We keep running new generations of our evolutionary search, and more informative relationships emerge; non-mainstream search engines where demand signals begin to pop-up, also mining pools, and even informational sites. Several iterations can be run, and if we do it over larger sample sizes (instead of the 200 sources, we use the original 1000, or even 10.000) and more time periods (temporal steps measured in weekly or daily returns, instead of months), likely many other interesting relationships will appear.
Model 2
\((BTC_{returns})>(BCH_{returns})=\left(BTC2poolin-BTCAsksearch\right)>\left(BCH2poolin-BCHAsksearch\right)\)
Model 3
\((BTC_{returns})>(BCH_{returns})=(BTCcryptofacts−BTCbitcoincom)>(BCHcryptofacts−BCHbitcoincom)\)
This diversity points to an interesting fact: we should perhaps trust model ensembles, rather than standalone models. And this makes sense: while in classical Economic Complexity theory one deals with relatively unchanged basket of products that are produced by the same countries, in a crypto economy new sources of attention are born and die constantly, and the sinks of that attention (the economies of each network) also are created at any time that a fork occurs and a community rallies behind the new coin. Our best anti-fragile ranking system is that one that is flexible and robust — as the crypto-economies that are being analyzed themselves.
Conclusions
In essence, both Say’s and the ECI approach are about aggregation of dispersed resources, and that’s what makes those so relevant to the study of decentralized systems.