Michiel van de Panne from the University of British Columbia graciously took the time to answer my questions regarding his work on Ski Stunt Simulator, perhaps my all-time favorite physics game. His gives his thoughts on academia/game industry collaboration, his varied success with commercializing Ski Stunt Simulator, low-level physical play control, his development process, and more.
FM: First of all, thanks for taking the time to do this interview! I’m sure your job at the University of British Columbia must be very demanding. It’s great to have a chance to talk with you.
MvdP: Sure, you’re welcome.
FM: Ski Stunt Simulator was very popular when it was first released. Did you expect the game–especially the Java version–to get as much traffic as it did? Did you consider sales of the standalone version a financial success?
MvdP: The java version was intended as a way to promote the PC version of the game. It began with an experiment to see how fast the inner simulation loop would run in Java. Cedric [Lee] decided it would be easier to port most of the system over to Java rather than just the inner loop needed for timing tests. In about two weeks we had a functional Java version of the game. Some aspects, such as the audio and scripting structure, took longer to port. It fairly quickly became obvious that the Java version would work quite well. The traffic it attracted did surprise us — after about 6 months, the java version was being played for an average of 300,000 runs per day. The standalone PC shareware version had many downloads but only about 500 copies purchased. So this was not a success for a number of reasons. We didn’t spend enough effort on addressing piracy issues — a few individuals began to publish registration codes and cracked versions of the game. There are further hidden costs involving in dealing with credit card fraud. And in the end, I was interested in returning to more forward-looking research rather than the daily minutiae of running a business, although academia involves plenty of minutiae as well.
FM: Ski Stunt Simulator appears to have been conceived as an interactive simulation/game. A lot of your other work is more theoretical in nature without the end goal of creating an application like Ski Stunt. Which do you enjoy more–research for the sake of research or working towards a focused goal by creating a playable simulation like Ski Stunt?
MvdP: They can both involve significant creativity, which is the real attraction. One of my core interests remains control — how do humans and animals control our muscles (all 150-600, depending on how you count) in order to walk, ski, etc. ? But with ski-stunt simulator, the goal was in part to show that a lot could be done to go beyond the current game-pad interfaces with buttons to ‘do-a-flip’. Many real sports such as snowboarding, skiing, or skateboarding involve a mix of creativity and skill. The creativity is in asking the question ‘is it possible to do a half twist followed by XYZ’. The answer to this then depends on both physics and the skill of the person. Ski Stunt Simulator thus tries to embody this, which is very different from discovering hidden button combinations in traditional sports games. But what I can do with my research work in the Department of Computer Science at UBC is then ask “How do humans and animals plan and execute their motions? How can we employ machine learning to have robots ‘learn’ about balance? How can we develop 3D modeling tools that `know’ how to interpret a hand drawing of an object or building?
FM: Cedric Lee from Relic Entertainment has credits in the paper accompanying the game. Was this your first partnership with someone from the video game industry? Do you find that the game industry is often interested in your work? What are your general thoughts on academia’s involvement with the industry?
MvdP: Cedric did a fantastic job working for Motion Playground and later moved on to work with Relic Entertainment. With respect to more general game-industry / academia: there is some interaction at conference — some academics go to GDC and some from the game industry go to conferences such as SIGGRAPH. But much of the transfer of ideas and algorithms happens through graduate students who choose to work in the game industry. More recently, there are more examples of game industry employees going back to university to get Masters and PhD degrees. And of course academia can learn about specific problems that are currently being faced in industry. A big difference is that of timeframe. At UBC, I can focus on tackling longer term problems what won’t pay off for 5 or 10 years. Deeper collaborations may be possible, but are burdened by intellectual property issues, etc. So, it’s a challenge. About 3-4 years ago I gave talks about ski-stunt simulator at a number of game companies, but nothing came of it. The standard model used by larger game companies seems to be that of purchasing companies that have already successfully proved particular ideas in the marketplace.
FM: The controls feel very well-tuned. How different is the final version from your initial implementation? What was your process for tuning the controls and the articulated figure’s constraints?
MvdP: The interface took a bit of tuning, but this was not excessively difficult. More time went into thinking about classes of motion where a character could be easily controlled using a continuous two-dimensional input space. But it is tuned for mouse or tablets rather than the standard joystick polar coordinates (direction and magnitude). We did experiment with controlling a physics-based swimming fish that the user could make swim, turn, and even jump out of the water to do a flip. But the ski jumps clearly had the most appeal.
FM: Players in Ski Stunt Simulator must rely on “chunking” component motions into higher-level verbs such as “front flip” rather than simply pressing a button to perform a pre-recorded action. Do you think mainstream retail games will be able to evolve from simple button presses to more advanced control schemes like Ski Stunt Simulator’s?
MvdP: That’s a good question. I’m hopeful, but it will require an incremental path such that interfaces other than the standard game pad become more widely adopted. I’m the most optimistic about future camera-based interfaces because of their low cost. However, the tracking will need to be more sophisticated than the Sony Eye Toy of a few years back. Another incremental path is for the game to offer a separate "advanced mode" that uses the physics and a low-level control interface. Thus it becomes an add-on rather than replacement for the easier-to-control standard interfaces. The attraction with each style of interface is very different.
FM: Did you have a lot of other people play the game early in its development for feedback? Or did you rely more on your own instincts for when things just felt “right”?
MvdP: Really just our own instincts.
FM: What would you change or do differently if you were to do the project again today? What did you learn from Ski Stunt that helped you in subsequent projects?
MvdP: Developing a version playable on consoles would probably be important. Also perhaps a payment model for the java version that lets one buy batches of runs and then use password-access to `spend’ these runs. At one cent per run, the java version could have a sufficiently deep self-sustaining source of revenue to continue development.
FM: The paper mentions you also experimented with using articulated figures as a platform driver and a person on a swing. What was the inspiration to use skiing as the context for a physics-based character simulation?
MvdP: I’m a skier myself, and it was one of the best fits in that an unsuccessful attempt was entertaining all by itself.
FM: The custom stunt creator in the game is great, but unfortunately there’s no way to save your creation to share with others. I know a lot of people would greatly enjoy being able to create new content for the game. Have you thought about updating Ski Stunt Simulator to be more extensible, or even releasing the entire project as open source?
MvdP: Right, many many people have inquired about this. If development were still ongoing, this would clearly be at or close-to the top of the list of things to add. I haven’t decided yet about releasing the source code.
FM: Have you played a lot of other physics-based games? Do you have a favorite?
MvdP: I’ve played the knock-the-man down the stairs game, but between my family & work, life is just too darned busy.
FM: Would you like to do another playable game project like Ski Stunt Simulator again in the future? What happened with Motion Playground?
MvdP: I’m winding down Motion Playground. And I’ll perhaps advise on some game projects if the opportunity comes up. But I’m focused on research at present, looking at ways in which advancing knowledge in machine learning, perception, and interfaces can help simplify content creation for computer graphics.
FM: Thank you again for your time!
Subscribe to comments with RSS or TrackBack to 'Michiel van de Panne, Ski Stunt Simulator'.