By: Oscar O'Brien
Purpose
Rocket League is an esports game that has been rapidly gaining popularity recently after becoming a free-to-play game on the Epic Games Store. It was even one of two games selected to be featured in a tournament (now likely canceled due to COVID-19) hosted before the Tokyo Olympics with a prize of $250,000. Its simple core-concept of cars playing 3-on-3 soccer lends itself to being an approachable esport for people that have never heard of the game. However, the gameplay mechanics are incredibly nuanced and have evolved enormously since the game’s release in July of 2015. The professional gameplay from the Rocket League Championship Series (RLCS) Season 10 looks like a different game from that of Season 1. One aspect of Rocket League that limited its exposure for many years was its lack of large-scale tournaments. The company Psyonix that developed the game was not as big as many other popular esports companies. Since Psyonix was purchased by Epic Games, its access to more significant resources has allowed it to host more frequent professional tournaments. However, due to COVID-19, all tournaments are isolated to individual regions because travel is impractical. Moreover, playing online across regions inherently disadvantages one team because of their poor connection to the server hosting the game. As a result, since the game was released, the top RLCS tournament with teams from North America (NA) and Europe (EU) has occurred eight times, with five wins to Europe and three to NA. Because so few games are played between these two regions, characteristics that make teams successful could differ between the two places.
The purpose of this article is to analyze the gameplay of successful teams in North America and Europe to see if there are striking differences in how the game is played in each region. This article’s data comes from ballchasing.com, and the games are from the RLCS X Split 1 Major for both NA and EU (319 total games). Metrics such as demolitions, boost usage, scoring, movement, and positioning will be used to compare teams. Team success is based on win percentage because that ultimately allows teams to advance deeper into tournaments. The rosters of the analyzed teams are provided on this google sheet.
Breakdown By Each Metric
Demolitions
Demolitions or demos in Rocket League are a game mechanic that occurs when specific conditions are met during the collision of two cars. Essentially, if one player initiates contact while their car is traveling at “supersonic speed” (nearly the car’s max speed), the opposing player’s car blows up. The opponent’s car then respawns a few seconds later, but that time can create an advantageous situation for the player that was not demolished.
Over time, demos have become a greater part of the Rocket League “meta,” or strategies that promote success in Rocket League. Some professional teams, such as Spacestation from North America, emphasize demos in their playstyle, with almost 4.5 demos per game on average (1.5 demos above the league average). It is hard to argue with their mindset from the fact that they won the North America region during RLCS X split 1. However, from the side-by-side graphs below of win percentage compared to average demos inflicted by region, there are many successful teams that hardly utilize demos.
In the community, there is a stereotype that the North American region prioritizes demos more than Europe. There is some truth to that idea because the mean demos inflicted in this dataset for NA is 3.06, while it is 2.85 for EU. Even though the average shows a little discrepancy, the graphs above show the spread between the two regions is fairly equal. Moving the correlation between team success and demos inflicted, the correlation between the two variables is .179 for NA and .071 for EU. These values make sense because there is no obvious linear trend shown on the graphs for both regions.
The two histograms below remove the teams’ groupings from the previous scatterplot and instead display demos inflicted by game outcomes for the two regions.
The data is right-skewed for both regions, but with more than 150 games from each region, the histograms have some degree of normality. Using the assumption that this data represents a somewhat random sample of recent professional Rocket League games, and that the large sample size takes care of the data being discrete, rather than continuous, a two-sample two-sided t-test can be carried out with an alpha value of .05. Within each region, the difference between the mean demos inflicted by winning and losing teams can be compared. But the resulting p-values are both larger than .50, implying that there is no significant difference between the average number of demos for the winning and losing sides. However, it is possible that although there is no difference within each region, there could be a noteworthy difference between North America and Europe. A t-test between the average number of demos inflicted for both regions yields a p-value of .17. As a result, there is not a significant difference in the average number of demos in North America versus Europe with the aforementioned assumptions in mind.
