3. Research Question
• What are the game features that aect a
player evaluation of story interestingness
and consistency?
Petri Lankoski
4. Method
• Literature review isolating possible candidetes
• Formal analysis, classifying games in terms of
certain features
• Mixed effect ordinal regression
– combining formal analysis and questionnaire data
about players evaluation of story
– Ad-hoc quotas
– Iterative model selection using AIC
Petri Lankoski
12. Conclusions
Story Consistency
• Cut-scenes showing
romance (positive)
• Pre-scripted character
development (positive)
• Player-guided character
development (positive)
• Appearance
customization (negative)
Story Interestingness
• Cut-scenes showing
romance (positive)
• Pre-scripted character
development (positive)
• Support different play
styles (positive)
• Moral choices
(negative)
Petri Lankoski
13. Conclusions
Story Consistency
• Romance modeling
(negative?)
Story Interestingness
• Interactive dialogue
(positive?)
• Player controlled character
development (negative?)
• Appearance customization
(negative?)
Petri Lankoski
Mixed-method research: combining formal gameplay analysis and player modelingContents:1 Why story2 What is this research about3 Method4 Results5 Conclusions
Story: interpretation of the game events as a storyWhy: Many AAA games are “story-driven”. Understanding aspects of game stories
Formal analysis results are shown in tableNot all of the features could be used in the later stages. For example, voluntary side quest with only one example in the no group needed to be dropped. interactive dialogue had two categories that collapsed as one.
Black line is predicted distribution of answers to the question “story was consistent” by male players.Greyed lines are 5%-tile and 95%-tile playersDotted lines show the distribution of the actual answers in the data
Green line shows baseline. This is part that the model does not associate the scores to the specific features of the game. Red and blue lines show the evaluated impact of a specific formal features.
We see that model is not so good as story consistency one. Here we have large confidence intervals, especially in threshold coefficients.Model is more complex and improving model would require larger N.
Black line is predicted distribution of answers to the question “story was consistent” by male players.Greyed lines are 5%-tile and 95%-tile playersDotted lines show the distribution of the actual answers in the dataHere we see the impact of the low quality of the model. In case of Batman Arkham Asylum the predicted probabilities and data does not correspond well, but with Dragon Age, Mass Effect & Skyrim we have better match. N is also higher there.