Testing Sentence Mining for Particle Nuance

I've been tracking my error rate with particles. I noticed a significant 恩恵 after switching from grammar rules to mining sentences from real content. In the 後半 of my latest reading session, I correctly identified 15/15 directional particles. Seeing them in context prevents the usual grammar confusion from being a 邪魔 to my flow. Has anyone else measured their accuracy shift like this?