Competitive video game playing, an activity called esports, is increasingly popular to the point that there are now many professional competitions held for a variety of games. These competitions are broadcast in a professional manner similar to traditional sports broadcasting. Esports games are generally fast paced, and due to the virtual nature of these games, camera positioning can be limited. Therefore, knowing ahead of time where to position cameras, and what to focus a broadcast and associated commentary on, is a key challenge in esports reporting. This gives rise to moment -to -moment prediction within esports matches which can empower broadcasters to better observe and process esports matches. In this work we focus on this moment -to -moment prediction and in particular present techniques for predicting if a player will die within a set number of seconds for the esports title Dota 2 . A player death is one of the most consequential events in Dota 2 . We train our model on 'telemetry' data gathered directly from the game itself, and position this work as a novel extension of our previous work on the challenge. We use an enhanced dataset covering 9,822 Dota 2 matches. Since the publication of our previous work, new dataset parsing techniques developed by the WEAVR project enable the model to track more features, namely player status effects, and more importantly, to operate in real time. Additionally, we explore two new enhancements to the original model: one data -based extension and one architectural. Firstly we employ learnt embeddings for categorical features, e.g. which in game character a player has selected, and secondly we explicitly model the temporal element of our telemetry data using recurrent neural networks. We find that these extensions and additional features all aid the predictive power of the model achieving an F1 score of 0.54 compared to 0.17 for our previous model (on the new data). We improve this further by experimenting with the length of the time -series in the input data and find using 15 time steps further improves the F1 score to 0.62. This compares to F1 of 0.1 for a standard RNN on the same task. Additionally a deeper analysis of the Time to Die model is carried out to assess its suitability as a broadcast aid.

Esport is a large and still growing industry with vast audiences. Multiplayer Online Battle Arenas (MOBAs), a sub-genre of esports, possess a very complex environment, which often leads to experts missing important coverage while broadcasting live competitions. One common game event that holds significant importance for broadcasting is referred to as a team fight engagement. Professional player's own knowledge and understanding of the game may provide a solution to this problem. This paper suggests a model that predicts and detects ongoing team fights in a live scenario. This approach outlines a novel technique of deriving representations of a complex game environment by relying on player knowledge. This is done by analysing the positions of the in-game characters and their associated cameras, utilising this data to train a neural network. The proposed model is able to both assist in the production of live esport coverage as well as provide a live, expert-derived, analysis of the game without the need of relying on outside sources.

In team sports, it can be a non-trivial task to compare players and understand how much they contribute and in what way, especially when those players are fulfilling significantly different roles on a team. What metric do you use to compare a Pitcher to a Batter? A Point Guard to a Center? Understanding what degree players are performing and how they compare to one another allows for greater depth in storytelling for the broadcasting team. It also provides audiences with overall performances across team positions.

Work in AI-based explanation systems has uncovered an interesting contradiction: people prefer and learn best from why explanations but expert esports commentators primarily answer what questions when explaining complex behavior in real-time strategy games. Three possible explanations for this contradiction are: 1.) broadcast audiences are well-informed and do not need why explanations; 2.) consuming why explanations in real-time is too cognitively demanding for audiences; or 3.) producing live why explanations is too difficult for commentators. We answer this open question by investigating the effects of explanation types and presentation modalities on audience recall and cognitive load in the context of an esports broadcast. We recruit 111 Dota 2 players and split them into three groups: the first group views a Dota 2 broadcast, the second group has the addition of an interactive map that provides what explanations, and the final group receives the interactive map with detailed why explanations. We find that participants who receive short interactive text prompts that provide what explanations outperform the no explanation group on a multiple-choice recall task. We also find that participants who receive detailed why explanations submit reports of cognitive load that are higher than the no explanation group. Our evidence supports the conclusion that informed audiences benefit from explanations but do not have the cognitive resources to process why answers in real-time. It also supports the conclusion that stacked explanation interventions across different modalities, like audio, interactivity, and text, can aid real-time comprehension when attention resources are limited. Together, our results indicate that interactive multimedia interfaces can be leveraged to quickly guide attention and provide low-cost explanations to improve intelligibility when time is too scarce for cognitively demanding why explanations.

Teamwork is challenging in cooperative digital games, especially between strangers. In many online cooperative games, teams have a short-lived existence and ever-changing membership. Our study explores how short-lived ad hoc teams of strangers communicate and investigate its effect on team performance. We analysed 2256 text message instances produced by teams during a 45-minute interaction. Our findings show that team communication is negatively related to performance, and affects performance over and beyond prior experience. A content analysis shows that teams generally have higher task-related communication than socio-emotional communication. This pattern is consistent throughout the duration of the interaction period. The results are discussed in the context of previous research on team communication and performance, and we draw parallels with communication patterns in real-world groups such as aviation crews.

