Factors affecting audience demand for professional football game videos: an analysis of post-game highlight videos on streaming platforms

The accessibility of online media across various devices has transformed media consumption, making it one of the leading internet activities (Dwyer, 2010). Sports fans now have more access than ever before, with streaming becoming a preferred viewing method, especially among younger audiences (Deloitte, 2021). Streaming technology enables the transmission of audio and video over the internet via platforms like Netflix, Hulu, Spotify, and YouTube, allowing users to watch content in real-time or on-demand. Social networks like Instagram and Facebook also offer streaming services, including video posts and live streams.

Illustrating this evolving dynamic in sports content consumption are notable examples like Amazon’s substantial investment of US$1 billion annually to stream the NFL’s Thursday Night Football and Apple TV’s contracts to stream MLB’s Friday Night Baseball and Major League Soccer (Deloitte, 2021). Furthermore, leagues and regional sports networks (RSNs) are exploring the launch of their direct-to-consumer streaming offerings for sports content (Deloitte, 2021). Their objectives revolve around gaining better control over their destiny and engaging fans in a more personalized manner.

Traditionally, sports media focused on showcasing only the most crucial live games, often leaving many other games in the background. This approach typically involved organizing and broadcasting live coverage of select high-profile events, with only a handful of important games receiving additional attention through highlights—condensed summaries of sports games showing key moments such as crucial goals, pivotal plays, and other notable events.

One distinguishing feature of streaming platforms compared to traditional broadcasts is that there is no limit on shelf space. This means that as much content as needed can be uploaded to the online streaming platform and be ready for fans to consume. This has led to a significant increase in non-live sports content, such as highlights, documentaries, interviews, and locker room scenes (Nielsen, 2022), which would not have been available to viewers through traditional broadcasts constrained by time-bound TV schedules.

This change has profound implications for the sports media industry. Because highlight videos can be produced for every single game on streaming platforms, sports enthusiasts have the freedom to click and enjoy any game of their choice as many times as they desire, catering to the varied interests of a dynamic and diverse fan base. The demand for highlight videos has risen as consumers seek to watch numerous video clips within a limited timeframe. This trend is particularly notable in professional sports, where fans are restricted by time and geographic locations that prevent them from watching every game in real time (McCammon, 2021). It is therefore common for fans to watch highlights of every game from the sports leagues or teams they follow.

The economic importance of sports highlight videos lies in their ability to drive fan engagement (Vale and Fernandes, 2018), foster loyalty (Fang et al., 2018), and create lucrative advertising opportunities for sponsors and teams (Cornwell, 2020). As digital content consumption grows, these highlights attract large, diverse audiences, making them a valuable platform for advertisers (Nielsen, 2023). By incorporating sponsored content or advertisements, highlight videos effectively connect brands with passionate sports fans, providing high visibility and targeted exposure (Da Silva and Las Casas, 2017). Consumed on platforms like YouTube and Instagram, these highlight videos effectively target specific audiences, making them highly profitable for advertisers (Wu, 2024). By attracting high viewership, highlight videos offer sponsors significant exposure and visibility for their brands and products (Cornwell, 2020).

Given the rise in streaming highlight videos, understanding the factors driving highlight demand on streaming platforms is as crucial as understanding the determinants of live sports demand. Only recently has a small amount of research dealt with the factors influencing streaming highlights. Butler and Butler (2023) examined fan preferences for streaming YouTube highlights of EPL games from 2019 to 2021, during which the COVID-19 lockdown occurred; they found that games played in empty stadiums were associated with greater highlight viewership. Han, Kim, and Kim (2021) examined the determinants of highlight videos for Korea’s professional football league, K-League 1, during the 2018 and 2019 seasons using data from Naver, a popular domestic streaming platform in Korea. Ryu, Hwang, Jeong, Jang, Lee, and Pyun (2023) also examined the demand for streaming highlights of the League of Legends Champions Korea games, one of the most followed esports leagues globally, and reported various ex-ante and ex-post variables predicting highlight viewership. Unfortunately, the results of these studies cannot be generalized as they were conducted within rather special contexts: Butler and Butler’s (2023) study examined the impact of the COVID-19 pandemic, Han et al.’s (2021) study was confined to a domestic streaming platform, while Ryu et al.’s (2023) study focused on esports leagues.

Existing literature on live sports demand (Gasparetto and Barajas, 2018; Li et al., 2023; Salaga et al., 2022; Sung et al., 2019) provides some insights, but these factors may not fully explain the demand for streaming highlights due to the unique characteristics of highlight videos and streaming platforms. Most notably, highlights are made available to fans after the game is over, so various ex-post factors that reflect the ever-changing dynamics and the unscripted drama generated during the games play a significant role. In contrast, for the telecasts of live games, ex-ante factors become important in influencing fans’ viewing behavior because fans’ decisions to tune into the broadcast need to be made before the game begins. Additionally, unlike televised live games, streaming highlights can be viewed repeatedly at fans’ convenience. As a result, variations in viewership of streaming highlight videos across games may be larger than those of live sports TV ratings.

