Time to Die: Death Prediction in Dota 2 using Deep Learning

Esports have become major international sports with hundreds of millions of spectators. Esports games generate massive amounts of telemetry data. Using these to predict the outcome of esports matches has received considerable attention, but micro-predictions, which seek to predict events inside a match, is as yet unknown territory. Micro-predictions are however of perennial interest across esports commentators and audience, because they provide the ability to observe events that might otherwise be missed: esports games are highly complex with fast-moving action where the balance of a game can change in the span of seconds, and where events can happen in multiple areas of the playing field at the same time. Such events can happen rapidly, and it is easy for commentators and viewers alike to miss an event and only observe the following impact of events. In Dota 2, a player hero being killed by the opposing team is a key event of interest to commentators and audience. We present a deep learning network with shared weights which provides accurate death predictions within a five-second window. The network is trained on a vast selection of Dota 2 gameplay features and professional/semi-professional level match dataset. Even though death events are rare within a game (1\% of the data), the model achieves 0.377 precision with 0.725 recall on test data when prompted to predict which of \textit{any} of the 10 players of either team will die within 5 seconds. An example of the system applied to a Dota 2 match is presented here. This model enables real-time micro-predictions of kills in Dota 2, one of the most played esports titles in the world, giving commentators and viewers time to move their attention to these key events.

Modelling Early User-Game Interactions for Joint Estimation of Survival Time and Churn Probability

Data-driven approaches which aim to identify and predict player engagement are becoming increasingly popular in games industry contexts. This is due to the growing practice of tracking and storing large volumes of in-game telemetries coupled with a desire to tailor the gaming experience to the end-user’s needs. These approaches are particularly useful not just for companies adopting Game-as-a-Service (GaaS) models(e.g. for re-engagement strategies) but also for those working under persistent content-delivery regimes (e.g. for better audience targeting). A major challenge for the latter is to build engagement models of the user which are data-efficient, holistic and can generalize across multiple game titles and genres with minimal adjustments.This work leverages a theoretical framework rooted in engagement and behavioural science research for building a model able to estimate engagement-related behaviours employing only a minimal set of game-agnostic metrics. Through a series of experiments we show how, by modelling early user-game interactions,this approach can make joint estimates of long-term survival time and churn probability across several single-player games in a range of genres. The model proposed is very suitable for industry applications since it relies on a minimal set of metrics and observations, scales well with the number of users and is explicitly designed to work across a diverse range of titles.

Predicting skill learning outcomes in a large, longitudinal MOBA dataset

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. In this paper, 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.

Predicting Retention in Sandbox Games with Tensor Factorization-based Representation Learning

Major commercial (AAA) games increasingly transit to a semi-persistent or persistent format in order to extend the value of the game to the player, and to add new sources of revenue beyond basic retail sales. Given this shift in the design of AAA titles, game analytics needs to address new types of problems, notably the problem of forecasting future player behavior. This is because player retention is a key factor in driving revenue in semi-persistent titles, for example via downloadable content. This paper introduces a model for predicting retention of players in AAA games and provides a tensor-based spatio-temporal model for analyzing player trajectories in 3D games. We show how knowledge as to trajectories can help with predicting player retention. Furthermore, we describe two new algorithms for three way DEDICOM including a fast gradient method and a semi-nonnegative constrained method. These approaches are validated against a detailed behavioral data set from the AAA open-world game Just Cause 2.

Predicting Victory in a Hybrid Online Competitive Game: The Case of Destiny

Competitive multi-player game play is a common feature in major commercial titles, and has formed the foundation for esports. In this paper, the question whether it is possible to predict match outcomes in First Person Shooter-type multiplayer competitive games with mixed genres is addressed. The case employed is Destiny, which forms a hybrid title combining Massively Multi-player Online Role-Playing game features and First-Person Shooter games. Destiny provides the opportunity to investigate prediction of the match outcome, as well as the influence of performance metrics on the match results in a hybrid multi-player major commercial title. Two groups of models are presented for predicting match results: One group predicts match results for each individual game mode and the other group predicts match results in general, without considering specific game modes. Models achieve a performance between 63% and 99% in terms of average precision, with a higher performance recorded for the models trained on specific multi-player game modes, of which Destiny has several. We also analyzed performance metrics and their influence for each model. The results show that many key shooter performance metrics such as Kill/Death ratio are relevant across game modes, but also that some performance metrics are mainly important for specific competitive game modes. The results indicate that reliable match prediction is possible in FPS-type esports games.

Win Prediction in Multi-Player Esports: Live Professional Match Prediction

Esports are competitive videogames watched by audiences. Most esports generate detailed data for each match that are publicly available. Esports analytics research is focused on predicting match outcomes. Previous research has emphasised pre-match prediction and used data from amateur games, whichare more easily available than professional level. However, the commercial value of win prediction exists at the professional level. Furthermore, predicting real-time data is unexplored, as isits potential for informing audiences. Here we present the first comprehensive case study on live win prediction in a professional esport. We provide a literature review for win prediction in a multi-player online battle arena (MOBA) esport. The paper evaluates the first professional-level prediction models for liveDotA 2 matches, one of the most popular MOBA games and trials it at a major international esports tournament. Using standardmachine learning models, feature engineering and optimization,our model is 85% accurate after 5 minutes of gameplay. Our analyses highlight the need for algorithm evaluation and optimization and how professional data differs from amateur data.Finally, we present implications for the esports/game analytics domains, describe commercial opportunities, practical challenges,and propose a set of evaluation criteria for research on esports win prediction.

Predicting Player Churn in Destiny: A Hidden Markov Models Approach to Predicting Player Departure in a Major Online Game

Destiny is, to date, the most expensive digital game ever released with a total operating budget of over half a billion US dollars. It stands as one of the main examples of AAA titles, the term used for the largest and most heavily marketed game productions in the games industry. Destiny is a blend of a shooter game and massively multi-player online game, and has attracted dozens of millions of players. As a persistent game title, predicting retention and churn in Destiny is crucial to the running operations of the game, but prediction has not been attempted for this type of game in the past. In this paper, we present a discussion of the challenge of predicting churn in Destiny, evaluate the area under curve (ROC) of behavioral features, and use Hidden Markov Models to develop a churn prediction model for the game.

Rapid Prediction of Player Retention in Free-to-Play Mobile Games

Predicting and improving player retention is crucial to the success of mobile Free-to-Play games. This paper explores the problem of rapid retention prediction in this context. Heuristic modeling approaches are introduced as a way of building simple rules for predicting short-term retention. Compared to common classification algorithms, our heuristic-based approach achieves reasonable and comparable performance using information from the first session, day, and week of player activity

Predicting Player Behavior in Tomb Raider: Underworld 

imagesdddThis paper presents the results of an explorative study on predicting aspects of playing behavior for the major commercial title Tomb Raider: Underworld (TRU). Various supervised learning algorithms are trained on a large-scale set of in-game player behavior data, to predict when a player will stop playing the TRU game and, if the player completes the game, how long will it take to do so. Results reveal that linear regression models and other non-linear classification techniques perform well on the tasks and that decision tree learning induces small yet well-performing and informative trees. Moderate performance is achieved from the prediction models, which indicates the complexity of predicting player behavior based on a constrained set of gameplay metrics and the noise existent in the dataset examined, a generic problem in large-scale data collection from millions of remote clients.