Unpredictable
A mystical religious sect
It took years for me to realize that certain areas of mathematical statistics are essentially no different in organization and nature that those who “scientifically study” Tibetan Tantras or similar subjects.
The whole fields consist mostly on repeating that someone said or published before them without any connection to actual reality whatsoever. They even proudly asknowledge that even if unapplicable, their theories are “sound”.
They just continue pile up abstract bullshit about abstract entities from their books - exactly what Tantric “scholars” do.
Reality checks, however, show that not a single sophisticated million-dollar cost statistical model outperformed the Market, and the profits institutions post are due to manipulations and primitive risky bets.
Like any good Tantric doctrine, so called risk-management in speculative financial and especially crypto markets is just highly sophisticated sectarian bullshit full of abstract (which captures nothing about reality) integral and differential equations, deeply nested.
Any scientific experimental testing would inevitably fail to validate any probabilistic model which is supposed to be able to predict any future outcome in the Market, but high social status of a priest of advanced statistics binds a sect together.
Principles
There are some notions from the old-school AI which tried to model systems.
A system is fully observable iff all the factors which affect it are known and could be defined mathematically.
Any dice or card or a game with a discrete set of fully deterministic rules is such a system.
Any physical system, subject to the finite set of “forces” and “laws” is also fully deterministic, given that any change in the environment will yield a proportional change is its observable behavior.
As long as a system is fully observable and fully deterministic (ether discrete or continuous) and mathematically modeled as such, there is well-understood frequentionist approach to observe and describe it.
A system being fully deterministic does not imply that it does not fluctuate or have pertubations within. A mountain river changes its flow daily, depending on time of the day, rains and seasons, but remains fully observable and fully deterministic.
Markets are nothing like that.
A Market is a dynamic system, defined as follows:
- The factors that affect its observable behavior are NOT fully observable or even could be enumerated.
- New (unseen before) factors continuously evolve and never remain stable..
- Observed (modeled) factors change its nature over time (do not stay the same).
- The “weights” of observed (modeled) factors change over time or even completely disappear.
The Market therefore is not a stable system. All the models form physics (which model an apriory stable system) are fundamentally wrong.
The ongoing activity of modeling them with probabilistic models is just (no different from) a religious sectarian activity performing sophisticated, vastly complex and costly rituals in minute details.
Notice that adequately modeling one or many aspects of it does not yield anything more than just one particular aspect.
Measuring “random” perturbations
While sounds reasonable and “logical”, permutations of a dynamic system yield only temporary approximations.
Even measuring cloud formations in a certain mountain are would produce a certain bounds about size, frequency and thinkness, exactly because there are a finite set of physical factors involved.
Everything is good and “reasonable” until an extremely rare cyclone from the sea reaches the area, and suddenly all the prior measurements and estimated probabilities got fucked (this is a technical term).
Notice, that these permutations are within some bounds, which are determined by the finite set of physical constraints on this particular planet (determined, in turn, by its size, mass, distance from the Sun). These are stable (for us and life itself).
In modern Markets there are no such stability, which implies that perturbations could be unbound.
The crucial principle is that, unlike physics, completely new factors, not limited to sums of exiting or observed ones, could be added just the very next minute. This fact invalidates any assumptions of stability at the right edge of a chart.
Not an opinion
It is not a matter of opinion, whether the Market actions can be consistently predicted. It is a conclusion based on grounded on the Causality Principle logic.
Religious sects, however, are rely for their very existence on having conflicting, clashing opinions, while not a single one could be proven correct. This is precisely what is going on.
Again, religious phenomena are well-understood and demonstrably reducible to supervised learning on socially conditioned inputs. There is literally noting but a supervised and then reinforcement learning.
Unpredictable
The fact that a large number of participants, including institutions and exchanges, with different anticipations and conflicting interests are simultaneously present in the Market makes it unpredictable.
This implies that we should not even look towards possibility of a consistent predictions, and focus on reactive side, just like every biological system has been evolved to do.
Biological systems, including us, rely on quick and dirty hard-wired heurisitcs, which reflect the constraints of stable physical environment, which looks like a universal stable strategy, evolved independently every time.
Yes, I would miss a $200k position of stats wizard at a Wall Street hedge fund (destined to inevitably collapse with a next major market crash), but the truth is dearest.
I will write a simple trading assistant which reacts adequately to What Is.
And yes, any unsupervised learning will fail in principle, because it requires a stable structure of underlying physical reality (learning just one or two aspects won’t do).