This paper examines the factors influencing the demand for streaming highlight videos of Chinese Super League (CSL) football games. Data was obtained from Tencent Sports and CCTV Streaming, the two major streaming platforms producing highlight videos for CSL games, and from CSL’s official Weibo account, the most influential social network service (SNS) in China that posts CSL highlight videos. Three negative binomial regression models were used to regress a series of independent variables on the highlight view counts for each of the three streaming platforms, respectively. The findings offer brands and rights holders useful insights into how to effectively engage fans with game highlights on streaming platforms.

Demand for live game vs post-game highlights

The enjoyment of the media entertainment experience relies on the audience’s interest in the content presented by the media (Ross and Nightingale, 2003). Therefore, the demand for media content is proportional to the interest of potential viewers in the content and the enjoyment they derive from it (Vorderer et al., 2004). Viewers select a program based on their interest and anticipation of its entertainment value, and they are likely to continue watching if the actual viewing experience is enjoyable and entertaining. Therefore, fans are willing to sit in front of the screen on time to consume the live sports content they find engaging and enjoyable, and the viewing experience must deliver the expected pleasure to maintain fans’ interest and viewership (Da Silva and Las Casas, 2017).

Given the inherent unpredictability of sports outcomes, the demand for live sports content is primarily shaped by factors known beforehand because the decision to tune into a game on TV is made in the lead-up to the game. These ex-ante factors have been well identified through existing studies that investigate the demand for stadium viewing and live TV broadcasts (Borland and MacDonald, 2003; Villar and Guerrero, 2009).

Outcome uncertainty has been one of the central pre-game expectation factors investigated in televised football demand studies. However, the empirical evidence does not provide consistent results, with some studies supporting the outcome uncertainty hypothesis (Schreyer et al., 2018; Sung and Mills, 2018) and others reporting mixed findings (Buraimo and Simmons, 2015; Schreyer et al., 2017) or finding no support for it (Jang and Lee, 2022; Li et al., 2022; Sung et al., 2019). Additionally, factors associated with team loyalty (Yoshida et al., 2021; Besters et al., 2019), such as the age of the club (Borland and MacDonald, 2003), and team attributes representing on-field performance, such as team salary (Bond and Addesa, 2020), club rankings (Madalozzo and Berber Villar, 2009), and past team successes (Besters et al., 2019), have been shown to affect the demand for live sporting games. These team-related factors tend to influence the persistent nature of audience habits in sports viewing.

Recently, with the widespread adoption of online streaming platforms, accessing highlight videos of games has become increasingly convenient. Unlike the experience of live games, sports highlights on streaming platforms are no longer just an “instant” experience but are “infinitely replayable” for viewers at any time after the game is over (Buehler and Marston, 2023). Therefore, viewers can access “self-scheduled, time-shifted, and on-demand” streaming of highlight videos (Spilker and Colbjørnsen, 2020), significantly contributing to viewing convenience and, consequently, the likelihood of viewership. Additionally, highlight videos produced and posted after live games typically focus on post-game editing, highlighting the most interesting scenes from the game (Pan et al., 2002). Viewers of streaming highlight videos can replay them repeatedly and even control the playback speed to review the most captivating moments (Chen et al., 2018).

This is in sharp contrast to televised live games, which selectively showcase specific games, with an even smaller number of games featured in highlights that fans need to tune into at the time determined by broadcasters. Therefore, the viewing behaviors of televised live games and highlights are unlikely to be the same as the viewing patterns of the highlight videos available on streaming platforms. This suggests that the determinants influencing the demand for live games may differ from those affecting streaming highlight videos.

Some variables, such as those associated with team attributes, influence both the demand for televised live games and post-game streaming highlight videos. For instance, factors such as historical team performance and the presence of star players—typical team attribute factors—have been found to significantly impact both live TV ratings (Schreyer et al., 2018) and highlight views on streaming platforms (Butler and Butler, 2023; Ryu et al., 2023). However, for live games, the decision to watch is made before the game starts, whereas post-game highlights viewership is influenced by what happens during the game. Therefore, while ex-ante variables are crucial in determining the demand for live games, ex-post variables significantly impact the demand for post-game highlights.

Determinants of post-game streaming highlight videos

As mentioned earlier, the demand for streaming highlight videos is predominantly influenced by ex-post variables, specifically the attraction or interest factors that unfold during the game; these factors are represented by in-game statistics that capture the essence of the game by highlighting important moments, key decisions, and player performances that occur during the game. The narrative and drama embedded around such statistics give post-game highlight videos a deeper appeal and drive viewers’ motivation to watch and replay the videos.

The most notable in-game statistics that may significantly influence post-game highlights viewership include factors associated with scores (Madalozzo and Berber Villar, 2009; Martins and Cró, 2018; Paul and Weinbach, 2007; Pawlowski and Anders, 2012), fouls (Butler and Butler, 2023; Jewell, 2009), and unexpected outcomes (Card and Dahl, 2011; Ge, 2018; Matti, 2021). These factors are key elements that add enjoyment to watching unscripted drama by generating suspense and arousal, leading to an increased demand for highlight views (Peterson and Raney, 2008). This is because the literature on sports spectatorship suggests that more scoring moments increase fan engagement (Jang et al., 2018; Madrigal, 2003), a greater number of fouls reflects the toughness of the game (McCarrick et al., 2020; Zillmann et al., 1989), and unanticipated defeats or victories evoke stronger emotional responses (Archer and Wildman, 2020).

