Project Card Race is another game prototype I developped while at Gameduell. It is still copyright Gameduell, and is an unreleased project.
This one was done using unity, and code in c#.
The goal of the project was to create an adaptation of RoboRally, but for two players, on phones. In order to achieve the same fun, I still had to heavily modify the rules. If players are still granted 5 cards, they can play from 0 to 4 cards on every turn. Players also have an extra action with a one turn cooldown, in this case a shot of glue, shot in the direction they are facing.
The goal of the game is still to pass through the three checkpoints in the right order, and be the first to reach the third checkpoint. To achieve that, each player would have a set of cards with a certain amount of moves, and up to them to achieve the best results.
I started by developping a paper prototype of this project that I tested with different people until I reached a point where the game was considered solid enough at its core. I then moved on to have a digital prototype. I decided to use Unity for its simplicity of use, and the possibility to deploy easily on all sorts of devices.
One of the concerns I had to address was that players would take too much time to make their decision. However, by setting the interface as shown on the screenshot, I could quickly run a series of tests on different players and see how long they needed to make their decision. The sweetspot when not confronted to an opponent, was between 7 and 9 seconds. The theory was that by leaving just enough time for median players to make their choices, we would avoid waiting, as players would concentrate on planning their moves first, then have barely more than 2 seconds of waiting.
A prototype of the digital version is available on my github. This prototype was created to showcase a first version of a basic AI for the game. Using an AI for first user tests would have permited us to put the players in more real conditions, without worrying about how to find an opponent for them.
Even though the AI is still a bit buggy, I think it works quite well. It was done by adapting Goal Oriented Action Planning to our needs. The AI tries to determine which is the best action to take in order to reach the first checkpoint. It weights the cards that it is given in that fashion and uses the one available with the highest score.