World's first: Discovery of Neural Patterns Strongly Linked to Esports Match Outcomes
Predicting "similar-level match" and "upsets" with approximately 80% accuracy

News Highlights:

Tokyo - July 18, 2024 - NTT Corporation (Headquarters: Chiyoda Ward, Tokyo; Representative Member of the Board and President: Akira Shimada; hereinafter "NTT") has achieved a world first by identifying neural oscillation patterns that associated with esports match outcomes and has succeeded in predicting with high accuracy the outcome of the match using this EEG data.
　These results show an optimal brain state right before competitions and demonstrate that brain information is effective in predicting competition performance. This significant achievement is expected to improve performance through feedback on ideal brain states for individuals active in various fields, such as sports, medicine, and education. Additionally, it paves the way for skill transfer through digitization of brain states of experts performing advanced tasks.
　Details of neural oscillations related to esports win/loss were published in the American journal iScience on June 16, 2023, and details of win/loss prediction technology were published in the Dutch journal Computers in Human Behavior on July 18, 2024.

1. Background

NTT Communication Science Laboratories is dedicated to researching how the brain regulates the mind and body to enable individuals to maximize their abilities. Athletes strive to perform at their best under intense pressure during competitions by optimizing their mental state, a process commonly referred to as "mental conditioning". The brain state of athletes as they approach competition is believed to be closely linked to their mental state. Previous research has investigated the relationship between brain state and momentary performance across various sports. However, due to the challenges of measuring brain activity during actual competition, the brain states associated with outcomes in competitive sports have not been identified.
　In recent years, advancements in sports analytics have significantly improved the accuracy of predicting competition outcomes based on past performance data. Despite these advancements, predicting the outcomes of "similar-level matches" or "upsets"¹ remains challenging due to numerous uncertain factors.

2. Results of the research

In this study, we focused on a fighting video game in esports, which emphasize the mental game aspect over differences in physical motor skills and are well-suited for brain measurement. We observed the brain states of players during matches using electroencephalography (EEG)². This allowed us to investigate the brain activity patterns that give rise to the ideal mental state for athletes as they approach competition. We successfully identified specific brain activity patterns in players' pre-match EEG data that are strongly linked to winning or losing. Consequently, we achieved an approximately 80% accuracy in predicting match outcomes based on EEG data. This approach has the potential to predict match results in scenarios difficult for traditional prediction technologies, such as "similar-level matches" and "upsets."

2.1 Discovery of brain activity patterns strongly associated with match outcomes

To investigate brain activity patterns associated with match outcomes, we measured the neural oscillations (EEG) of skilled esports fighting video game players during matches conducted under real competition conditions using a two-round first-pass system. In esports, strategic decision³ and emotional control⁴ are considered crucial for success. According to survey results from players, the importance of strategic decision is high just before the first round, while the importance of emotional control is high just before the third round. Therefore, we focused on EEG patterns related to strategic decision and emotional control in the first and third pre-round periods and examined their association with match outcomes (Figure 1, top). We found that the left frontal gamma oscillations⁵ in the first pre-round period, associated with strategic decision-making, and the left frontal alpha oscillations⁶ in the third pre-round period, associated with emotional control, were increased in winning matches, respectively (Figure 1, bottom).

Figure 1 Pre-round EEG correlates with match outcomes in fighting video game.

2.2 High-accuracy match outcome prediction using models trained on EEG data

Next, to evaluate how accurately pre-match EEG patterns can predict match outcomes, we constructed multiple machine learning models using EEG data recorded just before matches conducted under a one-round first-pass system (Figure 2A). The results revealed that the best-performing model identified EEG features related to match outcomes and was able to predict match results with approximately 80% accuracy. This accuracy was consistent for predicting both wins and losses (Figure 2B). Furthermore, when predicting outcomes for similar-level matches and upsets, the best-performing model trained on EEG data maintained a high predictive accuracy of about 80% (Figure 2C). In contrast, the best-performing model trained on traditional data, such as players' past match information, showed low predictive accuracy (Figure 2D).

Figure 2 Results of match outcome predictions using pre-match EEG data.
(A) Schematic of machine learning models using either EEG data or traditional data.
(B) Prediction accuracy and the corresponding confusion matrix.
Comparison of prediction accuracy for all matches, similar-level matches, and upsets, using models trained with either (C) EEG data or (D) traditional data.

3. Outlook

This study highlights the existence of an ideal brain state in competitive situations, providing important insights into how individuals can cope with pressure in various scenarios. We aim to enhance people's well-being through bio-information-based mental conditioning across various fields such as sports, healthcare, and education. It will also contribute to the digitalization of the skills of experts in various fields and the transfer of advanced skills in the R&D of digital twin computing⁸ under the IOWN concept⁷.

[Paper Information]

¹Upset: A match when a lower-ranked player unexpectedly defeats a higher-ranked player. In this experiment, an upset is defined as a match where a player with a lower in-game ranking wins against a player with a higher ranking.

²Electroencephalography (EEG): A noninvasive method of recording electrical activity in the brain. Electrodes are attached to the scalp to detect weak electrical signals from the brain. This allows researchers to examine different brain states and responses.

³The ability to choose the best way to achieve a strategic decision objective. In esports, it refers to an attempt to plan a behavior pattern in advance, considering the player's individual characteristics and character compatibility.

⁴Emotion control: An individual's ability to consciously regulate his or her emotions and express them appropriately. In esports, it refers to a conscious attempt to control the emotional turmoil of a match, such as fear of defeat.

⁵Gamma oscillation: A type of electroencephalogram that refers to high frequencies above 30 Hz.

⁶Alpha oscillation: A type of neural oscillation that is as low as 7-13 Hz.

⁷IOWN Concept: A concept of a network and information processing infrastructure, including terminals, capable of providing high-speed, high-capacity communications and enormous computing resources utilizing innovative technologies centered on optics, to transcend the limits of conventional infrastructure, optimize individuals and the whole based on all kinds of information, and create an affluent society that embraces diversity. URL:https://www.rd.ntt/e/iown/0001.html

⁸Digital twin computing: To digitally construct twins of the real world by digitally expressing things and people. Combinations that could not be treated comprehensively until now can be reproduced with high precision, and make future predictions. URL:https://www.rd.ntt/e/iown/0003.html

About NTT

NTT contributes to a sustainable society through the power of innovation. We are a leading global technology company providing services to consumers and businesses as a mobile operator, infrastructure, networks, applications, and consulting provider. Our offerings include digital business consulting, managed application services, workplace and cloud solutions, data center and edge computing, all supported by our deep global industry expertise. We are over $97B in revenue and 330,000 employees, with $3.6B in annual R&D investments. Our operations span across 80+ countries and regions, allowing us to serve clients in over 190 of them. We serve over 75% of Fortune Global 100 companies, thousands of other enterprise and government clients and millions of consumers.