In addition to the aforementioned ex-post variables, factors associated with team attributes can also influence post-game highlight viewing behaviors. For example, team loyalty tends to have relatively stable and enduring impacts on viewing behaviors and is determined by various team attributes such as past winning records and history, the presence of star players (Besters et al., 2019; Martins and Cró, 2018; Uhrich and Benkenstein, 2010; Yoshida et al., 2021), and the derby status of the game (Barajas et al., 2019; Buraimo and Simmons, 2009; Madalozzo and Berber Villar, 2009; Gasparetto and Barajas, 2018). These factors, reflecting team attributes, represent fans’ preferences that are relatively stable and enduring and can affect fans’ viewing behaviors independently of in-game statistics that represent what has happened during the game.

There are also factors external to the teams and the games themselves that still have significant influences on the demand for streaming highlight videos. The availability of the live game on TV is one such factor. The live telecast of a game may influence the demand for post-game highlight videos in both directions. On the one hand, fans’ consumption of highlight videos on streaming platforms may drive interest and excitement in the sports league (Novy-Williams, 2023), which, in turn, boosts the demand for live telecasts of the league games. On the other hand, watching live sports programming and viewing game highlights on streaming platforms may act as substitutes for each other (Zilber, 2016); this is because people, especially the younger generation aged 18–24, increasingly prefer watching highlights rather than the entire sporting program (Sim, 2022).

In addition, when live viewing becomes inconvenient due to the game schedule, it may result in a greater number of viewers turning to post-game highlights, suggesting a substitution effect. Likewise, game schedules such as weekend or prime-time games may influence highlight viewership, as they affect the demand for live TV sports (Buraimo et al., 2018; Cox, 2018; Martins and Cró, 2018). Finally, researchers have found that the release date of a video affects its streaming demand, with longer release durations correlating with higher viewership (Munaro et al., 2021). The findings from the studies reviewed in the literature are summarized in Table 1.

Study context

This study analyzes the factors influencing the demand for post-game highlight videos of CSL games on CCTV Streaming Media, Tencent Sports, and CSL’s official Weibo account—the three official streaming partners of the CSL. It is worth investigating the demand for streaming highlight videos in the CSL context. First, China is an emerging market in the global football industry. With significant investments in football infrastructure, development programs, and international partnerships, China is rapidly growing its presence and influence in the global football industry (Lee and No, 2022; Sullivan et al., 2022; Ye and Jarvie, 2023). Studying the demand for football highlights in this burgeoning market therefore provides valuable insights for those seeking business and marketing opportunities in association with football in China.

Second, China offers a unique context for examining streaming highlight videos due to its stringent control over unauthorized streaming content. The state enforces strict regulations to ensure that only authorized platforms can distribute highlight videos (He, 2022; Li et al., 2012). As a result, the three streaming platforms analyzed in this study—Tencent Sports, CCTV Streaming, and Weibo—are the exclusive providers of CSL highlights online. This exclusivity allows for a controlled environment to examine the factors influencing viewership, free from the confounding effects of unauthorized content distribution.

Third, understanding the dynamics of highlight video consumption in China can offer broader implications for digital media strategies in more open and competitive environments. The insights gained from China’s controlled market can be used as benchmarks by platforms operating in other markets to help optimize their content strategies and improve user engagement with sports highlights.

Finally, the three streaming platforms analyzed in this study cater to distinct audience demographics. CCTV, a traditional TV company operated by the Chinese state, enjoys a broad and loyal viewership base. In contrast, Tencent Sports, owned by the internet giant Tencent, and Weibo, China’s largest SNS platform, each boasting hundreds of millions of active users, primarily attract the digitally savvy younger generation. These differing core consumer profiles among the platforms may lead to varying patterns of demand for highlight videos, offering valuable insights for tailoring media production strategies to audience characteristics.

Data

The data was collected during the CSL seasons of 2022 and 2023. Viewership data for the CSL highlights were gathered from CCTV Streaming and Tencent Sports, the two official streaming platforms of the CSL, as well as from the CSL’s official Weibo account, China’s most popular SNS, on which CSL-produced highlight videos are streamed. Global streaming platforms such as YouTube are not available in the Chinese market, and unauthorized streaming of CSL videos is strictly controlled (He, 2022; Li et al., 2012). Therefore, there are no legal alternative methods to view CSL highlights online other than the two official streaming platforms and CSL’s Weibo account. Highlight view counts were collected in 2024 after the two football seasons under examination had finished. The highlight views of CSL games peak on the date of release, followed by a rapid decrease over 2–3 days, and almost level off after a week. The data was collected after the view counts for the highlight videos had leveled off.

In the 2022 season, the CSL had 18 teams, each playing 34 rounds of games, resulting in 306 games. In the 2023 season, the CSL had 16 teams, each playing 30 rounds of games, resulting in 240 games. Therefore, the total number of games played during the two seasons was 546. Tencent Sports produced highlight videos for 401 of the 546 available CSL games, while CCTV Streaming was more selective, producing only 305 highlight videos for its streaming platform. The CSL’s Weibo account posted a total of 519 highlight video clips. For this study, the view counts of each highlight video from the streaming platforms of Tencent Sports and CCTV Streaming, and the CSL’s Weibo account were collected. All data were collected at the individual game level. Therefore, all available highlight data during the analysis period were collected instead of sampling a subset of the population.