The graph below emphasizes the lack of a difference between the two continents by plotting a histogram of differences in demos inflicted versus those taken from the perspective of the team that won the game from each region.
The blue and red bars have a very similar shape, center, and spread, which lend themselves to the conclusion that the demo meta in professional Rocket League is similar in NA and EU. As a result, the validity of the stereotype surrounding North America’s Rocket League prioritizing more demolitions is called into question. This is especially noteworthy because a non-negligible proportion of the Rocket League community looks down upon demos as if they are an immoral method to achieve success and uses the demo stereotype to argue for European Rocket League superiority.
Boost Usage
Boost is the only resource in Rocket League that can be found around the field. It is very important because it allows players to choose from a wide variety of mechanics to accelerate the ball that are otherwise impossible. It also enables players to move around the field more quickly for offensive and defensive purposes. There are two kinds of boost pads, big and small ones, that are collected when a player drives their car over it and then do not function for a set amount of time. The big pads completely refill a car’s boost storage to 100. Small boost pads increase the car’s boost amount by 12. There are six big boost pads around the edges of the field arranged in a similar manner to the pockets of a billiards table. The small boost pads are scattered all over the place. The image above notes the six locations with big boost pads with red arrows and the two blue arrows highlight two of the many small boost pads around the field. If one team constantly has more boost than the other team, it is easy for the team with the boost to continue an offensive attack because the opponents will have a hard time accelerating the ball for a counter-attack.
One method of measuring boost usage is through boost per minute (BPM) by all three players of a team. The side-by-side scatterplots below depict win percentage versus BPM for both North America and Europe.
It is clearly no secret that successful teams, on average, are able to use a large amount of boost per minute as the top teams from each region are squarely in the upper-right hand corner. However, not all teams that use a lot of boost have a high win percentage. For both graphs, there is some degree of a positive linear relationship that is reflected by correlation values of .321 and .518 for NA and EU, respectively. In terms of the comparison of BPM between the regions, a two-sample two-sided t-test with the same random population assumption as above reveals a p-value of .38. Overall, the average boost usage for each region looks very similar.
Transitioning to focusing only on small boost pad collection by region, there is a noticeable difference in the average value for each region from the plot below.
The mean amount of boost collected from small pads for North America was 2207.17, while in Europe, the value was 2324.91. The histogram below plotting small pads collected supports this difference in means as the red European bars are more skewed right. However, both regions still have fairly normal distributions, with Europe having more high-side outliers. A t-test similar to the carried out for boost per minute reveals a startling p-value of .001. This statistic implies that there is a difference in small pad collection in the two regions. Greater small pad boost collection allows players to maintain a greater midfield presence as they do not have to rely on big boost pads on the edge of the field or in their own corners. It also discourages players from finding themselves out of position defensively if they are not pressured into taking a risk and restocking at a big boost pad.
Building on the idea of defensive problems that stem from offensive pressure, analyzing the amount of boost stolen by game outcomes could describe offensive tendencies for each region. Boost stolen is any boost that is collected on the opponents’ half of the field. The side-by-side histograms below illustrate the amount of boost stolen separated by game outcome for both regions.
Both graphs have a large red spike indicating that the losing team has very little variation in the amount of boost that they are able to steal. This trend makes sense because the losing team often is not the one that sets the pace of the game because they are typically reacting to the winning side. Although the amount of boost stolen by the winning team has a greater spread for NA and EU, the difference between the mean of winning and losing teams within their own region does not produce a significant p-value. However, across the regions, the difference in the amount of boost stolen is supported by a two-sample two-sided t-test value of .018. European teams tend to steal more boost from their opponents than their North American counterparts. One possible effect of keeping opponents with low boost amounts is the ability to have longer offensive attacks that rely less on risk because they are more a battle of attrition rather than one flashy play.
