sample/whole source final bitrate with crf

[martinfrombern]

I have a source with ~90000 frames, and created a sample with with ~5000 frames (using selectrangeevery(...,75)). i did crf 21.5 encode of the sample and got final bitrate 1850 kbps. same crf encode of the whole source resulted in bitrate 1600 kbps.
my question: is this normal that there is such (relatively) huge difference between bitrates of these two encodes? Somehow I believe that 5% of the source should be a good indicator of its compressibility, so this difference is a bit of a shock. Is there any other logical way to explain this, except the obvious one (sample didn't represent the whole source good enough)?

[Audionut]

Say your source has 60% of high motion scenes. Your 5% sample might only catch 30% of those high motion scenes.

[Atak_Snajpera]

compressibility test is waste of time! You are still living in Xvid era. Why do you need this?

[martinfrombern]

compressibility test is waste of time! You are still living in Xvid era. Why do you need this?because i wanna know what would be (approximately) final bitrate using crf.
and no - i'm not targeting any specific bitrate (not even within a specific range), i just wanna know what would it be.

[cogman]

Scene changes are expensive operations for encoding. By using select every, you basically ensure that you capture pretty much every scene change in the video (and possibly create some artificially).

I think what you are seeing is the heavy weight of your scene change, when you have the rest of the video, that weight sinks into the background.

[Dark Shikari]

Scene changes are expensive operations for encoding. By using select every, you basically ensure that you capture pretty much every scene change in the video (and possibly create some artificially).If you select groups of keyint frames, you won't have that problem.