Models and analyses

Three regression models were employed to analyze the factors influencing the demand for CSL highlight videos on the three streaming platforms. The dependent variables in these models are the highlight view counts on Tencent Sports (i.e., TENCENT) and CCTV Streaming (i.e., CCTV), and the CSL’s Weibo account (i.e., WVIEW), respectively.

When the dependent variable is count data, as in the current study, appropriate estimation techniques include the Poisson model (Coxe et al., 2009), the negative binomial regression model (Liu et al., 2023), or the zero-inflated negative binomial model (Loeys et al., 2012).

In the current study, the zero-inflated negative binomial regression analysis, typically used when there are excessive zeros in the dependent variable, was considered inappropriate because the highlight view counts in our dataset do not contain zeros. Consequently, we examined the variance-mean ratio to assess the suitability of the Poisson and negative binomial regression models. These models vary in their assumptions regarding the conditional mean and variance of the dependent variable.

The Poisson model assumes equality of the mean and variance of the data and cannot adequately handle over-dispersed data (Hanssen and Jørgensen, 2015; Didegah and Thelwall, 2013). Over-dispersed data indicate that the variability in the data exceeds the variance predicted by a specific statistical model, which occurs when there are extra sources of variability or when the model assumptions are not fully met. Negative binomial regression models, in contrast, do not assume equal mean and variance and are particularly suited for over-dispersed data, where the variance exceeds the conditional mean (Piza, 2012; Osgood, 2000).

In our study, we initially conducted Poisson regression models for the three dependent variables. As will be reported later, the data appeared to be over-dispersed as the variances were larger than the means in all three Poisson models. Consequently, negative binomial regression models were employed to analyze the data. Additionally, the selection of variables in the regression models was based on theoretical reasoning and previous empirical evidence to ensure that the models capture the true relationships between the variables without being confounded by endogeneity. The means and standard deviations of the number of teams covered by each of the three streaming platforms were also inspected. The mean number of games covered per team, with standard deviations in parentheses, was as follows: 35.5 (SD = 4.27) for CCTV, 60.9 (SD = 1.59) for Weibo, and 45.79 (SD = 3.19) for Tencent, suggesting no clear pattern of any platform disproportionately covering certain teams.

Independent variables

A total of 17 independent variables were regressed on the highlight video view counts. These independent variables were grouped into three categories: game attributes, team attributes, and external factors. The selection of these categories was based on the limited literature on streaming media demand research (Butler and Butler, 2023; Ryu et al., 2023), as well as relevant studies on sports broadcasting demand (Buraimo, 2006, 2008; Tainsky and McEvoy, 2012; Tainsky, 2010). Table 2 displays the list of variables and their operational definitions. Unless otherwise stated, all explanatory variables related to CSL games were obtained from the CSL website (www.thecfa.cn).

The explanatory variables related to game attributes include SGOAL, DGOAL, YCARD, RCARD, BETRO, and THEIL; these are ex-post variables reflecting in-game statistics related to scores, fouls, and unexpected outcomes, which capture the unscripted drama of the games and can only be known after the game is over. SGOAL and DGOAL represent the sum of goals and the difference in goals between the two teams at the end of the game, respectively. A significant discrepancy in goals or a game with a higher number of goals can appeal to the audience, prompting them to click on the highlight video after the game. YCARD and RCARD measure the number of fouls in terms of yellow and red cards received, reflecting the intensity of the game. BETRO is a variable representing the reverse bet outcome, referring to games where the actual outcome is contrary to the prior information or predictions indicated by the betting odds. This allows us to examine whether surprising outcomes in the games attract more fans and viewership. THEIL is a measure of outcome uncertainty suggested by Theil (1967) based on the betting odds data times 100, which has been used in previous broadcast demand literature (Buraimo and Simmons, 2015; Schreyer et al., 2017; Schreyer et al., 2018). The betting odds data used for the variables BETRO and THEIL were obtained from the “BetExplore” website (www.betexplorer.com).

The explanatory variables related to team attributes include WEIBO, CHA, AGE, SALARY, STARP, DERBY, SPOINT, and DPOINT. These team attribute variables are known before the game begins and are therefore ex-ante variables, reflecting fans’ preferences that are relatively stable and persistent over time. WEIBO represents the number of Weibo followers of the two teams and was recorded from the Weibo website (https://m.weibo.cn). WEIBO reflects the teams’ social media presence and fan engagement outside of games. Therefore, a larger following on Weibo suggests a stronger fan base and greater interest in the team’s activities, potentially leading to higher view counts for their game highlights. CHA represents the cumulative number of times both teams have won the CSL championship, serving as an indicator of the historical success and prestige of the teams. This variable was chosen because the historical performance of teams can significantly influence viewer engagement and perception of game quality (Jane, 2016). AGE is the sum of the two competing teams’ ages. This variable was selected to capture the overall experience and maturity level of the teams participating in the games. Older teams may have amassed more knowledge and skills, which can enhance their performance and appeal to the audience over time.

