Back in 2004 when I worked at a newly-founded hedge fund with the tech head of SAC capital's algo trading to replicate the profits they'd been making using statistical arbitrage models, I learned a very important lesson: If all you've got is conventional statistical methods, TAQ history and a good real-time market data feed, you're going to fail.

You either have "inside" sources at or near the traded entities or you have to spend an enormous amount of money to trade tiny price differences at milli-, micro- or nanoscale time intervals. And your models will have to wield stochastic computations at high speed. Now widely known as HFLL trading.

Of course, as everyone in the equity trading world now knows, there are only a handful of entities that can muster the talent and technology to consistently make money using only probability models. Or you incorporate things like sentiment from chat groups, news sources, other exogenous events that move markets.

Crypto is largely dominated by gamblers, so-called "technical traders". This pseudo-probabalistic practice is largely a religion and not a science. And this is what is described here.

The net state of making money on crypto is trying to guess better why anything is happening. Although, there are changes. For instance, after Shi Jin Ping put the kabosh on ICOs and local coin trading in China, there was a big movement of Chinese money into BTC, ETH, etc. In other words, there was an observation of a non-statistical event or condition. But this was one-off and the effect subsided.

The only means of "investing" --as opposed to informed gambling-- in crypto is to believe demand will keep going up (eg., in proportion to the amount of dark money and contraband/laundering exchanged) or to look for the rare events as China's change of policy.

Everything else is "technical". In other words, there is generally no such thing as investing in crypto. So, good luck making real money.