Post-GDC: Physical Gameplay in Half-Life 2

Author: by Daniel Sánchez-Crespo Dalmau

The Half-Life saga has been praised worldwide for its innovations in gameplay mechanics, from storytelling and physics to revitalizing the FPS genre. So, there was great expectation over Jay Stelly's talk on physical gameplay design in Half-Life 2 last Thursday at the Game Developers Conference. It was a session that examined Valve's approach to game mechanics design, and how physical interaction methods were devised for last year's blockbuster.

Stelly started by describing the initial prototypes his team started building and, more importantly, the methods used to validate them. With titles ranging from “glue gun” to “toilet crossing” or “skeet shooting”, these simple tests tried to experiment with physics concepts to create novel gameplay mechanics. After creating this series of prototypes, Valve realized they needed a scientific way to allow designers validate the suitability of each technique for the game. So, they developed the idea that a game design is just a set of experiences that:

So, creating good game designs actually boils down to improving the training of the player more efficiently and allowing him to prove his skills in more creative ways. From this initial assumption, Stelly moved on to analyzing the different types of training methods that can be applied to gameplay.

The first training method was allowing the player to learn by example: establishing some rules or mechanics by using a visual reference so he understands the new rules. A good implementation of this is showing the said activity in a game cinematic, so the player sees an NPC performing an action and learns that he can perform the same action. A second way would be to establish clues, and then allow the player to learn by deduction.

Scenarios where this may be interesting can be learning to use items encountered in the adventure: the player may find several pieces (like the gravity gun and the rotating saws in Half-Life 2) and figure out the way to use them together. A third, less subtle training method is using an explicit test, such as a clearly presented tutorial or sequence where an NPC tells us something, and then we are forced to pass a simple test to ensure we got it right.

These training methods, said Stelly, can be used to implement new game design constructs which are unfamiliar to the player and make sure they are integrated within his body of knowledge of the game. This was the case with the gameplay physics: very few games had done it before, and clearly none using the design decisions used by Half-Life 2, so a great effort was taken to ensure the player understood the new rule set.

Additional training methods included the currently popular sandbox mode, where the player has a virtual playground (in the logical sense, not the spatial sense of the word) to experiment with the new rules. In Half-Life 2 this was used in the first areas where you had access to the gravity gun: enemies were relatively weak, and there were lots of items to encourage testing the gun's performance.

Still, training the player can only be successful under some strict circumstances. The speaker explained how their testing showed that the learning rate diminished if the player was put under any type of pressure during training time or, even worse, exposed to any sort of peril or even combat situations. Clearly, stress makes us go back to basic survival strategies rather than trying new ones.

On top of that, putting the player in situations where he has to make decisions also seems to stifle learning significantly as he needs to focus in the decision-making process and not so much in acquiring new rules or info. To address these concerns, Valve made sure the player understood from the very beginning that it was okay to fail during training, and that no bad consequences would result: suggesting experimentation, as can be seen in the gravity gun tutorial you perform in-game with another NPC.

Speaking of the design of the physical gameplay, Valve begun by listing the constraints needed for the system. First of all, it should allow breakable objects, due to the heavy use of the crowbar during the game. Second, and perhaps most important, the physics engine should interact with the core combat mechanic, so you can use physics in the middle of a fight. Throwing objects at your enemies causes damage, as the collision speed inflicts a loss of hit points and thus creates a new, unique way of fighting.

A side effect of this is that both players and NPCs can use physics in combat for cover, which also creates new gameplay situations. Last, but not least, physics had to be integrated in such a way that they could interact with the core puzzle engine so new, physics-based puzzles, could be implemented. Many games have followed Valve's trail in this one in the last year or so.

At the end of the day, Valve created a blockbuster, but learned some valuable lessons about game design and physics along the way. They learned that integrating physics into a game is hard both from a technology and game design standpoint. Technologically speaking, they ran into numerical problems, some objects getting stuck, and even full physics simulator breakdowns. Still, design problems were at least up to par with these: training game designers to work in terms of torque, impulse and force was challenging, but decomposing each system in the game in physically correct blocks was even more challenging.

As a final piece of advice, Jay Stelly recommended turning your designers into “gameplay engineers”, so they use engineering methods to create, test and improve their gameplay constructs. In the end, this was a very complete session about one of the games that have recently raised the bar of what a first-person game can do.

_____________________________________________________