SALARY represents the sum of the average salaries of players from both teams, reflecting the financial resources and investment in player talent by the teams. Higher salaries are associated with greater on-field performances of the players (Carmichael et al., 2011; Hall et al., 2002), making the game more attractive to audiences. STARP is the sum of the number of star players from both teams. The top ten international players and the top 10 domestic players were identified as star players based on transfer fees provided by Transfermarket (www.transfermarkt.com). All top ten domestic players are part of the Chinese national team and include the most influential Chinese players. Star players are recognized for their exceptional skills and reputation and often draw attention and generate interest among fans (Gasparetto and Barajas, 2018); therefore, their presence can elevate viewership for highlight videos.

DERBY indicates games between teams from the same local city. DERBY games often generate heightened excitement and interest among fans due to local rivalries and historical significance (Tyler and Cobbs, 2015; Wilson, 2012).

SPOINT is the sum of both teams’ total points per game until the previous game in the current season, assessing the current performance of the two teams involved in the game. DPOINT is calculated as the absolute difference between the two teams’ total points per game until the previous game, reflecting the difference between the two teams’ performances in the current season. These variables were chosen because they reflect the current standing and competitiveness of teams in the league. Teams with higher point totals or smaller point differences are likely to be involved in closer games with greater performances, which can attract more viewership for their highlight videos.

Finally, there are explanatory variables not related to the in-game statistics of the game or team attributes but that may still affect the demand for highlight views; these factors are named “external factors” in Table 2 and include CCTV5, WEEKEND, NIGHT, S2023, and VTIME. Specifically, CCTV5 indicates whether the live game has been broadcast on CCTV5. WEEKEND and NIGHT represent whether the game was held during the weekend and night, respectively. S2023 is a dummy variable representing the football season (2022 vs 2023). VTIME represents the number of days from the release of the highlight video to the time of data collection. Highlight videos that have been available on the streaming platform for a longer duration are likely to have had more opportunities to attract viewers and accumulate views. VTIME was included to control for the effects of the time lapse between the upload of the highlights and the time of data collection.

Descriptive statistics

Table 3 presents the descriptive statistics of the key variables. The highlight view count of CSL games on the three platforms varies from 981 to 1,783,068 views, indicating significant disparities in viewership. It is worth noting that the highest viewership count for a single-game highlight video is observed on Tencent Sports.

The number of followers on Weibo for teams participating in the CSL exhibits significant variation. The maximum value is 13.555 million, while the minimum value is only 41 thousand, indicating a substantial difference in the fan bases of the CSL teams. Moreover, there is a notable disparity in the number of CSL titles won in the past, with the top two teams, Shandong Taishan F.C. and Guangzhou F.C., having a combined total of 12 titles. The age of participating teams can vary significantly, and if it involves two experienced teams, their combined team age could reach up to 60 years old. Additionally, the number of star players in a game can reach as high as 11 for certain teams, while other teams might not have any star players. Moving on to the game attributes, it is worth mentioning that a single game in the CSL can result in up to 12 yellow cards being awarded. Furthermore, the maximum goal difference in a single game has reached as high as eight goals.

Model tests

A total of 18 independent variables were regressed on the highlight view counts of Tencent Sports (401 games), CCTV Streaming (305 games), and the CSL’s Weibo account (519 games). As the dependent variables were count data, Poisson regressions were initially performed on each of the three dependent variables, respectively. The means and variances of the dependent variables TENCENT, CCTV, and WVIEW were examined to ascertain adherence or proximity to the assumption of equality between the mean and variance. The results show that the mean value of the dependent variable TENCENT is 28,130.57, and the variance is 8.83e + 09. This indicates that the variance is much larger than the mean value, suggesting the data is over-dispersed, and that the Poisson model cannot be used. Similar results are derived from the remaining two dependent variables, with means of CCTV and WVIEW being 87,800.66 and 19,067.98, and variances 9.51e + 09 and 1.40e + 09, respectively.

Accordingly, negative binomial regression models were performed, and the null hypothesis predicting that the over-dispersed parameter alpha equals zero was tested. The estimated value of alpha for TENCENT is 0.894, which is greater than zero, and the 95% confidence interval is (0.727, 1.099). Therefore, the null hypothesis is rejected, suggesting that the negative binomial regression model is appropriate. The regression model was statistically significant (LR χ²(19) = 151.19, p < 0.01). Similar results were found for the other two dependent variables, CCTV and WVIEW. The estimated alpha for CCTV and WVIEW were 0.986 and 0.413, respectively, which were both statistically significant at a 95% confidence interval. The regression models were statistically significant for both CCTV (LR χ²(19) = 82.62, p < 0.01) and WVIEW (LR χ²(19) = 90.41, p < 0.01).

Finally, the variance inflation factor (VIF) was used to examine multicollinearity between variables. The results indicate that the VIF of the independent variables was well below 10, suggesting no serious multicollinearity between the variables (Midi et al., 2010; Senaviratna and Cooray, 2019). Additionally, tolerance tests were performed. The tolerance value, which is the inverse of the VIF, estimates the proportion of variance that can be explained independently by a given independent variable. Lower tolerance values (usually < 0.1) indicate that the independent variable may have a collinearity problem (O’Brien, 2007; Wondola et al., 2020). The tolerance values of the independent variables range from 0.879 to 0.210 for the CCTV model, from 0.945 to 0.202 for the Tencent model, and from 0.936 to 0.261 for the Weibo model. Also, the correlations among the 18 independent variables did not show extremely high correlation coefficients, with the highest correlation observed between SALARY and STARP (r = 0.80, p < 0.01), followed by the correlations between WEIBO and AGE (r = 0.66, p < 0.01), and between WEIBO and CHA (r = 0.63, p < 0.01). These findings do not provide clear evidence of collinearity.