An interesting final comparison of boost usage between the two regions is the boost amount used while at a supersonic speed. Supersonic speed, in this case, is the same supersonic speed from the demolition discussion above, which means players are at or nearly at their car’s maximum speed. If a player is driving straight on the ground at supersonic speed, they will maintain supersonic speed until they turn or brake. There is a slight increase in speed if a player boosts while at supersonic speed, but generally, it is considered wasteful. Boost used at supersonic speed is typically more useful in other situations such as accelerating from low speeds or as a way to quickly recover defensively.
The mean boost amount used while at supersonic speed is 1047.98 and 964.26 for North America and Europe, respectively. The histogram above shows that the blue North American bars have a center that is shifted to the right of the red European ones. This leads to a p-value of .0007 from a two-sample two-sided t-test, which strongly supports the conclusion that there is a difference in boost used while at supersonic speeds between the two regions. However, it may seem easy to argue that the NA playstyle wastes more boost than the EU playstyle, but supersonic speeds can also be achieved in the air, and usually boost is needed to maintain supersonic speed while airborne. Further analysis is required to decide if NA’s use of more boost while supersonic is productive or just wasteful compared to EU’s.
Scoring & Team Plays
Another offensive statistic to consider is offensive efficiency through average shooting percentage. For both North America and Europe, the average shooting percentage for most teams is between 20% and 30% as shown in the two plots below.
The regional champions from each region surprisingly have very mediocre shooting percentages. However, shooting percentage alone does not account for the total volume of shots or the quality of shots. The North American team G2 has the highest shooting percentage from this dataset, and their offensive approach is often highlighted as riskier than that of most teams. They are more willing than most teams to consistently commit all three players to an offensive possession, but their reward seems to be more scorable opportunities. Comparing both regions, the correlation value for North America is .612, and it is .565 for Europe. Intuitively, shooting a higher percentage helps win games in Rocket League similarly to nearly all other sports with a goal or point system. Nevertheless, a comparison of the regions with a p-test is difficult because the distributions of shooting percentages are not normally shaped. As it is not uncommon for a team to be held scoreless, a lot of data points are apart from the main grouping of data at zero percent.
Similar to the average shooting percentage, the distribution of the average assist percentage is even less normally distributed. The histogram below plots the assists per goal by region.
Roughly 80% of the total data is plotted at an assist percentage of 100%. It is no secret at the professional level that even the best offensive players often are not able to score on defenses alone. It normally takes the coordination of passing and boost control to wear down a defense before a goal is conceded. As was the case with the last metric, a two-sample p-test cannot be used since the data is not remotely normal in shape, even with a fairly large data set. Another stereotype in the Rocket League community is that European teams rely more on team passing plays than North American teams that use more solo plays to score. This idea is not supported by the data below. Clearly, offensive success in both regions is founded on effective passing because so much of the data sits at a 100% assist percentage.
Movement
Another variable to consider between Europe and North America is movement around the field. Both movements on the ground and in the air could highlight strategies that vary between the regions. The graphs below plot the average total distance traveled by teams versus their win percentage split between NA and EU. The units of distance traveled are called unreal units because the game engine of Rocket League is called Unreal Engine.
Immediately, it is clear that the spread between the data from Europe is much greater than the North American data. The European teams BDS, Vitality, and Triple Trouble’s average distance traveled across the 1.7 million unreal units threshold, which no North American teams have surpassed. Interestingly, the correlation between these two variables is starkly different in the two regions. For North American teams, there is no obvious linear trend with a correlation value of -.015, although the top three teams in terms of win percentage tend to cover more distance than most teams. However, for European teams, a positive linear association exists with a correlation value of .711. It should be noted though, that the high and low outliers in Europe definitely strengthen what would otherwise be a much weaker relationship. Removing the grouping by teams, these two sets of data produce a p-value of .035, which implies that there is a difference in the total distance traveled between the regions.
Team movement can also be characterized by the speed at which teams tend to drive around the field. The histogram below displays the average time at supersonic speed for European and North American teams.
