Thread:Talk:Cunobelin/Danger function

The hitScans list and distance/danger weighting accumulation in getDanger() is very interesting. It looks like this is to avoid the sparse data problem caused by heavy segmentation which is exacerbated in the case of surfing by very few data points.

And I right in assuming this implementation is to achieve a similar effect to the kd-tree you would use in a mega bot?

In any case this gives me an idea, which like most of my ideas will probably come to nothing, but you can now feel guilty for making me waste another couple of weeks on robocode ;-)

I can't think of an easy/small way of doing KNN with this style of implementation, so I guess my distance weighting algorithm had better be a good one.