AI: The Secrets of Balancing Levels — Desperado Diaries

AI: The Secrets of Balancing Levels

The popularity of a game depends on how successful the user’s journey is. We return to our favorite games because we enjoy the gameplay they offer. But how do we determine the order and difficulty of the game’s levels?

Casual games such as Woka Woka and Viola’s Quest have a great number of levels (only Woka Woka has over 2000), so it is important to balance levels appropriately. A good user journey is what makes players come back to a game and the foundation of that experience is balancing levels.

The golden middle is our goal, based on the central rule that we shouldn’t serve players with levels that are too easy or too difficult. This sounds like a logical conclusion but to reach the projected ‘level heaviness’ is not so easy to accomplish in practice. However, there are some rules that can help us attain our goal.

After a hard level, it’s best to serve several easy ones

Balancing levels  is in fact a constant balance between frustration and boredom, satisfaction and thrill. We would like to provide the maximum level of fun and enjoyment but that won’t happen if players  are frustrated and bored. After a hard level that takes a lot of effort, it is best to follow up with several easy ones so that players  can take a break and the difficulty gradually rises. . We should also take into account the niche of casual mobile games, the difficulty of the levels is limited because we do not provide endless new mechanics that allow players to  practice their skills and adapt to new challenges. For this reason, it is not easy to define the curve or ideal theoretical concept for constructing the right equilibrium. At the end of the day, the focus of our work are the players themselves. 

In-game metrics as constant directives

In-game metrics: conversions, retention and long term monetization make our jobs more complex. Metrics turn the game into a system, which needs almost endless adjustment. Making a small change to a single level is enough to create a different feel for players but it can also potentially disrupt all business-relevant KPIs.

Players’ experience is most important

Of course, no one plays games the exact same way and it depends on the skill level of every player. Some players may find certain levels very difficult, while others complete them with no effort at all. Skill represents an important component in the whole story and  can be developed through time but also cultivated by  new mechanics and concepts. The  question is: is it possible to create a system that  facilitates the work of game designers and simultaneously personalizes players’ experience  on the basis of their skill? Should we just focus on new levels or should we also modify the existing ones?

The majority of data scientists focus on creating new levels adapted to players’ skill.  For creating new levels, the system provides game designers with information about players’ experience for the set configuration of levels, average number of essays to pass the level and similar data. But why do we need 100 levels that  are nicely balanced, if the 200 that precede them are not? We need to make certain changes to  previous levels as well. The next question that emerges is if we should redo levels based on the whole population or personalize them according to the skill of every player? In the second scenario, the user journey  automatically adapts to individual players, so two players who are playing the same level can actually get a different configuration of the upcoming ones. 

All of these questions need to be thoroughly thought out before they  get implemented. If possible, find answers before creating, since they are important for the success of the system.

Danijel Mišulić Data Scientist