How A.I. will bring us closer to an Ideal Player Journey

It’s all about the Journey. A captivating experience is what we are looking for when we read books, watch movies, or play games. What separates the games from other media is interactivity. In other words, the ability to take an active part in the experience. Influencing or even changing the Player's Journey.

Sounds really exciting, right? The freedom to shape your own Journey is a dream worth chasing. Are we close to this dream already? Not really, in my opinion. I’ll try to explain why and propose a possible solution.

Same Journey Problem

Modern games are working hard to provide multiple ways to play them. Upgrades, weapons or activities that you choose are defining the way you play the game - a Playstyle that suits your personality. In a way, different Playstyles should already produce different Player Journeys but if we look closely it’s clear that they are not.

Actual events happening in the game, challenges and rewards do not change regardless of the Playstyle you choose. Every player gets the same Journey.

Different Playstyles are facing the same challenges and events.

Of course, game developers understand this issue and work hard to resolve it. Here are 2 most common ways to fix this problem:

Manual branching

The game and its levels are designed already with several different Playstyles in mind. This approach provides the smoothest experience for most players but has some downsides:

Each Playstyle is getting different challenges and events but at the cost of stilted gameplay. Example: Dishonored

Randomization

Events and challenges are randomly presented to the Player from a list of possibilities. This provides varied and replayable Player Journeys but, again, not without downsides:

Challenges and events are different but are not influenced by the Player. Example: Spelunky

Both of these approaches have brought some really memorable and enjoyable Player Journeys to us. But still, I feel we can do much better. One important aspect is missing from all modern video games. They do not know their player.

Power of Personalization

This is a quote from the “Dungeon Master’s Guide” - a manual for the world’s most famous role-playing game. D&D is almost 50 years old and this principle was driving it from the start. Indeed, you must understand and recognize your player to offer them an ideal Player Journey.

The modern term for that is “personalization”. You can see more and more examples of it in the everyday life. Personalized news feed your phone builds for you, movies suggestions on Netflix or automated product recommendations on Amazon. All these examples are using advanced A.I. tech at their core.

The Ecommerce industry is especially advanced in this field. Online shops are rightfully considering personalization an essential component of optimal User Experience. The numbers speak for themselves:

When it comes to controlling the Journey games are still using decades-old technologies. Taking the cutting-edge A.I. research from Ecommerce world and adapting the solutions for Gaming will certainly bring revolutionary benefits. And this is exactly what our team is doing.

Ideal Player Journey

We are building an A.I. framework with a powerful Machine Learning component. Among other things, it will provide ways to personalize experiences in real-time. In my previous post empowering Game Analytics insights with A.I. I have introduced the basics of the framework.

Its main power is the ability to easily classify your players and extract their Playstyle. Naturally, each Playstyle is engaging with the game in a different way. You can measure this engagement with the game as a whole but also in detail. For example, in relation to a particular Game Event.

Now back to the Same Journey Problem we are aiming to solve.

Step 1

Let’s first train our A.I. system to understand the problem. We will run an imaginary playtest. Challenges from the examples above will be delivered to our test players in random order. The simplified example below shows an analysis of playstyles A and B during the training phase:

Each Playstyle is obviously enjoying different game events. We may use this information to personalize the game.

Analyzing the engagement we can clearly see the challenge preference for each of the clusters. Of course, given the colors, this seems to be a trivial task, but real-life design problems are more complex and do not have labels attached.

Step 2

Now, after the training phase is done we may use the A.I. output to personalize the gameplay. The resulting Player Journey is getting much closer to an ideal one:

Different Playstyles enjoy different challenges and events

This two-step process shows how a game can self-improve based on the actual players’ behavior. These steps are easy to automate and they could be repeated after any change to game mechanics. The game will be able to re-balance itself and keep providing an optimal experience.

Conclusion

I believe that almost any game can benefit from the approach presented in this post. Personalization and self-learning A.I. systems are driving forward the innovation in almost every industry and it’s time for games to join the movement.

In the next post, I’ll switch from chasing ideal dreams to more vital (for most) use cases. Stay tuned for a shot at Difficulty Balancing and In-Game Store Optimization.