Ok, taking my seat again for the next round incoming.
Hey tok _ n
I was thinking about our recent interaction, and I am really thinking that if you were to consider a kind of weighted average to the trade volumes of exchanges, then that could be more satisfactory overall...
I've thought about this in the past too. I tried to model it as a supply/demand system.
I failed miserably, but the journey was enlightening
I understand that there has been various discussion about this point of weighted trade volume, and I understand that it could be quite a bit of work to attempt to establish any kind of exact system, and therefore, probably good in concept, even though in practice we may want to merely attempt a kind of ballparking of such weighted trade volume considerations.
1) there is some actual faked volume easily visible watching trades, verified by people who should know etc. This makes modelling (and therefore weighting) very inexact.
In the end, each of us is working with various kinds of imperfect information; however, yes, as you say, from time to time, some insiders will reveal some of the insider practices (such as faking trade volume), and various kinds of corruption and deception such as this is going to vary from exchange to exchange (and some will be attempting to be on the up and up), so we just would need to ball park approximate with the best that we can.
2) as in all demand/supply models you either take into account some external resource limitations, or you accept an infinite demand when the price is zero. This means the zero transaction fee volumes are limited by unknowable things like traders' bandwidth, bot speeds, exchange processing limitations, etc. Weighting this is pretty much unknown (and even if you could, it wouldn't last long).
I understand that these kinds of factors can exist, sometimes, and sometimes these kinds of factors may need to be considered in various scenarios, but other times, maybe they constitute too many trees for missing the forest.... I mean big things and big lumping of factors is likely better than getting caught too much into the details, in my thinking.
3) where the fee is above some threshold, it starts to become doable. I was looking at something like measured volume = a 'base' volume scaled up by some measure of the jiggling around of the price by traders with the start and end price over an analysis period (volume bar times). That base volume was to be just the volume 'swept' out under the effective supply/demand curves at the start of the analysis as the price moved (no retracements) from the starting price p0 to the end price p1 in the analysis period. I finally realized that the offered volume would, itself, probably be dependent on the fees (?).
The scaling up would need to be done by adding randomness, restricting possible transactions to those that actually made a profit greater than their fee plus maybe a factor expressing the drop in demand for transactions simply because of the fee (a trader sentiment kind of thing). The effective demand/supply profile would probably need to be approximated by the exchange order book.
I gave up looking at all the bad approximations needed in that lot. Lol.
TL;DR zero fee volumes can't be used sensibly unless you model those fake and external influences. Different fee volumes probably can't be modelled the way I described but maybe the basic idea is usable.
I personally think that it would be easier not to get caught up too much with any particulars of the exchange, just down grade them a bit from 100% if they have questionable practices. For example a place like Okcoin or Huobi have extraordinary trade volume, so maybe just count them as 10% or 20%, and maybe even less, if you come to the conclusion that the trade volume does not really represent anything meaningful...