"mean" is far more useful to estimate how many transactions are likely to fit in a block.
Nobody said that it wasn't.
translation
"Q:guys how do we hide the fact that transaction sizes will bloat when people do calculations of the average blocksize vs average transactions per block after all the proposed features are included..?
"A:dont talk about averages, dont use 'mean', we can manipulate numbskull opinion by talking as if we are suggesting average but actually quote a median number.
"Q:how does that work
"A: well if we had 0,1,2,226,227,228,229 the median is 226.. if we have 0,0,0,226,5000,10000,500000 the median is still 226... if we have 0,226,1023435453 the median is still 226
"Q:so why should we pick 226 as a special number..
"A:because that is a safe minimum transaction size, its not the absolute minimum, but its a safe minimum people expect to see.. and if we try to talk about this minimum in a way that makes people presume 226 is expected atleast 50% of the time. or the majority of the time.. we dont have to explain real data because then it is revealed that us blockstreamers are actually the "bigblockers".. where we offer less transactions per megabyte then the simple blocksize increase alone"
Yup, "I don't see blocks with 4000+ transactions as the median 226 would imply."
But, as pointed out it does appear to be the median.
I have asked Greg if he knows the average transaction size as that figure would indeed be more relevant here.