Machines, systems and methods for compression ratio estimation are provided. The method comprises selecting a plurality of sample points in a data stream to compress sample data selected at said sample points according to a first compression method; and computing an average compression ratio contribution over the selected plurality of sample points, wherein compression ratio contribution of a sample point is determined based on the first compression method and data in locality of the sample point, and wherein the locality for the sample point is defined by locations in the data stream, such that said locations potentially affect the sample point's contribution to the compression ratio.