Happy Trails, Part Deux

0 Comments | This entry was posted on Nov 12 2009 by

Today we eliminated the shrinking trails problem by moving all of the trail particles and related calculations to the GPU.  Long trails are back again.

Happy Trails

0 Comments | This entry was posted on Oct 13 2009 by

So we were getting some performance issues from the boid trails.  It turns out that updating, drawing, and tossing one trail particle per boid per frame into the garbage collector is inefficient at 150+ boids.  Who knew?

Anyway, we adjusted the system to reduce the number of trail particles as the processor began to choke.  This leads to the trails lengthening and shortening, gradually, during gameplay.  As having a bunch of one kind of trail is cheaper than many different types of trail, this means that as you win, and more of the boids become your type, the trails get longer.  It’s totally an intentional reward system.


0 Comments | This entry was posted on Oct 02 2009 by

Sometimes, at the very beginning of a game, you run straight at an enemy an die, and that game is a waste.  Not anymore!  With the new respawn option, a neutral boid simply becomes a new you!


0 Comments | This entry was posted on Sep 18 2009 by

Each boid type has different attributes.  One is fast.  Another has a smaller alignment radius, which means it’s easier to extract from flocks.  In fact, we have ten different stats that we can manipulate to change the way the boids move, interact, and react to the mouse.  Balancing the boids, so that players can choose the one they like the best, without having any one oid be clearly superior to any other, looks like a herculean task.  Until today.

Today, I created a test program, where we can set the boids’ stats to whatever values we want, and then have the AI play a bunch of games against itself really fast (yes, just like in war games).  If, after 10000 games, the boid with speed 3 and agility 2 won more than its fair share, we know that it’s unbalanced, and can, say, reduce its speed to 2.5.

There are some potential shortcomings to this approach, the biggest of which is that the AI plays somewhat differently from humans, and so boids that are balanced for AI battles might not be for people.  Also, it takes quite a while to play 10000 games, even without actually displaying the graphics on the screen, so each test we run basically has to run overnight.  Hopefully, though, this system will enable the multiplayer to be fairly fair, with people able to choose whatever boid type they fancy without being handicapped, right out of the gate.