so it would seem, so it would seem. i was just pointing out that as an example from my own encoding tests. but let's consider something, on an encoding with these VBR settings:

minimum bitrate = 1000
average bitrate = 4000
maximum bitrate = 9000

with CCE running multiple passes, it "sees" which frames have a lot of motion, and hence, it allocates the more bits (so it is technically a higher bitrate for the specific high-motion scenese), and it does this more accurately than if it were a single pass (CBR). reason being: more passes means it gets a much more accurate idea of how many bits need to be allocated at what positions in the video.

not sure how familiar you are in the field of mathematics, but if you know anything about limits and parabolic equations, you can think of it this way... a video (in AVI format) is like a parabolic line, and when doing an encoding to MPEG-2, you're basically assigning square blocks to approximate (as closely as possible) how the source material works.

for a more visiual explanation, take a look at this graph on HowStuffWorks.com...

[img=http://img205.exs.cx/img205/9758/cdsample16yf.th.gif]

this graph was actually made to represent a comparison of analog audio (on vinyl, e.g.) vs. digital audio. as you can see, digital audio actually approximates the closest possible "bits" to the analog source, although it isn't a perfect replication.

likewise with encoding video to MPEG-2 (from a different source material, such as AVI format), the encoding software approximates the closest possible "bits" to the source material. now, the more the software can learn about the source video (AVI), the better it can approximate where and how many "bits" are necessary. i hope that makes sense.