Play-Persona: Modeling Player Behaviour in Computer Games
Play-Persona: Modeling Player Behaviour in Computer Games
Play-Persona: Modeling Player Behaviour in Computer Games
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
<strong>in</strong>to which sections of the levels that players utilize, and can also be used to locate areas where<br />
players are confused about how to progress.<br />
Note that higher-order variables such as these selected here can sometime mask underly<strong>in</strong>g details<br />
<strong>in</strong> the behavior of the player – it is important to consider exactly what <strong>in</strong>ferences that can be made<br />
from a given gameplay metric. For simplicity, <strong>in</strong> this case study it is assumed that such underly<strong>in</strong>g<br />
patterns do not occur <strong>in</strong> the gameplay metrics data.<br />
There are several approaches that can be applied to locate patterns <strong>in</strong> sample-based datasets (such<br />
as gameplay metrics datasets). The majority of these are based on multivariate <strong>in</strong>ferential statistics,<br />
for example variance analyses, cluster<strong>in</strong>g techniques, ord<strong>in</strong>ation methods and similar approaches<br />
from population studies. Additionally, factor-based algorithms and neural network algorithms can<br />
be utilized to f<strong>in</strong>d the patterns of play <strong>in</strong> game metrics datasets (provided that any patterns are<br />
present).<br />
One of the most direct approaches towards evaluat<strong>in</strong>g if there are any underly<strong>in</strong>g patterns (or<br />
clusters) <strong>in</strong> the collected data is to use a cluster<strong>in</strong>g method. K-means cluster<strong>in</strong>g and Ward´s<br />
method, the latter utiliz<strong>in</strong>g Euclidean coord<strong>in</strong>ates, form two possibilities. A cluster analysis will<br />
<strong>in</strong>form whether there are any strong group<strong>in</strong>gs <strong>in</strong> the way that the underly<strong>in</strong>g variance of the<br />
metrics dataset is organized.<br />
A different approach considers divid<strong>in</strong>g the ranges of each variable <strong>in</strong>to categories (after remov<strong>in</strong>g<br />
outliers). For example, if the “number of deaths” varies between 1-100, a simple b<strong>in</strong>ary division<br />
could def<strong>in</strong>e two ranges: 1-50, 51-100. Care must be taken when manually devis<strong>in</strong>g these ranges<br />
to ensure that they are mean<strong>in</strong>gful. If for example 99% of the players die between 80-90 times<br />
dur<strong>in</strong>g the game, the rema<strong>in</strong><strong>in</strong>g 1% could be considered outliers (or a very m<strong>in</strong>or group of players<br />
behav<strong>in</strong>g <strong>in</strong> a non-typical fashion). By allocat<strong>in</strong>g ranges to all variables of <strong>in</strong>terest, it is possible to<br />
allocate players <strong>in</strong>to play-persona categories based on the requirements def<strong>in</strong>ed <strong>in</strong> the design<br />
phase. Experiences with TRU and other games suggest that there will always be some players who<br />
fall outside the def<strong>in</strong>ed personas. This may or may not be an issue, depend<strong>in</strong>g on the nature and<br />
number of these outliers. However, the ulterior goal is to evaluate if the play-personas <strong>in</strong>tended,<br />
actually manifest <strong>in</strong> the metrics data for a statistically significant portion of the players. If not, this<br />
means that there are problems with the game design, and <strong>in</strong>-depth analysis of the data may be<br />
necessary. It is possible for example that certa<strong>in</strong> game-levels facilitate the k<strong>in</strong>ds of behaviors aimed<br />
for, while others restrict the players´ agency more. This may or may not be a problem and always<br />
requires case-by-case evaluation.<br />
265