After analysing the data we are confident that our results corroborate our hypothesis. In almost every case the analysis is in line with our expectations, both Null hypotheses returned where expected and Alternative hypotheses where expected. Essentially, wherever we expected a ‘control’ value to demonstrate no difference between results (for example, level of Immersion between scenarios) we kept the Null hypothesis. This is very useful as for the most part our results are very consistent within the two groups and compared between the two groups. In one particularly compelling case we returned a Null hypothesis for players’ willingness to comfort the NPC. We had expected that the players would attempt to console our NPC early in the simulation, but this was not the case. This is something worthy of study in its own right; while players demonstrated that they liked and were engaged with our NPC, this did not extend to comforting or consoling her.
Another particularly interesting element of the analysis was players’ willingness to speak to our NPC. A Null hypothesis was returned where we had expected a large difference; we had theorised that players who experienced the Bad end first would want to know the NPC’s reasons for betraying them, and they did, but their curiosity was matched by Good end players wanting greater detail on the background of the simulation. This is interesting because the qualitative data gathered (what the players wanted to know) was considerably (whether or not players wanted to speak to her). From simply the quantitative data we’d have missed a valuable conclusion from that data: almost every player wanted to know something, the Bad ending players simply wanted to know something more relevant to the project.
Also relevant was the comparison between players opinion of our NPC, those that did the good end first had a much harsher reaction to the betrayal ending, than those that experienced it first. We suspect that this is because people who played the good end first had a better expectation of what the NPC might do. Those that played the bad end first entered into the negative version of the scenario with no preconception of the NPC. We expected the players’ opinion of the NPC to be lower initially after the first play through if they did the bad end first. But lowest overall if they did the good end first, because it would seem like a greater betrayal when they experienced the bad end. We also expected a slight rise in opinion by those that played the good end second.
The performed analysis of our data, for the most part, allowed for many relevant conclusions to be made and we feel that by using the two-tail comparison analysis method we were able to do this effectively. The analysis however also made apparent that some questions were maybe more redundant than others, with qualitative data received from topics such as running out of ammunition and questioning why candidates fired at enemies usually giving the same negative response. This was perhaps due to our test candidates being primarily gamers, a target audience we had planned to aim for during the creation of the test plan. Despite this, questions linked to character engagement and immersion gave results that generally allowed us to reject the hypothesis that was not in line with our own, ultimately leaving us with solid evidence that can be used to aid in backing up our original project statement.
Click Here to download the teams final group report.
Do you like this article? Share It!
Post a Comment