“Moneyball” in Esports?
You will be able to digest and enjoy the following post better if you have some background knowledge of the game League of Legends(I will refer to it as LOL from now on), but I tried to write the post using accessible terminology so that anyone can understand the overall context.
MVP? Although players who make flashy plays might leave deep impressions on their audience, they may not be selected as game MVPs. Players that are favored by head coaches might not be their teams’ best players. In addition, we might underrate players because of their ill-tempered personalities.
Given that no one can avoid the “danger” of such subjective decision-making, is there any way to make totally objective decisions? If you want to approach gameplay from a data driven perspective, where should you start?
Win Contribution
Imagine that you are the General Manager(GM) of a professional team. What`s keeping you up late at night?
1. Who are the good and bad players on our team? If we plan on scouting, which position should we concentrate on?
2. Are our players competitive in the league?
3. What’s the reasonable compensation level for our players?
These are some of the questions that you must consider as a general manager. However, the lack of objective data makes answering them extremely difficult. A team’s winning rate does not reveal exactly how well the players are playing, and individual KDA (Kill Death Assist) [1] is too simple of a data-point to represent a player’s playing abilities.
We need a new index and it needs to answer the three questions listed above. In my opinion, we need an index that enables us to evaluate match processes in specific contexts. In other words, we need to evaluate the “real” skill of the players. This raises the following important question: how should we define “skill”? I would like to define it as “a player`s contribution level to his or her team ’s winning.” Let’s consider the below metrics.

By isolating each position’s win contribution, we can construct graphs like those above. These graphs lead to the following conclusions.
1. The blue bar represents the average win contribution. The red bar represents win contribution per game.
2. A score of 1 means that your skill is exactly at the league median.
3. This team has strong TOP and ADC players. With a 1.2 in TOP and 1.2 in ADC, the TOP and ADC players’ abilities are 1.2 times better than the abilities of average league players.
4. However, if you look closely, TOP has ups and downs. In his lows, his score dips to 0.6, and, in his highs, the score rises to 2.0. However, ADC has no ups and downs. He consistently contributes to victory in the 1.1 range.
5. The weakest link on this team is SUP because his win contribution is the lowest. At this performance level, we need to reconsider whether SUP has sufficient skill to play in the league. We also need to take into consideration his/her interaction with ADC. SUP’s play is easily influenced by how ADC plays.
6. JUG is playing okay while completing his part of the game. He also doesn’t have any ups and downs. However, we can’t expect JUG to produce a play that could overturn the game situation.
7. The team’s total score is 4.9 and the average score is 0.98. Therefore, the roster is weaker than the league average.
8. We can dig into the reasons for TOP’s ups and downs. For example, TOP’s performance may dip and rise due to different growth rates depending on the lane fight situations.
This is a fairly basic translation of data, but it comes much closer to an objective analysis than a GM or head coach commenting “I think he/she is not playing well.”
The GG Function — GG(x)
GG(x) is an engine that scientifically calculates win contributions. GGtics(http://ggtics.com) has spent the past two years working to ensure its accuracy. [3] If you feed GG(x) with match content data, you can get the match win contributions. The development of the GG(x) engine was based on several different principles, which I outline below:
1. Every move during the match will have an impact on the game’s outcome. Moves can have either positive impacts or negative impacts. Each positive impact will receive a “+” score and each negative impact will receive a “-” score.
2. The same move can have a different impact depending on the time progression. Intuitively, we know that a death just before the end of a match has a much more significant negative impact on the game than a death that happens even before lane fight begins. A death right before the end of the game may cause a team to lose.
3. Different expectations apply to different positions [2]. For example, if we state “ADC needs to commit at least 10 damage in a game,” then this is our expectation. If ADC commits exactly 10 damage, then we give ADC 1 point, and we give ADC 2 points if he/she commits 20 damage, and so on. Additionally, the expectations for ADC differ from those for SUP. For example, the expectation that ADC will “commit damage to the enemy champion” is high, while the expectation of “widening vision by warding” is much higher for SUP.
From this perspective, what moves should the engine really track? Does the engine track every move we make in a game? The answer is “No.” The engine only tracks the key index; the engine decides factors like coefficiency, measurement accessibility, improvement possibility, and more.
“I want to see my dodge rate.”
Q. “Why do you want to see your dodge rate?”
A. “Because I want to know my survival rate.”
Q. “Why do you want to know your survival rate?”
A. “Because I want to survive while committing significant damage in team fights.”
Q. “Why is it necessary to survive in a team fight?”
A. “Because I need to survive in order to pursue objectives like the Tower, Baron, or Dragon.”
… (omitted)
Q. “Why do you need to ….?”
A. “To win.”
Repeating questions like this over and over always eventually brings up a common final reponse: the desire for victory. Victory in LOL involves destroying the Nexus. Given this common goal, let’s trace the process of destroying the Nexus in reverse order.