The metagame is an overloaded term with no unified definition despite its importance and its common occurrence across different fields such as game design and behavioural economics. In our research we provide a unified and compact definition of the term metagame and metagaming by firstly highlighting their historical evolution. Although the original definition of metagame meant multiple things such as the environment surrounding the game, it has come to mean a perceived optimal or dominant playing strategy that is usually popular within an esport at that specific point in time. Metagaming as a verb is distinct and refers to a number of ways, external to the game's environment, a player can affect the outcome of a game. We focus on how these terms crystallised in the world of digital entertainment (esports) by providing multiple examples of metagames and metagaming in competitive settings. We additionally highlight the benefits of metagame shifts from the point of view of game developers. Finally, we provide a theoretical framework on the life cycles of metagames, as well general guidelines for understanding the current metagame of League of Legends and Dota 2. We conclude that by understanding the highly fluctuating metagame(s) of an esport at specific points in time, researchers will gain a better historical context of that game’s space. This in turn will give them insight into the decision making of professional esports players, game developers and tournament organisers. Additionally, this work will help researchers create better analytics tools and machine learning algorithms.

Unlike traditional physical sports, Esport games are played using wholly digital platforms. As a consequence, there exists rich data (in-game, audio and video) about the events that take place in matches. These data offer viable linguistic resources for generating comprehensible text descriptions of matches, which could, be used as the basis of novel text-based spectator experiences. We present a study that investigates if users perceive text generated by the NLG system as an accurate recap of highlight moments. We also explore how the text generated supported viewer understanding of highlight moments in two scenarios: i) text as an alternative way to spectate a match, instead of viewing the main broadcast; and ii) text as an additional information resource to be consumed while viewing the main broadcast. Our study provided insights on the implications of the presentation strategies for use of text in recapping highlight moments to Dota 2 spectators.

Multiplayer strategy games are examples of imperfect information games, where information about the game state can be retrieved through in-game mechanics. One such mechanic is vision. Within esports titles of this genre, such as League of Legends (LoL) and Dota 2, players often gather map information through the use of friendly units called wards. In LoL, one of the most popular esports title worldwide, warding has hitherto been evaluated only using a heuristic called vision score, provided by Riot, the game’s developer. In this paper, we examine the accuracy at LoL’s vision score at predicting the overall game-winner within the context supported by the game. We have ported LoL’s vision score to Dota 2, a similarly popular esports title, and compared its performance against a novel warding model. We have compared both models not only at predicting the overall winner, but also the current state of the game and their ability to predict and reflect short term game advantage and events. We found our model significantly outperformed LoL’s vision score. Additionally, we trained and evaluated a model for predicting the value of wards in real-time through the use of a Neural Network.

Esports (competitive videogames) have grown into a global phenomenon with over 450m viewers and a 1.5bn USD market. Esports broadcasts follow a similar structure to traditional sports. However, due to their virtual nature, a large and detailed amount data is available about in-game actions not currently accessible in traditional sport. This provides an opportunity to incorporate novel insights about complex aspects of gameplay into the audience experience – enabling more in-depth coverage for experienced viewers, and increased accessibility for newcomers. Previous research has only explored a limited range of ways data could be incorporated into esports viewing (e.g. data visualizations post-match) and only a few studies have investigated how the presentation of statistics impacts spectators’ experiences and viewing behaviors. We present Weavr, a companion app that allows audiences to consume data-driven insights during and around esports broadcasts. We report on deployments at two major tournaments, that provide ecologically valid findings about how the app’s features were experienced by audiences and their impact on viewing behavior. We discuss implications for the design of second-screen apps for live esports events, and for traditional sports as similar data becomes available for them via improved tracking technologies.

The exploration of the relationships between behavior and cognitive psychology of game players has gained impetus in recent years because such links provides an opportunity for improving user experiences and optimizing products in the games industry. At the same time, the volume and global scope of digital game telemetry data has opened up new experimental opportunities for studying human behavior at large scales. Prior research has demonstrated that a relation exists between learning rates and performance. Although many factors might contribute to this correlation at least one may be the presence of innate cognitive resources, as demonstrated in recent work relating IQ and performance in a Multi-player Online Battle Arena game. Here, we extend this work by examining the relationship between early learning rate and long term performance using a 400,000 player longitudinal dataset generated by new players of the widely-played MOBA League of Legends. We observed that the learning rate of new players in a competitive season explains a significant amount of variance in the performance at the end of the year. This analysis was then extended by training two multivariate classifiers (Logistic Regression, Random Forest) for predicting players who by the end of the season would be considered masters (top 0.05%), based on their performance in the first 10 matches of the same season. Both classifiers performed similarly (ROC AUC 0.888 for Logistic Regression, 0.878 for Random Forest), extending the time frame for skill prediction in games based on a relatively sparse sample of early data. We discuss the implications for these findings based on preexisting psychological studies of learning and intelligence, and close with challenges and direction for future research.

