Artificial intelligence (AI) is used to play critical roles in various industries, ranging from healthcare to automation. AI technologies supplement human efforts, and the positive impact observed in these sectors demonstrates that AI technologies have also helped to improve the efficiency of most human processes.
eSports’ rise has been remarkable, but recently, AI technologies have been introduced to eSports to provide a brilliant twist to gaming. As a result, eSports platforms like www.esports-betting.pro have seen significant traffic increases. eSports are no longer considered a subset of the sports industry. It is, however, now a thriving and self-contained industry.
The current trends indicate that eSports are here to stay, but the future holds much promise. So what does the future hold for eSports, according to the crystal ball? Many industry experts have attempted to unravel the future of eSports in light of its impressive numbers.
Artificial Intelligence in Service of Changing the Known World
In a nutshell, AI (artificial intelligence) refers to systems or machines that mimic human intelligence in order to perform tasks and can iteratively improve themselves based on the data they collect. AI manifests itself in a variety of ways. Here are a few examples:
- Chatbots use artificial intelligence to better understand customer problems and provide more efficient responses.
- AI is used by intelligent assistants to parse critical information from large free-text datasets to improve scheduling.
- Based on users’ viewing habits, recommendation engines can provide automated recommendations for TV shows.
AI is much more about the process and the ability to think faster and analyze data than it is about any specific format or function. Although images of high-functioning, human-like robots taking over the world conjure up images of AI, the technology is not intended to replace humans.
Its goal is to vastly improve human capabilities and contributions. As a result, it is a precious business asset reshaping industries as we know them and the world around us.
Getting the Most Out of Artificial Intelligence
The true power of AI was not in predicting the outcome of the status quo but in determining how best to change it!
For example, the model’s single most important feature was its price. Lowering a product’s cost increased the models’ forecasted sales while decreasing profit-per-unit. The model was worth building because of the art of balancing these two aspects. The price-setting team could consult the models to understand how various scenarios would play out or simply ask for the best price.
The same is true in eSports. If a model’s sole function was to tell you that “Team A had a 75% chance of winning the game,” this information would be meaningless unless you’re a professional gambler.
There must be a “so what?”
A useful realization. Hopefully, the examples I’ve provided below will help to clarify what we mean and provide food for thought.
A Quick Overview of Predictive Models
AI is divided into two categories: supervised and unsupervised.
Prediction falls under the former category, in which we provide the model with features (such as price in my previous example) and a value to predict (such as sales).
The model must then learn how to make the best use of the given features to make this prediction. It is best practice to then ask the model to forecast new values using data that was not used to train it to fairly evaluate how well it has made these predictions.
From simple linear models like Linear Regression to complex “black-box” approaches like Deep Learning, the “how” varies. There are many well-written breakdowns of predictive models available, and I would recommend further reading if you are unfamiliar with this topic.
The Use of Predictive Models in eSports
One of the most useful applications of AI in eSports is the prediction of game outcomes. This could be done days before the game, with the primary features being team-based, such as recent performance or player information. A step up from this is a model that predicts the outcomes just before the game begins. This would include all of the previous features, as well as new eSports-specific features such as the map to be played, which side the teams are starting on, and which characters/loadouts they’ve chosen.
Even more complex models for in-game predictions can be built, incorporating all previous information as well as new live information, such as the resources held by the teams at any point in the game.
However, keep in mind that, aside from betting, predictions provide no value to players or coaches. What matters is understanding why and applying this knowledge to make better decisions. For example, suppose we are interested in predicting the game’s outcomes immediately after the players have decided which characters they will play.
If you’re more familiar with first-person shooters, replace “loadouts” with “characters.” To accomplish this, the model has two features:
- The characters’ most recent overall performance
- The players’ most recent performance as their chosen character
There will most likely be two relationships. To begin with, a character who has won 60% of their previous games should be more likely to win than one who has only won 40% of their previous games. Isn’t it self-evident?
Furthermore, a player who has won 60% of their games on their character should be more likely to win than one who has only won 40% of their games. Although it may be reassuring to know that the data supports our assumptions, we will not be adding any real value.
The power in my previously mentioned financial model came from being able to optimally trade off margin and volume. Here, we must strike a balance between the character’s strength and the players’ ability to control it.
For example, is it better to have a character with a 60% recent win rate but the player has only won 40% of their games on it than a character with a 40% win rate but the player has won 60% of their games on it?
This is about as basic as it gets in terms of examples. Two variables, both of which have obvious and straightforward relationships.
