Tweaking some more...

Looking over the risk system I put into it, it does not seem to take into account enemy 'nearness' at all in the risk evaluations. Meaning it doesn't really take into consideration the distance to the enemy when determining which way to go. In fact it doesn't take this into consideration period. It doesn't even have a catch to avoid ram bots.

The only close avoidance method it has is its minimum risk movement when there isn't any waves (which when you are very close is a percent of the time), and its natural tendency to try and spiral away from the enemy.

It only cares about the waves. Which is fine, and works great obviously (considering the rankings). It just seems to be really good at dodging bullets™. That aside, I am wondering if I should try to hammer some enemy distance risk into it. Considering how potent the movement is already, tweaking anything could really really break it, especially something as key as the risk evaluation.

I (and I believe Voidious) had extremely good experiences with incorporating some sort of way that weights closer points to the enemy as more dangerous. I multiply by botwidth, which effectively divides by distance. I then divide by the min(distance_to_wave_centre,distance_to_enemy) from the test point. I also then weight by how long it will take the bullet to arrive at that point.

Hey, thanks for the advice. I added a really tiny bias to avoid being to close to the minimum of those two things and it is looking really good so far. I hope it can keep this ranking with full pairings.