Analysis of Tencent sports

The results of the negative binomial regression model on highlight view counts of Tencent sports are shown in Table 4. Along with the regression coefficients, the incidence rate ratios (IRRs) of the negative binomial regression model were also calculated and presented.

It appears that nine variables (i.e., YCARD, RCARD, BETRO, CHA, AGE, SPOINT, DPOINT, S2023, and VTIME) controlled by the model had a significant influence on the view counts of highlight videos on Tencent Sports. All other independent variables were not significant.

Concerning game attributes, highlight videos with fewer yellow cards (β_YCARD = −0.073, p < 0.05; IRR = 0.930) but more red cards (β_RCARD = 0.430, p < 0.10; IRR = 1.537), and unexpected outcomes (β_BETRO = 0.359, p < 0.05; IRR = 1.433) led to higher view counts. In terms of IRR, an additional yellow card resulted in a 7% decrease [(1 − 0.930) × 100] in the highlight view counts, while an additional red card led to a 53.7% increase [(1.537 − 1) × 100] in the highlight view counts. Unexpected outcomes resulted in a 43.3% increase [(1.433 − 1) × 100] in the highlight view counts. These results indicate that viewers are attracted to intense gameplay and surprising results, as reflected by the influence of red cards and unexpected game outcomes. It is interesting to note the seemingly conflicting findings for yellow and red cards, which will be further elaborated on in the discussion section.

Regarding the team attribute variables, viewers favored highlight videos of teams with better historical performances (β_CHA = 0.087, p < 0.05; IRR = 0.1.091), smaller sum of team age (β_AGE = −0.032, p < 0.01; IRR = 0.971), and greater sum of (β_SPOINT = 0.020, p < 0.01; IRR = 1.020) and greater difference in (β_DPOINT = 0.015, p < 0.05; IRR = 1.015) the two teams’ points per games, reflecting fans’ preferences for more dynamic teams with greater past success and more unbalanced match-ups.

Finally, for the external factors, season 2023 (β_S2023 = −0.449, p < 0.10; IRR = 0.638) had a negative impact on view counts, suggesting that the highlight videos of 2023 games had fewer view counts compared to those of 2022. Additionally, the longer the highlight video has been available on the streaming platform (β_VTIME = 0.011, p < 0.01; IRR = 1.010), the more likely the fans would view them.

Analysis of CCTV streaming

The results of the negative binomial regression model on highlight view counts of the CCTV Streaming are shown in Table 4. The results indicate that six variables (i.e., BETRO, WEIBO, CHA, WEEKEND, S2023, and VTIME) controlled by the model had significant influences on the highlight view counts of the CCTV Streaming. All other independent variables were not significant.

Concerning game attributes, highlight videos with expected outcomes (β_BETRO = −0.250, p < 0.05; IRR = 0.779) were associated with higher highlight view counts, suggesting viewer attraction to highlights with greater predictability. Regarding the team attribute variables, viewers favored highlight videos of teams with a greater number of Weibo followers (β_WEIBO = 0.001, p < 0.05; IRR = 1.001) and a smaller number of past championships (β_CHA = −0.155, p < 0.05; IRR = 0.857). Finally, highlight videos for the weekend (vs weekday) games (β_WEEKEND = 0.252, p < 0.05; IRR = 1.287) and games held in season 2003 (vs 2002) (β_S2023 = 0.437, p < 0.05; IRR = 1.547) were associated with more highlight view counts. In addition, the longer the highlight video has been available on the streaming platform (β_VTIME = 0.005, p < 0.05; IRR = 1.001), the more likely the fans would view them.

Analysis of CSL’s Weibo account

The results of the negative binomial regression model on highlight view counts of the CSL’s Weibo account are shown in Table 4. The results indicate that eight variables (i.e., DGOAL, BETRO, CHA, SPOINT, WEEKEND, NIGHT, S2023, and VTIME) controlled by the model had significant influences on the view counts of CSL’s Weibo account highlight videos. All other independent variables were not significant.

Concerning game attributes, highlight videos with greater goal differences (β_DGOAL = 0.098, p < 0.05; IRR = 1.103), and unexpected outcomes (β_BETRO = 0.222, p < 0.05; IRR = 1.248) led to higher view counts, indicating viewer attraction to surprising results with great goal difference. Regarding the team attribute variables, viewers favored highlight videos of teams with better historical performances (β_CHA = 0.074, p < 0.05; IRR = 1.077), and greater sums of the two teams’ points per game (β_SPOINT = 0.012, p < 0.05; IRR = 1.012), reflecting fans’ preference for teams with greater historical success, as well as their recent success during the observed season’s match-ups. Finally, highlight videos for weekend (vs weekday) games (β_WEEKEND = −0.297, p < 0.01; IRR = 0.743), games held after 6 p.m. (β_NIGHT = −0.241, p < 0.05; IRR = 0.786), and the games held in season 2003 (vs. 2002) (β_S2023 = 0.559, p < 0.01; IRR = 1.749) were associated with more highlight view counts. In addition, the longer the highlight video has been available on the streaming platform (β_VTIME = 0.007, p < 0.05; IRR = 1.007), the more likely the fans would view them.

