How Data Drives Champion Balance - Inspired by An Internal Talk with Riot Champion Balance Team

After interviewing a friend who works for the Riot LoL Champion Balance Team that consistently makes strategic changes to balance the game by reading and analyzing various data, I'd like to share with you all a few keys to understanding the correlations between data and champion balance, from a daily League player and a professional game designer's co-perspective.

First of all, let's define game balance, win rate, pick rate, and ban rate, and then dig into the nuances among these concepts. One of the most important goals of the balance team is for summoners to succeed with any champion. When you pick your champion, ideally you're not supposed to fully predict how good it will be before the game even starts. Under different circumstances, some champions will always be stronger than some of the rest due to team composition, strategies applied, and enemy champion picks. All of these are the most critical components of League games. The ideal bottom line is that players should never feel compelled to pick a certain champion due to how extensively strong it is.

There are so many different aspects to be considered when we rate and categorize how strong a champion is, but eventually, it all comes down to how fast and easy it is to blow down the nexus, that is, to gain the Victory. Different champions might be at their best power at different stages of a particular game: early, mid, or late; some champions might be more useful during team fights; some champions might get penta-kills more easily from certain skills or a wombo-combo; and of course, when players acquire different resources on the map, champions will exert different powers, too. The definition of a strong champion will not stay identical all the time. In a particular ranked game, it depends on the champion definition and player skills. For example, Kog'Maw is not perceived as strong all the time, but in a particular team composition that peels for it, Kog'Maw is capable of melting the enemy team within a few seconds. Similar to how we define the killing and escaping abilities of a champion with K/DA, we also consider how fast and easy a champion can help take down towers and nexus at the end.

You must have wondered at a certain point what win rate means, and whether it is unbalanced when we see win rates not strictly at 50%. If we define the absolute accuracy of the measurement using how strong a champion is perceived, then the relative accuracy can be represented by the champion win rate. Win rate is the possibility to win a ranked game within the current champion and summoner ecosystem. The win rate data is mainly controlled by two factors: champion power and player skills. Neither is in a purely linear relationship with win rate. The mix of non-linear factors and causal relations reveals that a champion could still be balanced without having a 50% win rate. For example, the win rate of champions on weekly free rotation might have been negatively affected by players who lack gaming experience. According to official Riot data, only an extremely small portion of Azir players has actually mastered this champion. Therefore an average win rate below 50% is expected. On the contrary, many Heimerdinger players are truly playing it well, and therefore the win rate goes naturally above 50% as the data has shown.

After knowing how the win rate is affected, the balance team should be dedicated to study and simulate the impacts of game changes on players from all ranks. But it doesn't necessarily mean they can come up with a universal change or solution to resolve all the problems across different ranks. Players from a higher rank are more likely to consider more gaming information and execute it on a more accurate basis. That's the reason the balance team will be closely observing and analyzing data from high-rank players to sufficiently understand game balance and pattern. They believe that the key to the best gaming experience hides in the data from the high-level but not always 100% perfect plays. This also motivates players to improve their champion masteries. Taking Twisted Fate as an example, he's not as easy to play as it seems at the first sight, however, we find him extremely powerful in some games if players have a better understanding and a higher mastery at him. I believe players can gain a great sense of satisfaction from increasing mastery until they truly master the champion, at the same time, the win rate data will be more accurate as well.

We hope you've gained a better understanding of win rates now. Let's move on to the pick rate. Pick rate is the possibility that a champion will be selected in a game. If players only care about victory, pick rate can be simply interpreted by win rate with a play-to-win mindset. While in fact, a lot of players are more likely to pick the champions that attract them more since they desire to enjoy a game itself just as much as a victory. For example, Ahri has a higher pick rate than Urgot as she seems more attractive and interesting to play in a lot of players' perspective. The balance team should also pay enough attention to champion diversity, which effectively reduces the situation where a champion is selected in almost every game, such as Jinx. The basic rule is to never sacrifice balance for diversity, that is, they shouldn't buff champions simply because some champions are more popular among players or having more skins, vice versa.

We've talked a lot about the relations between data and balancing. Apart from cold data, feedback from players is also extremely critical in my opinion. When everyone is complaining about how unbalanced Akali becomes after the rework, the balance team should start to shift their attention to Akali and evaluate if certain adjustments need to be put in place. How does Riot team know if players are complaining in the first place? Apparently the data team can grab some useful information from the ban rate, which is the possibility that a champion will be banned in a game. At the same time, it's important to know that all the rates we've introduced above are mutually affecting each other. For example, Rumble is a strong champion with a relatively low ban rate as players are less likely to pick him over Mordekeiser or Darius. Based on previous experience, the best Ban-Pick plan aka the best plan towards victory is to ban the champions with both high pick rate and high win rate if the ally team doesn't play them. Up to this point, I'd like to assume you're already aware that the ban rate is not simply a result of pick rate and win rate, but it is also affected by players' judgment of how powerful the champion is, whether it is easy to master it, the risk of playing with it, and how bad players might feel if they lose with it. Le Blanc was once on top of the banning list, but she's not the optimal ban at ranks lower than Master/Challenger. Janna as another example should be the most worth banning based on pure data but she has an actual ban rate of only less than 1%. Knowing how different rates work on each other and affect the gaming data itself in return as a circle, the balance team should rely on the ban rate for some of the changes, but not merely on the numbers themselves.

No matter how important the data is, they're still just factors in an equation. I believe the data will consistently help improve the gaming experience for all the players if used correctly, and Riot is aiming to achieve this great goal. But it's a never-ending journey, and in my opinion, another significant factor to boost gaming experience is to improve players' own mentality and perception of win or loss.

Thank you to DOJO Digital, a Bristol SEO Agency for creating this comprehensive blog post and interview. For any questions regarding data and data-driven analysis, please contact their Bristol Digital Marketing team for a free consultation.