Machine learning for mobile games: GameTune versus askblu.ai

Now that we have more information about Unity’s GameTune here is a quick comparison of GameTune and askblu.ai, the AI-SaaS solution launched in beta 2 months ago by our team at Happy Blue Fish. Our information about GameTune is public information, available to anybody on the Unity web site and their SDK documentation.

Usage for mobile game studios

GameTune will tune in real time different options of a game design element: different character speeds, different IAP values, different button colors in order to reach a certain target or “reward”: better retention, better conversion, more revenues, etc. It will help game designers to save time while experimenting with game design options.

askblu.ai is entirely designed around a unique goal: improving player retention. The solution is based on the “Cognitive Flow” concept: players stay in the game while they are in their “flow zone” (comfort zone) and leave the game (churn) for 2 main reasons (frustration- too difficult — boredom — too easy — ). This is why askblu.ai is focused on real-time tuning and personalization of game difficulty.

Setup and ease of use for the studios

To use GameTune, the studio must decide which player data to send to GameTune (with the SDK) then it has to create a question in the SDK as well as on the Web portal and finally configure the reward in the web portal. If you want to tune the difficulty of your stages, this becomes tedious: one question per stage in the SDK and in the Web portal. As in casual and hyper-casual games you quite often have more than 100 stages, this becomes quickly tremendous.

To use askblu.ai, the studio has only 3 events to setup in the game (with the SDK), and nothing to setup on the web portal. Stage names are a parameter, so in the web portal you can see the askblu.ai feedback (difficulty ok, too easy, too hard…) for all your stages by name. As soon as a stage is well tuned for the overall player audience, this stage becomes available for player personalization, thus improving the player retention above what can be achieved with just a good game balance.

Features (machine learning variables)

With GameTune, the studio must choose which user data to send and “GameTunes does automatic features discovery, selection, transformation and trains a model with game specific data”. (Features are important variables derived from the user data and used by the Machine Learning algorithms).

With askblu.ai, our team worked for months to find the best features coming from the simplest user data, making the solution as simple as possible for the studio (SDK with 3 events — no guessing — and no web portal setup) and as efficient as possible for the Machine Learning and predictive algorithms (optimized features).

Data

Another important topic: GameTune “is driven by your game’s data but also leveraging Unity’s dataset from 1.5b devices”.

askblu.ai uses only the data from real-players playing a game to tune this game. askblu.ai will never use another studio’s player data to help tune a game. Each game has its own set of features and its own Machine Learning model.

Conclusion

We think that AI and Machine Learning will give more and more competitive advantages to mobile game studios, especially for the casual and hyper-casual segments.

If you are looking for a solution to experiment with game design parameters tuning and have onboard someone with data competencies, ‭GameTune is the solution‬.

If you are looking for a user-friendly, efficient and real-time solution to optimize your game difficulty and improve your player retention through personalization, askblu.ai is waiting for you.