This study examines the factors affecting the demand for CSL highlight videos on three major streaming platforms in China—Tencent Sports, CCTV Streaming, and CSL’s official Weibo account. The results of three negative binomial regression models reveal different patterns of demand for the three platforms, suggesting that the demand for highlight videos depends on the specific streaming platform delivering them.

The likely difference in the core customer base could have led to such different patterns of demand. Statistics from the industry analytics company Qianfan indicate that a significant 74.04% of Tencent Sports users fall within the 24–40 age bracket, with males comprising 92.68% and females comprising 7.32% of the user base (Qianfna.tech, 2024a). Tencent Sports targets a digitally savvy, younger demographic, emphasizing quick, visually engaging highlights and individual player performances (Coe and Yang, 2022; Hutchins et al., 2019).

Conversely, CCTV’s streaming platform, serving as the inaugural national 5G new media flagship, has garnered over 526 million app downloads and boasts 214 million users as of 2023 (National Radio and Television Think Tank, 2023). Being a nationally recognized new media platform, it enjoys a broad audience spanning all age groups. Thus, CCTV Streaming, functioning as the state broadcaster’s streaming platform, caters to a more diverse and traditional audience, typically with a higher average age.

Meanwhile, Weibo, China’s premier social network, had garnered 80 million sport-interested users and over 350 million pan-sport enthusiasts, generating over ten trillion sport-related blog posts as early as 2017 (Social Marketing News, 2021). Among Weibo’s sports users, the primary age group tends to be 18–30-years-old, with football and basketball commanding the greatest attention. Notably, leagues such as the NBA, Europe’s major five, and the CSL enjoy fervent engagement from users (Data Extravaganza report, 2022). A significant 91.76% of Weibo users are under the age of 40, generally younger and more energetic (Qianfna.tech, 2024b), which aligns with the CSL’s primary target audience on the Weibo platform. According to the White Paper on the Chinese Football Association Super League, jointly released by the CSL and Deloitte, CSL fans are socially active and pay close attention to various aspects of the league, including games, player transfers, and league progress; these fans primarily engage with mainstream social media platforms, such as Weibo (Deloitte, 2021).

These differences in core consumers could have led to Tencent Sports fulfilling the immediate, highlight-focused preferences of its younger audience, CCTV Streaming appealing to viewers with more traditional sports viewing habits, and CSL’s official Weibo account providing a more professional and comprehensive information platform for loyal CSL fans.

More specifically, given the different profiles of the three streaming platforms, there are several seemingly conflicting findings that deserve further discussion. First, a greater number of yellow cards (i.e., YCARD) was associated with smaller highlight view counts, while an increased number of red cards (i.e., RCARD) was associated with greater highlight view counts on Tencent Sports. The positive impact of red cards is consistent with previous research findings (Butler and Butler, 2023), which suggest that significant events during games, such as player ejections or game-changing decisions by referees, generate heightened viewer intrigue and engagement. Given that Butler and Butler’s (2023) data were collected from YouTube viewers of EPL highlights, the findings suggest that typical YouTube viewers exhibit similar demand patterns to Tencent viewers, warranting further empirical investigation. Meanwhile, the observed negative impact of yellow cards, given up to 12 times a game as shown in Table 2, on highlight views suggests that Chinese football fans prefer to avoid too many interruptions in gameplay due to the provision of yellow cards, which are not as influential in changing game results as red cards.

Second, it is notable that while unexpected outcomes (i.e., BETRO) contributed to higher highlight view counts on Tencent Sports and CSL’s Weibo account, the opposite effect was observed on the CCTV streaming platform. Given that Tencent’s core users are typically younger and seek stronger stimuli, while Weibo users are often loyal and knowledgeable CSL fans, it is not surprising that they prefer unexpected game outcomes over easily predictable ones. This viewership pattern reflects the preferences of younger Chinese generations (Han, 2018; Xu et al., 2023). These findings align with Ryu et al. (2023), who found that esports viewers, who are typically young, enjoy watching highlight videos when underdog teams win. However, users of CCTV Streaming showed a preference for expected outcomes over unexpected ones, consistent with findings by Paul and Weinbach (2007) and Sung and Mills (2018). The typical users of the CCTV streaming platform are generally older, more conservative, and have diverse backgrounds, thus tending to avoid results that contradict their predictions and prefer outcomes that confirm their expectations (Li et al., 2023; Tang et al., 2009).

Third, interestingly, past championship records (i.e., CHA) positively impacted highlight view counts on Tencent and CSL’s Weibo accounts but negatively impacted views on the CCTV streaming platform. The positive relationship between Tencent and Weibo aligns with literature showing fans are drawn to successful teams. Wann and James (2018) note that winning breeds a sense of pride and attachment among supporters, which often translates into increased engagement with the team’s content. The negative relationship observed for the CCTV streaming platform may be attributed to the broad reach of the CCTV platform, which encompasses a much larger fan base beyond just supporters of past championship-winning teams. CCTV enjoys extensive coverage and influence in China, enabling it to capture a diverse audience with varying demographics and backgrounds (Meng et al., 2023). Therefore, for each highlight video, there might be a greater number of viewers who support teams with little to no championship history than those supporting teams with past championship trophies.

