Wednesday, July 25, 2007

Hotel Dusk: Play Session Analysis

This Blog has moved to

This particular Page has moved to

If you are a recurring visitor, please update your bookmarks & rss feeds. We apologize for the inconvience.


Blogger Danc said...

Hi Kristian

This is a very impressive batch of work. I particularly like the association with observed emotion with the various classified sections of the design. This is a great shorthand for describing an observed period of gameplay.

I suspect you've also gotten a hint about why this technique is used sparingly. It is quite labor intensive, especially when you are in an evironment when design changes are occuring quite rapidly.

The common solutions for the sake of efficiency are three fold
- Make the studies less rigorous. Instead of detailed observation and analysis, many teams substitute more informal playtesting. These still yield an immense amount of useful feedback.
- Focus detailed testing on isolated areas. Traditional usability tests take this approach. They'll test a set of tasks that cover the functionality of one section of the UI.
- Automated tests. This one is less common though I've seen more effort put into it recently. The goal here is to automate most of the game so that unpleasant points are highlighted automatically. This allows users to go in with more detailed usability or playtesting to analyze what the problem might be. An example are recent games by Valve, which logged when people were dying a lot. They went back in and tweaked that level to be more friendly. These can also benefit from aggregate statistics.

Great to see this sort of detail in a post.

take care

26 July, 2007 03:27  
Anonymous Impossible said...

It would be a bit more exciting if both, the follow-up verb and the question tokens were more meaningful - if the decision of whether to use them or not was not as straightforward.

This happens a bit more later in the game. I can recall times when picking certain dialogue options led to a game over. This tends to happen at least once near the end of most chapters, although it would certainly be nice if there were more ways to mess up by being careless with your conversation options in the rest of the game. Otherwise, it's not really gameplay, just reading where you pick the order you read things in.

26 July, 2007 08:16  
Blogger Krystian Majewski said...

Thanks you so much for your feedback!

Danc: I agree that this kind of testing might be over the top if you just want to check if a recent tweak in the sourcecode works better or not. On the other hand, I do not believe that we should give up so easily. Just because it is laborious doesn't mean it should be. The reason I quoted Gesche Joost is because she already went a step further and suggested developing software tools to streamline this kind of method, so it can become a fundamental part of the design process. At least - like your models - the very existence of such methods encourages different, more detailed thinking about the game. Just how long will the player stare at that particular screen? We already identified quite negative examples.

Impossible: Yes, it's true. For example you can instantly loose by asking Dunning about the date you find encircled on the calender in his office (Oops! Spoiler?). As you correctly suggested, it is not really gameplay yet: you just pick the order in which you read stuff, sometimes you might fall into a trap, sometimes you can get some nice additional info but otherwise it makes no real difference. The interesting thing is that it doesn't have to have impact on the story in order to feel like you are doing something. You do not necessarily need branching and alternative endings and stuff like that. I think Phoenix Wright shows very well how you can use clever interaction design to present a liniar story and still make the player feel he's in charge. I might write some more about it soon. Stay tuned. :)

27 July, 2007 19:45  

Post a Comment

<< Home