May

27

As long as there are useful idiots in the world like those that I often enumerate from Cal, and the twin former Harvard fund manager, there is no need to worry about A.I in our field.

Orson Terrill writes: 

The idea out there that there is no discoverable alpha, or soon there will be no discoverable alpha, because of so many quant funds, hft, AI algos, have absorbed it all…. is ridiculous!

People should be excited that no one can get to it all.

The factorial reality when trying to get the faintest idea of the market as a closed system was already beyond the grasp of current human ability. Attempts to increase precision with AI doesn't remove anomalies from the market at an increasing rate, it will slow the rate of removal. Deep learning adds a huge amount of variables (random weights, nodes, whatever you want to call it) to the computational load.

Say you want every basket of 10 stocks you could make with the SP500, for 10 years of data, OHLC, volume, ranges of various frequency, day of week, day of month, month of year, 1st 2nd 3rd (and on) differences of everything, some various filters such as averages, then you want covariances, the changes in those.

THEN! (Maybe!) You take those 200 octillion data points (seriously) and get to work with some AI? Seriously? Every computer in the world probably couldn't do this, and of course you'd still have over fit models in the end.

To make the point about AI:

Formulating a model, or an hypothesis about a condition, with all the data wrangling and basic transformations involved, that are then used to make orders of magnitude more data by fitting something as basic as a neural network auto-regression model, essentially means the potential combination of inputs, and computational demand, has left the universe. It's way beyond the scale of any computational feat happening even the most bleeding edge. So. Far. Beyond.

In other words, think, dig, have ideas, test them, and don't worry that you might be the only one trading on something. If what you're trading on has just a minutia of complexity, it probably wasn't found via brute force search by some high performance computing juggernaut. Bet on that. They have trillions and trillions of the most simplistic price change anomalies that they are still chasing/racing each other for all over the world.

The stuff that the Chair was doing in the 70s (before I was even born!) is still being chased today…and everything else (which is almost everything) is left on the table. The structural reality of: What computers are good at, returns to speed, large funds needing to focus on the most liquid assets for risk control, and because large funds need larger sources of liquidity to match against, guarantees most money is crowding in the same large liquid spaces. 

Stef Estebiza writes: 

Warren Buffett's favorite holding period is forever.

anonymous writes:

That is his favorite holding period for everyone else, particularly those holding his, as it reduces the directional liquidity to the upside and therefore reduces the the power of the viscous forces to resist change and increase slippage to the upside.


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search