The plots seem to be centered at roughly the same value, but the European distribution is more right-skewed with high side outliers. The p-value returned by a two-sample t-test is .307, which does not argue for a difference between the true means of these populations. This is an interesting result because the t-test from the average total distance traveled above indicated that there was a significant difference between this data. The game length in Rocket League is five minutes unless overtime is required to separate two tied teams. Since teams in Europe tend to cover more distance than their North American counterparts, but there is no discernible difference in supersonic speed between the teams, it may seem that games require overtime more frequently in Europe. However, the data also looks at what ballchasing.com defines as slow and boost speed. There is no significant difference between the regions for slow speed, but European teams tend to be at boost speed more often with a p-value of .008. It is interesting that one of the three speed ranges defined by ballchasing.com shows a significant difference. Theoretically, if the length of games was constant, then the time at boost speed would be total game length minus time at slow and supersonic speed. However, this is not the case because the time at boost speed was significantly different between the regions, which could be evidence that European games tend to be longer because of overtime. Additionally, more high side outliers for the European data could be artificially raising the mean of the sample data, which does not correspond to the true population mean.
Aeriel play is another factor of a team’s movement that is especially important to professional games as it poses threats that the defense has to proactively prevent. This facet of the game is particularly connected to boost management because boost is what enables a player to “fly” in the air to hit the ball. The graphs below show teams’ average time in the high air (above the crossbar of the goal) versus team win percentage for the two regions.
Similar to the average total distance graphs above, the European data has a much higher correlation value between these two variables of .766 compared to .265. On average, European teams spend five more seconds in the high air per game than North American teams with approximately 70 seconds versus 65 seconds. The p-value for a two-sample t-test for this variable is nearly zero, which stresses a difference in the population means between these two regions. Even looking at time in low air (below the crossbar), Europeans appear to spend more time than their North American counterparts, with a p-value of .0003. It seems like the general European playstyle relies more heavily on aerial plays than the North American playstyle.
Positioning
Positioning is another factor to consider in gameplay strategies from these two regions. Since there are no set positions in Rocket League, all teams employ something called a “rotation.” The general idea is that after a player has attempted to hit the ball (whether successful or not), that player usually should drive back towards their goal and collect boost while their teammates look to hit the ball. The image above shows the Rocket League field roughly split into thirds. The middle third is always called the neutral third. The outer thirds are either called the offensive or defensive thirds based on the perspective of either the blue or orange team. The side-by-side graphs below show how much time teams spend in the neutral third of the field on average compared to win percentage.
Midfield control allows a team to control the big boost pads at the center line and maintain offensive pressure. The scatterplots below have a similar trend to average distance traveled and time in high air above whether the correlation value is much higher for Europe than it is in North America. For EU, the value is .654, and for NA it is 0.056. If the only main grouping of teams in the middle of the European graph were considered, the correlation coefficient would actually be negative, but the high side and low side outliers create a generally positive trend. While the datasets have much different spreads just from looking at the x-axis, the p-value from a two-sample t-test is .143, which does not support a significant difference between the true population means. It is impossible to make any broad statements about some magical threshold for the right amount of midfield control, but the top 3 teams from NA and the top 2 from EU that had the most success in this dataset had above average time in the midfield compared to their regional means.
Switching gears slightly to comparing average time in the offensive third of the field, the histogram below breaks down the data for both regions.
The European data is more right-skewed than the North American data because it has more high side outliers. A two-sided t-test returns the value .049, which just squeezes under the arbitrary alpha value to judge significance. Visually it is not obvious that there is a difference in mean value between the regions, but this data argues that European teams tend to spend more time near their opponent’s net. This result could support the stereotype that European teams prefer a more “war of attrition” attack where they whittle down their opponents’ boost amount with constant offensive pressure.
Building on the belief above that European teams tend to use a more long-term pressure attack strategy, the variable time with ball possession supports this claim. Ball possession is measured by the time from when one team touches the ball until the other team manages to touch the ball.