Fourth, weekend games (i.e., WEEKEND) were associated with higher highlight views on CCTV but fewer views on CSL’s Weibo account. On weekends, people typically have more free time due to reduced work or school commitments. As a result, CCTV’s audience, which includes a broad demographic with ample leisure time, is more likely to watch sports highlights, leading to increased view counts. Conversely, followers of CSL’s Weibo account, who are predominantly younger and more loyal CSL fans, may prefer to watch live games, either on TV or at stadiums, during the weekends when they have more leisure time. Therefore, the lower highlight views on weekends for Weibo users can be attributed to their preference for watching live games rather than post-game highlights.

This study yields both managerial and theoretical implications. Managerially, the insights gained regarding team attributes and game dynamics carry significant implications for streaming platforms aiming to monetize sports content. Understanding the factors that influence viewer engagement with highlight videos enables streaming platforms to tailor their content strategies more effectively. By aligning their offerings with viewer preferences, platforms can refine their strategies based on platform-specific nuances and consumer behavior. This approach has the potential to enhance user engagement, stimulate demand, and foster a more devoted and loyal viewer base.

For instance, platforms can leverage data on team attributes, such as historical performance, to craft targeted marketing campaigns and offer personalized content recommendations. Showcasing pivotal moments involving popular teams or players can draw in more viewers and enhance the chances of content sharing and discussions on social media. This not only augments immediate viewership but also bolsters the platform’s visibility and attractiveness to a wider audience.

Moreover, acknowledging the influence of game dynamics, such as the intensity of gameplay evidenced by red cards, can assist platforms in crafting compelling narratives around highlight videos. By accentuating the most exhilarating and dramatic moments of a game, platforms can captivate and sustain viewer interest. This approach can prove particularly effective in engaging casual fans who might not watch complete games but are drawn to thrilling highlights.

The positive correlation between video upload time and view counts suggests an opportunity for managers to optimize content release strategies. Extending the availability of highlight videos can boost engagement, and experimenting with different release schedules and promotional tactics can help maximize view counts and retention. For instance, releasing videos soon after a game and promoting them consistently can encourage repeat viewings. Data analytics can also identify peak viewing times to adjust release schedules accordingly.

Distinct demand patterns suggest that platforms should tailor strategies for sports highlights. For Tencent Sports, catering to a younger, tech-savvy audience, shorter, high-impact highlight reels focusing on key moments like goals or red cards may drive engagement. CCTV, with an audience that prefers tradition and predictability, might benefit from comprehensive highlights offering game context and analysis. Weibo, serving loyal CSL fans, could enhance engagement through interactivity, such as live commentary, fan polls, and behind-the-scenes content. Promoting content during weekdays may also align better with Weibo’s younger audience.

Theoretically, this study represents one of the few endeavors to explore viewers’ demand for highlight videos on streaming platforms. While considerable research has delved into sports demand, primarily focusing on stadium attendance and live TV audiences, the examination of demand for highlight videos is a relatively novel area of inquiry. This research enriches the existing body of literature by bridging a gap in understanding how viewers engage with and consume sports highlights online.

By examining the factors influencing the demand for sports highlight videos, this study offers fresh insights into digital sport consumption behavior. The transition from traditional media to digital platforms necessitates a deeper understanding of how highlight videos are consumed, as these platforms provide unique viewing experiences compared to live broadcasts or in-person attendance. Highlight videos afford viewers the flexibility to watch key moments at their convenience, potentially impacting their overall engagement with the sport. Furthermore, this study broadens the theoretical framework of sports demand by integrating variables specific to digital consumption, such as the role of social media presence and the influence of video release timing. These factors are pivotal in the digital era, where viewers’ attention spans and consumption patterns diverge significantly from traditional media consumption. Understanding these variables can aid in devising more effective strategies for content distribution and fan engagement on streaming platforms.

This study has some limitations that suggest potential avenues for future research. First, the study’s timeframe is limited to a specific period, and the dynamic nature of digital media implies that viewer behaviors may evolve over time. Future research should adopt a longitudinal approach to capture changing trends accurately. Second, this study solely focuses on data collected from digital platforms. Future research should embrace a more holistic approach by considering the interplay between digital and traditional media in shaping viewer behaviors and preferences. Third, multicollinearity between variables was assessed by examining VIF values. However, it is important to note that VIF serves as an indicator and does not entirely eliminate potential collinearity issues. For instance, theoretically, one might anticipate a correlation between the number of fans and the number of titles won, and certain teams may receive more frequent coverage on CCTV5 or during specific schedules than others. Fourth, this study primarily focuses on CSL games. To generalize the findings to other sports leagues or cultural contexts, future investigations should explore various sporting events and leagues across different cultural and national contexts. Finally, there are additional variables that future research should consider examining. These include a dummy variable for games where a player scores three or more goals, as this typically indicates an outstanding performance and could boost viewership; the number of penalties in a game, given their association with drama and controversial referee decisions; the number of twists in the scoreboard, which adds to the drama and storytelling of a game; a variable indicating the game week, as higher demand for streaming might occur at the beginning and end of the season; an indicator for ‘beautiful’ goals, such as a variable for games featuring the ‘goal of the week’ or similar indices; and, lastly, comparisons of the impacts of international stars vs local stars.