Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Shape patterns in popularity series of video games

Leonardo R. Cunha, Arthur A. B. Pessa, Renio S. Mendes

TL;DR

This study investigates large-scale popularity time series for 5,840 Steam games to identify shape-based patterns in their evolution. After normalizing and smoothing the monthly popularity traces, the authors apply Dynamic Time Warping, UMAP, and Infomap to uncover five distinct shapes: decreasing, hilly, increasing, valley, and bursty, with nearly half the games decreasing over time. The results reveal genre associations and show that most patterns persist as games mature, while some transitions occur, notably constant patterns evolving into other shapes. These findings suggest universal dynamics in cultural content popularity and offer practical insights for developers regarding lifecycle timing and marketing strategies. The methodology and dataset enable scalable, shape-focused analyses of cultural dynamics across digital platforms.

Abstract

In recent years, digital games have become increasingly present in people's lives both as a leisure activity or in gamified activities of everyday life. Despite this growing presence, large-scale, data-driven analyses of video games remain a small fraction of the related literature. In this sense, the present work constitutes an investigation of patterns in popularity series of video games based on monthly popularity series, spanning eleven years, for close to six thousand games listed on the online platform Steam. Utilizing these series, after a preprocessing stage, we perform a clustering task in order to group the series solely based on their shape. Our results indicate the existence of five clusters of shape patterns named decreasing, hilly, increasing, valley, and bursty, with approximately half of the games showing a decreasing popularity pattern, 20.7% being hilly, 11.8% increasing, 11.0% bursty, and 9.1% valley. Finally, we have probed the prevalence and persistence of shape patterns by comparing the shapes of longer popularity series during their early stages and after completion. We have found the majority of games tend to maintain their pattern over time, except for a constant pattern that appears early in popularity series only to later originate hilly and bursty popularity series.

Shape patterns in popularity series of video games

TL;DR

This study investigates large-scale popularity time series for 5,840 Steam games to identify shape-based patterns in their evolution. After normalizing and smoothing the monthly popularity traces, the authors apply Dynamic Time Warping, UMAP, and Infomap to uncover five distinct shapes: decreasing, hilly, increasing, valley, and bursty, with nearly half the games decreasing over time. The results reveal genre associations and show that most patterns persist as games mature, while some transitions occur, notably constant patterns evolving into other shapes. These findings suggest universal dynamics in cultural content popularity and offer practical insights for developers regarding lifecycle timing and marketing strategies. The methodology and dataset enable scalable, shape-focused analyses of cultural dynamics across digital platforms.

Abstract

In recent years, digital games have become increasingly present in people's lives both as a leisure activity or in gamified activities of everyday life. Despite this growing presence, large-scale, data-driven analyses of video games remain a small fraction of the related literature. In this sense, the present work constitutes an investigation of patterns in popularity series of video games based on monthly popularity series, spanning eleven years, for close to six thousand games listed on the online platform Steam. Utilizing these series, after a preprocessing stage, we perform a clustering task in order to group the series solely based on their shape. Our results indicate the existence of five clusters of shape patterns named decreasing, hilly, increasing, valley, and bursty, with approximately half of the games showing a decreasing popularity pattern, 20.7% being hilly, 11.8% increasing, 11.0% bursty, and 9.1% valley. Finally, we have probed the prevalence and persistence of shape patterns by comparing the shapes of longer popularity series during their early stages and after completion. We have found the majority of games tend to maintain their pattern over time, except for a constant pattern that appears early in popularity series only to later originate hilly and bursty popularity series.
Paper Structure (5 sections, 19 figures, 1 table)

This paper contains 5 sections, 19 figures, 1 table.

Table of Contents

  1. Introduction
  2. Dataset
  3. Methods
  4. Results
  5. Discussion and Conclusions

Figures (19)

  • Figure 1: Temporal evolution of the number of games and total popularity. (A) The cumulative number of games monthly released on Steam is represented by a solid line and the dashed line stands for a power-law fit with a power-law exponent equal to $1{.}81$. (B) The sum of popularities of games for each month is represented by a solid line and the dashed line stands for a linear fit. In both panels, the data covers a period of 120 months from August 2012 to July 2022.
  • Figure 2: Number of games and players related to each game genre. (A) Cumulative number of games released on Steam belonging to each genre until July 2022. (B) Number of players per genre in July 2022. For each genre, the popularity of games in the month of July 2022 was summed.
  • Figure 3: Summary of time series of game popularity. (A) Probability distribution of the number of games as a function of the average of the popularity series. The gray circles represent the data and the dashed line stands for a log-normal distribution of parameters $\mu = 3{.}35$ and $\sigma = 2{.}30$. (B) Relationship between the averages and standard deviations of popularity series. The light gray circles are the data, the black circles are averages taken in equally spaced logarithmic windows, and the black dashed line represents a linear model fitted to the data ($R^2 =$ 0.92).
  • Figure 4: Shape patterns of game popularity series on Steam. The center panel shows the proportion of games belonging to each group that was detected by the clustering algorithm: 47.4% of games are in cluster 1 (blue); 20.7% in cluster 2 (red); 11.8% in cluster 3 (green); 11.0% in cluster 4 (purple); 9.1% in cluster 5 (orange). Panels around the center plot depict the average shape of times series belonging to each cluster and one standard deviation band.
  • Figure 5: Representation of genres and their relative proportion in the detected clusters. (A) Percentage of games in our database belonging to one or more of the 12 official genres on Steam. Relative differences in representation of game genres for games pertaining to shape clusters 1 to 5 are shown in panels (B)-(F).
  • ...and 14 more figures