To destroy the Nexus, you must destroy the Nexus Tower. To destroy the Nexus Tower, you must destroy the Inhibitor Tower. To destroy the Inhibitor Tower, you must destroy the Inner and Outer Towers. To destroy the Inner and Outer Towers, you must win in small and big team fights. This breakdown is somewhat simplified, but we can conclude that LOL is a game in which players gradually reach the goal of “destroying the Nexus” by repeatedly winning team fights.[4]
In this context, a player’s skill depends on how much the player can contribute to wins in team fights. This leads to the question: what kind of abilities do you need to win team fights? After 2 years of examining various perspectives and conducting different analyses, GGtics concluded the core abilities are “Positioning,” “Map reading,” and “Growth.”
“Positioning” refers to how well a player can position himself/herself in a team fight to participate in the fight by dodging opponents’ skills.
“Map reading” refers to how well a player can collect information from the game and use that information to make correct decisions.
“Growth” literally means the ability to grow and this includes winning or losing a lane fight.
Any player will agree that developing skills in these three areas will enable you to win team fights. We therefore draw the following conclusions:
Player skill = Win contribution = Positioning, Map reading, Growth = GG(x)
Recognition and improvement
Once we have confidence in the win contribution calculated by GG(x), we can better assess the current situation and improve it.

1. Understand whether a player lacks skill in positioning, map reading, or growth.
2. For example, a player who lacks positioning skill needs to check if the problem stems from hit, dodging, targeting, or moving issues. Review the index if the problem can be tracked with data, and if it is not for technical reasons, then the coaching staff can examine the situation.
3. Try to use feedback to improve. Improving problem areas will make team fights easier and will eventually result in more team fight wins.
4. Track win contributions and monitor changes; if necessary, repeat the whole process beginning with the first step.
This is similar to the startup cycle embraced by Lean Startup. The important thing to recognize here is that data doesn’t do everything for you. Data will help you see the situation objectively, but, ultimately, the coaching staff needs to learn the nuances of their players and help them improve.

ESports Moneyball

Esports has a short history. The sport’s newborn (relative to other leagues’ professional teams) professional teams are competing and progressing. The high ranked teams can maintain and manage their positions by tracking win contributions. The mid ranked teams can get the “push” they need to advance to higher rankings. The low ranked teams can use this approach as an opportunity to improve by objectively observing their specific situations. Although the win contribution index seems simple, it is extremely useful. If we continuously approach the games objectively using the win contribution index, we may see a revolution in Esports just like the one depicted in the movie, “Moneyball.”
Contact : support@ggtics.com
https://your.gg
Footnotes
[1] “KDA = (Kills + Assists) / Deaths” is a simplified mathematical equation. The more points you receive on Kill/Assists and the more time you play consistently, the more points you receive overall. This index is widely used among LOL gamers due to its simplicity.
[2] There are 5 positions in LOL. AnLOL team is similar to a company that has a business planner, a developer, and a designer who have the same goal but different professional expertise and responsibilities.
[3] GG(x)’s accuracy is measured in various ways. One way involves checking how similar GG(x)’s decision making is to the decision making of Master ranked players (top 0.1% in Korea). When we measured accuracy in this way, we found that GG(x)’s decisions were 90% similar to those of Master ranked player’s. We also measured accuracy in other mechanical ways.
[4] If you are unfamiliar with LOL, you can understand this as what you need to do to reach a certain goal.
