[rq1]

The Pi compression algorithm is better.

[luma]

Doubtful. The problem with the pi idea is that you need to include the offset, which will likely be as long as or longer than your data.