The data from the graphs below is visually centered further to the right for the European data. European teams tend to have 6.5 seconds more of ball possession than North American teams. It seems surprising that 6.5 seconds is enough to be considered a significant difference with a p-value of .010 when the means for the data sets are between 150 and 160 seconds. However, the data for the two regions have a low spread and are both slightly right-skewed. Such a low p-value suggests that North American teams rely on less pressure to score, or else their general defensive strategy limits teams from setting up the long-term offensive pressure. The Peeps are the only North American team that averaged more than 160 seconds of ball possession, which is something six teams in Europe accomplished.
Another way of measuring offensive pressure comes from how much time the ball spent in a team’s half of the field. This variable is different from ball possession because if a team on defense touches the ball in their own half to establish possession, the ball may never leave their half of the field if the hit is weak or the offensive team quickly regains control. The histograms below depict average time with the ball in a team’s side for each region separated by the game outcome.
Although the European data is much more right-skewed, a two-sided t-test p-value of .104 implies that there is no significant difference between the regions. Isolating the regions and looking for a difference between game outcomes reveals no significant difference in NA, but a p-value of .009 for EU does argue that the ball spends more time in the half of the losing team. This further supports the stereotype that the European offensive meta is more of a “battle of attrition” as one team tries to “boost starve” the other team into conceding a goal.
Team Color
A less serious statistic is the win percentage of each team color. Some players make comments about either orange or blue being a lucky team color for them. Team color is decided by a tournament administrator, so for pool play games, it can be assumed that the colors are randomly assigned. For bracket play games, the official rules do not specifically mention a color for the higher seed, so it is probably fair to assume team colors are still random for those games. In the North American data, the blue team won 52.6% of their games, which means the orange team won 47.4% of the time. The resulting p-value from a two-sided z-test is .428, so there seems to be no significance between these proportions. A similar conclusion is drawn for the European data where the blue team wins 52.8% of their games, while the orange team wins 47.2% of the games. At least across the regions, no color seems to be more “lucky” than the other.
Conclusion
Overall, there do seem to be differences between the playstyles in North America and Europe, mainly in terms of boost usage, movement, and positioning around the field. Although it is important to note that the data for this article was a large sample of professional games at the most recent regional tournament at the time, the games were not randomly selected. As a result, assumptions have to be made to interpret the p-values returned by two-sample t and z-tests.
Metrics that did not show a significant difference between the two regions were demolitions and scoring plays. For further analysis of demos, it would be interesting to look at their usage by both regions over time. By now, they are quite common for European and North American teams to employ, but it is possible that North American teams started using them earlier. That would explain the stereotype about NA teams looking for demos more often. Moreover, for goals per shot and assists per goal, the distributions were not normal enough to compare the sample means. Since shooting accuracy is a fundamental skill in Rocket League that is expected of professional players, it is no surprise that they all do it exceptionally well. Similarly, no professional team will consistently get beat by one player from the other team, which would explain why nearly all goals are assisted; defenses are too solid to concede that easily.
Boost usage was one metric that differed between the regions to some degree. The overall boost per minute used by teams was similar for the regions, but European teams tended to collect more small boost pads and steal more boost from their opponents. This European style supports the belief that European teams like to score through extended offensive pressure rather than a quick, flashy play. With movement around the field, North American teams typically do not cover as much total distance and spend less time in the air. Finally, comparing positioning on the field, European teams usually spend more time in their opponents’ third of the field and maintain more ball possession.
This analysis could be improved upon in the future by including more games in the dataset, so the risk of a large sample size to mitigate nonrandom data effects would be reduced. Additionally, it would be interesting to look at some of these trends over time to see when certain playstyles emerged and were adopted in these regions. Lastly, more analysis could be done with individual players on each team, instead of just looking at macro trends for each team. Ballchasing.com provides so much data that the future is bright for further data analysis that could be interesting to fans but also help teams study their weaknesses.
Source: Ballchasing.com
Comments