My case is accidentally I developed two completely independent systems with some positive correlation between them. They were based on very different assumptions/approaches.
When comparing their performance, I found they both work quite well, but one has relatively better performance than...