A Dynamic Difficulty Adjustment Algorithm With Generic Player Behavior Classification Unity Plugin In Single Player Games

With the game market growing year by year, game developers find themselves in an extremely competitive scenario. To draw players attention towards their game and to engage them even more during gameplay, one alternative is to apply a Dynamic Difficulty Adjustment algorithm. But the problem of the DDA approach is usually not the algorithm itself, but the player classification step. Therefore, we created a generic Unity Plugin that, allied with a Python API, will be able to classify players, using unsupervised and supervised Machine Learning techniques, based on game telemetry. We also implemented our own simple DDA algorithm, to test how it would work allied with the online classification process. The results show that the DDA version outperforms the standard one in the Video-Game category (CEGE Framework). The resultant classification was 63% completely accurate and 100% partially accurate. Moreover, no other work was able to create a generic plugin that simplified the use of ML in the game development context, allowing to test 28 different algorithm combinations.