Nonetheless, it is a difficult question to answer. It gets more difficult with each new feature we consider. However, for AI, this is a piece of cake. It can analyze thousands of games and discover subtle rules that favor one trade-off over another. Depending on the model used, it may even learn less obvious relationships.
For example, in most games, having a healer/medic on the team is advantageous, so increasing this variable is advantageous.
However, do you really want to field a team comprised entirely of support roles?
How much healing is too much, and does it differ depending on the map, game type, or opponent strategies?
These and other questions confront coaches on a daily basis, and making accurate estimates is also impossible for the human brain.
So we take all of the variables we can think of and train a model to predict the game’s outcomes. That is one aspect of the job of a Data Scientist. However, and perhaps more importantly, we must use these models to provide actionable insights to the team. Here are a couple of examples:
“The model predicts that we will lose a lot of games this season because our mid-range damage statistic is low; therefore, during the off-season, we should consider acquiring a player who excels in this role.” Here are some of the league’s best players who will help us win.”
“The issue with selecting this character is that, while it is performing well and our players are at ease with it, the model believes it will struggle against the opposition’s playstyle.” This alternate character has a lower win rate but has historically been a strong counter to aggressive teams.”
“We had a 60% chance of winning until this point in the game, and then it drops to 40%.” This was the point at which we concentrated on securing Objective A4 while allowing the opposition to secure Objective B2. According to the model, we could have increased our win chance to 65% if we had instead focused on defending B2 from their attack.”
The Data Availability
A model is only as good as the data it is fed. Traditional sports have been dealing with this issue for many years. Even though they wanted to deploy AI as soon as possible, they quickly realized that simple statistics like “goals scored” or “home runs” only produced simple models.
What was needed was not an AI advancement but rather data capture.
The ability to convert a physical world event, such as a player sprinting down the left field, into detailed data such as pace, stride, and exact position. This is not a cheap endeavor.
To begin, the footage must be captured from multiple angles using high-resolution video cameras. This data must then be engineered into a usable format before being run through Computer Vision models capable of converting millions of images into raw data.
This stage can be avoided, which is why eSports is in such a good position. Games are typically designed in such a way that raw information, such as a player’s position on the map or the abilities they use, is captured by default. Those who want to work on AI can start playing with data immediately, without the costly data collection investment required in traditional sports.
*Keep in mind that not all game publishers are the same. Some provide data via an API to anyone who requests it, while others charge a fee or do not release it at all.
AI’s Unavoidable Flaw in eSports
We hope that we have given you some reason to be optimistic about the future of a data scientist in eSports. However, before we proceed, we must point out one flaw. This comes up almost immediately when researching the field and will continue to overshadow your work.
Simply put, games evolve. Consider traditional board games such as Chess or Go, where the rules have remained unchanged for centuries. If you played against your friend last year, it’s safe to assume that the only difference between the two of you will be the next time you play.
This is why AI has been so successful in the competitive field. We were able to learn from millions upon millions of games that all followed the same rules. It will eventually learn even the most sophisticated stratagems required to defeat its opponent.
Consider how each piece in Chess would be subject to a new set of random changes each game. Pawns can suddenly take a piece directly behind it. Every third move, the Queen can only be moved. Knights take two steps forward, followed by two steps to the side.
The AI would falter with each change, undoing millions of finely tuned parameters that suddenly had no idea how to play, let alone win, the game. This is the situation that eSports is in, with the most popular games changing on a weekly basis.
Assume we trained our model to identify the strongest character (or loadout) to use at the start of the game. Yes, that character had won 60% of their games in the previous two weeks, but the game updated yesterday, and she now does half the damage she used to.
So, will she still win 60%?
Is it 30% or 30%?
What if you’re the first game after the change, and you have no data to work with? Even so, eSports is popular but not that popular. For the most watched game, there will only be 50 professionally played games per week, with no guarantee that they’ll even play the character — far from enough data to learn the new impact of the changes.
The eSports Secret Weapon
But don’t give up hope just yet! Of course, this is a challenge, but it is not insurmountable. Thus, the final advantage of eSports is that millions of people play the game every day!
Moreover, those professional eSports players who compete on stage each week also return home to play with the general public. Imagine if not only were millions of football games played across the country every day, but all of their data was reliably tracked and stored.
Even better, most games use a match-making system, which allows the most skilled players to play together while leaving the casual newcomers alone (for the most part). This allows data to be collected and hand-picked to include only those whose games most closely resemble the professional environment.
Of course, these “casual” games will never truly replicate the professional scene, owing to a) players’ lower risk tolerance on stage and b) teams’ improved coordination, particularly if they have a coach. However, this does not preclude you from devising clever ways to extract insights from this invaluable data source.