Hundreds of millions of people play intellectually-demanding video games every day. What does individual performance on these games tell us about cognition? Here, we describe two studies that examine the potential link between intelligence and performance in one of the most popular video games genres in the world (Multiplayer Online Battle Arenas: MOBAs). In the first study, we show that performance in the popular MOBA League of Legends' correlates with fluid intelligence as measured under controlled laboratory conditions. In the second study, we also show that the age profile of performance in the two most widely-played MOBAs (League of Legends and DOTA II) matches that of raw fluid intelligence. We discuss and extend previous videogame literature on intelligence and videogames and suggest that commercial video games can be useful as 'proxy' tests of cognitive performance at a global population level.

It has been suggested that the brain pre-empts changes in the visual environment through generating predictions, although real-time eletrophysiological evidence of prediction violations remains elusive. In a series of experiments we showed participants sequences of images that followed a predictable implied sequence or whose final image violated the implied sequence. Through careful design we were able to use the same final image transitions across predictable and unpredictable conditions, ensuring that any differences in neural responses were due only to preceding context and not to the images themselves. EEG and MEG recordings showed that early/mid-latency visual evoked potentials were robustly modulated by images that violated the implied sequence across a range of types of image change (expression deformations, rigid-rotations and visual field location). This modulation occurred irrespective of stimulus object category. Although the stimuli were static images, MEG source reconstruction of the early latency signal (N/M170) localised expectancy violation signals to brain areas associated with motion perception. Our findings suggest that the N/M170 can index mismatches between predicted and actual visual inputs in a system that predicts trajectories based on ongoing context. This has important implications for understanding the N/M170 and investigating how the brain represents context to generate perceptual predictions.

Multi-player online battle arena games (MOBAs) are large virtual environments requiring complex problem-solving and social interaction. We asked whether these games generate psychologically interesting data about the players themselves. Specifically, we asked whether user names, which are chosen by players outside of the game itself, predicted in-game behaviour. To examine this, we analysed a large anonymized dataset from a popular MOBA ('League of Legends') – by some measures the most popular game in the world. We find that user names contain two pieces of information that correlate with in-game social behaviour. Both player age (estimated from numerical sequences within name) and the presence of highly anti-social words are correlated with the valences of player/player interactions within the game. Our findings suggest that players' real-world characteristics influence behaviour and interpersonal interactions within online games. Anonymized statistics derived from such games may therefore be a valuable tool for studying psychological traits across global populations.

An unresolved goal in face perception is to identify brain areas involved in face processing and simultaneously understand the timing of their involvement. Currently, high spatial resolution imaging techniques identify the fusiform gyrus as subserving processing of invariant face features relating to identity. High temporal resolution imaging techniques localize an early latency evoked component—the N/M170—as having a major generator in the fusiform region; however, this evoked component is not believed to be associated with the processing of identity. To resolve this, we used novel magnetoencephalographic beamformer analyses to localize cortical regions in humans spatially with trial-by-trial activity that differentiated faces and objects and to interrogate their functional sensitivity by analyzing the effects of stimulus repetition. This demonstrated a temporal sequence of processing that provides category-level and then item-level invariance. The right fusiform gyrus showed adaptation to faces (not objects) at ∼150 ms after stimulus onset regardless of face identity; however, at the later latency of ∼200–300 ms, this area showed greater adaptation to repeated identity faces than to novel identities. This is consistent with an involvement of the fusiform region in both early and midlatency face-processing operations, with only the latter showing sensitivity to invariant face features relating to identity.SIGNIFICANCE STATEMENTNeuroimaging techniques with high spatial-resolution have identified brain structures that are reliably activated when viewing faces and techniques with high temporal resolution have identified the time-varying temporal signature of the brain's response to faces. However, until now, colocalizing face-specific mechanisms in both time and space has proven notoriously difficult. Here, we used novel magnetoencephalographic analysis techniques to spatially localize cortical regions with trial-by-trial temporal activity that differentiates between faces and objects and to interrogate their functional sensitivity by analyzing effects of stimulus repetition on the time-locked signal. These analyses confirm a role for the right fusiform region in early to midlatency responses consistent with face identity processing and convincingly deliver upon magnetoencephalography's promise to resolve brain signals in time and space simultaneously.