audience, please ensure you have one play-by-play commentator and one color commentator. For this talk, I watched a significant amount of eSports commentary and analysis, and there’s a major problem with commentators trying to do more than one thing and doing neither effectively.

On The Edge Of Human Ability

But what about audiences who aren’t really interested in the human drama angle?

Let’s take a fictional strawman I have constructed for this purpose. He likes basketball, but he doesn’t really like basketball as a sport. Instead, he watches whichever teams have the widely-accepted best players in the league, because he wants to see the sick dunks and unbelievable three-pointers that a basketball player in their prime can score.

For our hypothetical spectator, the tension of watching the sport comes from seeing people defy or validate expectations. If LeBron James is outstripped by a rookie, for example, that makes for a compelling story. Everyone loves a winner, and everybody loves an underdog.

Outliers make for compelling visuals and stories too. When you’re looking at a graph, and you see one little data point far from any other data points, that sparks your curiosity. You want to know more about why that data point exists.

For a real life example with a tragic angle, look at the story of Kim Se-yeon, also known as Geguri. A very skilled Overwatch player, Geguri was unfairly and publicly accused of cheating because her accuracy as Zarya was absurdly good. Vindicated through a public demonstration of skill she never should’ve had to do, she now plays on the Shanghai Dragons. With a more supportive eSports scene, her incredible accuracy would’ve made for a much more uplifting story.

Behind every absurd accomplishment or grossly negligent mistake, there is a story to tell. The reasoning of the player, their personal history, the events leading up to the fight; all of it is important. This is analogous to hidden information for the poker audience: by finding and telling the human stories that give rise to outliers, we contextualize them, humanize them. Jon Bois is particularly good at this.

This is our second takeaway: You can use outliers to find and contextualize interesting human stories. They provide you with an avenue to discover the tensions to promote with pro wrestling techniques.

Middle Finger To The Meta

Those who don’t care for players succeeding or the peak athlete performance, however, get their excitement from a much stranger source: the game’s meta.

For these audience members, tension primarily arises from how optimal or unusual a strategy is. While these are outliers, they are not quite the same as somebody doing well within the bounds of expected play. Rather, their expectations come from how players adhere to the way the game is played, and when players deviate from that formula and succeed, it excites them. Think Luffy playing Street Fighter with a PS1 pad while most players use arcade sticks.

For a spectator who derives tension from meta, the more novel your data set, the better.

Take this series of heatmaps from all 9 de_inferno matches at the FACEIT Major 2018. Each one shows the team color of the shooting player and the direction they were facing, and attenuates line transparency to damage dealt (no damage, no line).

The all shots fired in the match dataset is, naturally, huge.
The AK47 is a very commonly-seen Terrorist rifle, so while the dataset is smaller, it's still too large to find interesting snippets to pull.
The Scout rifle is not very commonly seen in competitive play, so it has a dataset you can understand immediately. Using a Scout rifle is generally not a good idea in CSGO, because it is very weak compared to its brother the AWP. But that makes the kills so much more impressive, so they make for good highlights.

For organizations who don’t have access to automatically-gathered data like these replay heatmaps - for example, if you are not affiliated with the developer, or there is insufficient API support - all is not lost. You can always gather data the old fashioned way: notes. Physical sports offer the best example of this, as a game like baseball has very detailed player metrics thanks to the dedicated work of sports statisticians and fans. It’s more labor intensive, but without that effort, baseball wouldn’t be able to give is Win-Above-Replacement, a close-to-universal ranking of a player’s skill which can be used with other metrics to create interesting analytical contexts.

This is our third and final takeaway: You can use data to provide hyper-specific contexts which are novel or strange to your audience. In fact, it’s through these hyper-specific contexts that the most interesting outliers are found. Just remember the lowly Scout.

Tangent: It would be a mistake, in a talk all about data, to not mention GDPR. This regulatory bill describes what kind of information you can retain, and gives players a way to better control their data. Since all of the tension building detailed above relies on institutional data-gathering, it’s important to make sure that the regulations are properly followed.

To ensure that you do not run afoul of GDPR, make sure that you sanitize and anonymize user data before it is added to your data sets. Do not include personally-identifying information such email addresses, and refer to player data through a user id rather than a username. Require an opt-in to view and save replays, and make sure you have an avenue for players to request deletion of personal information. By doing this, you can keep important game metrics while properly maintaining user privacy.

Using Data Effectively

If you want to implement these takeaways in your game or league, there are a number of options.

Consider the ways you use infographics and HUDs to communicate information to the audience. Simple, uncluttered panels filled with otherwise hidden information allow players to focus on the action while still getting a peek behind the curtain.
Use global metrics to create player profiles with novel, human-understandable information. This varies widely by game - since what information is important varies significantly, but there are plenty of metrics to consider. For a fighting game example, consider a crossup success chance versus certain characters or players.
Create a replay system with an API that allows for automated metrics gathering by third-party sources. The more robust and open your data gathering, the more likely your game’s community will do the data visualization for you. Account for future balance or game changes when designing the replays too; you don’t know what your game will look like after years of live operations, so try to future-proof.

By giving players access to robust datasets, visualizing that data in novel ways, and using it to contextualize the competition on screen, you dramatically increase the audience’s investment in the match and the tension that gives rise to hype. And the more excited your players are, the more they will tune in for your broadcasts, spend money on your game, and collaborate to build a thriving competitive community.