第一财经

AI is infiltrating esports: It doesn't play games for the players, but it can change the outcome of matches.

原文:AI渗透电竞:不替选手打比赛,但能改变胜负

Summary of Key Points

Artificial Intelligence (AI) is being utilized in esports to help teams identify hidden issues, quantify training metrics, and improve performance by analyzing details such as voice communication and hero selection strategies. For example, an AI tool developed in collaboration between Team Liquid and SAP can detect subtle problems like emotional distractions and language switches during conversations, as well as assist with battle plans (BP) formulation and post-match analysis. Esports, being a purely digital medium, have become a prime environment for the implementation of AI technologies. However, AI is not a shortcut to victory; it merely serves as a tool to reduce mistakes.

1. AI Identifies “Hidden Bugs” in Team Communication

In competitive team games, effective communication is crucial for winning battles, but many issues go unnoticed during manual post-match reviews by coaches. Take Team Liquid as an example: The team has three Korean players who tend to switch to Korean when the game becomes intense, making it difficult for their Argentine and American teammates to understand them and miss important instructions. AI analyzed the team's voice data and revealed that Korean accounted for a significant proportion of their communication. As a result, the team conducted training on using specific English slang in games, which significantly reduced the use of Korean and improved communication. AI can also pinpoint moments when players become emotionally overwhelmed and utter meaningless statements, such as former DOTA2 player Zai mentioning how stress can lead to providing irrelevant information that disrupts teammates.

2. Turning Vague Sensations into Trainable Metrics

Coaches used to rely on vague comments like “There was a communication issue just now,” but AI can transform these feelings into concrete data. For instance, it can analyze voice recordings to determine whether players' instructions were clear during intense battles or if emotional fluctuations affected their decision-making. By analyzing keywords and tone changes in the audio, AI can inform coaches that a particular defeat was caused by unclear instructions from a specific player, leading to incorrect positioning of teammates. This data can then be used as training targets—such as reducing the frequency of ineffective communication during stressful situations or standardizing the use of English terminology.

3. AI Goes Beyond Communication

AI has other applications in esports beyond just analyzing voice data. For example, the AI Draft Bot can assist teams in evaluating their battle plans by predicting which heroes opponents might choose and recommending the best combinations. Joule Agents can quickly organize players' performance data, highlighting issues such as economic disadvantages or poor positioning. Additionally, AI-assisted post-match analysis tools integrate billions of data points from 190,000 matches into visual dashboards, providing coaches with immediate insights into areas for improvement.

4. Esports: A Natural Testing Ground for AI

Esports are an ideal environment for AI because everything about the game is digitally recorded—every move, word, and skill use is quantifiable, unlike in traditional sports like football where some actions are harder to measure. Thomas describes this as a “blue ocean” situation, where the vast amount of data allows AI to develop more accurate models. Analyzing 10 terabytes of raw data to identify winning patterns is challenging in other industries.

5. AI: Not a Victory Shortcut, but a Supportive Tool

While AI can enhance team performance, it cannot perform actions on behalf of players. Thomas emphasizes that Team Liquid invested 1%-2% more in technology than their opponents, but in top-tier competitions, even small advantages make a difference. Victory still depends on the players' skills, on-the-spot decision-making, and even luck. The role of AI is to reduce mistakes by turning incidental issues (like language switches) into sustainable training topics, helping teams avoid basic errors rather than achieving effortless victories.

In summary, AI is making esports more “scientific,” but it can never replace the hard work and talent of the players. Just as professional athletes use technology to train better, champions are still determined by their own efforts and abilities.