Over the past couple weeks, between returning from Japan and having my wisdom teeth removed, I’ve been pouring over NA LCS VODs (amounting to about 30 hours of straight data collection) from the 2018 Spring Split regular season. I was watching and recording data in order to experiment with some trend visualizations that I thought might be useful for everyone from teams looking to scout opponents, to desk analysts trying to visualize team strengths. Having been fairly successful, I wanted to give a very quick work-in-progress update on the types of things I’ve been looking at, and a look at the visualizations themselves. I should note that all of the visualizations shown here are taken from real data from the 2018 Spring Split, but I estimate that there’s about a <3% error in the exact numbers simply from mistakes due to manual data scraping. Second, everything in the visualizations from colors to size is completely customizable, and I list some potential areas of interest at the end of this article. If anything seems unclear, feel free to reach out with questions/ comments!
But anyways, not wasting too much time, let’s begin with our first visualization(!):
Optic Gaming’s 2018 Spring Split. In the diagram, an arrow pointing from “player A to player B” represents the number of assists from “player A to player B” throughout the entire season. In instances where more than one player provide an assist on a kill, it is counted for each player individually, so the sum of the assists to a player is likely greater than the number of kills.
This first figure (above) is a chord diagram of the assists for Optic Gaming during the spring split. Chord diagrams are incredibly useful in visualizing the relationships between different variables, and are often used in the analysis of genomic data. For competitive League of Legends, I’ve just begun scratching the surface of their ability to represent team tendencies in different ways. In the diagram above, I have highlighted both Arrow’s and PoE’s share of impact on Optic’s kills and assists during all play for the entire season. While Arrow’s and PoE’s assist lines to other players may not be super significant, the visualization does a good job of displaying their central focus for the roster (PoE assumes more influence than LemonNation and Optic’s top lane combined!) based on their share of the circle’s circumference.
Clearly, this is not a perfect visualization, and might just tell you something you already know about Optic’s play style. But the visualization is very attractive and does communicate a lot of information without too much work for the viewer. For a lot of these chord diagrams that I’ve been playing with, their value comes from their customizability and robustness. I’m sure, like me, you have 1,000 ideas about how to use these diagrams to not only visualize something we already might know, but how to reveal something that we haven’t yet realized about a certain team. It’s exciting!
Riding the ideas train for a moment, I wanted to take a closer look at some specifics for teams. Most League games are broken up into discrete phases, especially during “laning,” so being able to visualize team trends during subsets of play is possibly more useful than a full synopsis:
100 Thieves’ laning phase
Echo Fox’s laning phase
Here, another strength of the chord diagram (and most forms of charts) shines though: the comparison of two teams. Visualized and highlighted above (in team colors!) is the assist breakdown for 100 Thieves’ and Echo Fox’s junglers (Meteos and Dardoch respectively) during the laning phase. This is where the chord diagrams really start to shine! They can convey a seasons’ worth of team strategies in such an easy way and give you an easy platform for comparing teams (still, with knowledge you might already have). If assists are an indication of focus, Dardoch is more focused on his mid-lane than his other lanes combined, while Meteos spreads his focus between bot and mid a bit more evenly. Although that is the focus of the visualization, I have two more notes: Adrian’s influence on midlane is much larger compared to Aphromoo’s, and for most teams, the toplane was the lane with the least focus (bot not so for Echo Fox!). Obviously there is a ton to unpack with each of these diagrams, and so getting better at finding surprising trends is certainly a critical next step for me personally.
Finally, I’ll call attention to a simpler, but nonetheless interesting set of charts, which is a comparison of assists binned by game time in a histogram:
A histogram of TSM’s assists for the Spring Split, binned by time
A histogram of Clutch Gaming’s assists for the Spring Split, binned by time.Right off the bat, I’ll re-state that these are a work in progress, and there are many flaws with these charts as of now, especially for comparison purposes. However, the charts are still useful, and for those that accuse TSM of having a very slow play style that leverages late-game team fights, you may be vindicated (at least for last split). All joking aside, I’ll mainly let you draw your own conclusions about these distributions, but I wanted to highlight the different play styles that successful teams can have. Both Clutch and TSM ended the season with the same record before tiebreakers, but clearly try to play to different strengths for their rosters. Will there be a predominant strategy for the top teams vs. the bottom teams, though? I’m not sure, which adds another layer to the interesting conclusions from these simpler types of plots.
Again, without wasting too much time, I’ll reiterate that in this short article, I’ve really only begun to scratch the surface of these visualizations and the analysis I can provide. I think there’s an incredible potential in these and other types of data visualizations and analyses for competitive League of Legends— for everyone from team analysts to entertainers. If you happen to be one of the people who agree and would like to learn more before I begin to publish more thorough insights and visualizations, please feel free to reach out! I am always looking for great conversations and projects surrounding competitive League and would love to chat.
Areas of Interest:
Having most of the data that I currently want for visualization and analysis (better data sets is ket for this type of work), I have a bunch of areas of interest that I will explore in the near future. I have very vaguely highlighted some here with the specifics to be revealed once I explore more:
-
Laning Phase Tendencies (Lane focus, strategy changes over time/patch/roster, timing, etc.)
-
Late-game Tendencies (Carry focus, strategy changes over time/patch/roster, objective contest, etc.)
-
Improved Chord Diagram Formatting
-
Time-binned Histograms for Kills (not assists)
-
Comparing International Opponents (via similar domestic teams) (Would require a lot of nice data)