Back to Community
Generally required level of complexity

I was just wondering on thoughts in the community around the level of complexity one would expect to be required to make a successful strategy. Perhaps anything that doesn't use highly sophisticated AI is already arb'd out? Thanks!

4 responses

The Quantopian platform doesn’t support sophisticated AI (although users may be applying it offline and uploading the signals). Personally I won about $1K in the contest without anything fancy so I’d just give it a go, just for yucks. It can be a nice little hobby.

To quote Einstein, " Everything should be made as simple as possible, but not simpler. "

I wouldn't doom your quest before it's even started by assuming "anything that doesn't use highly sophisticated AI is already arb'd out". Start with a hypothesis (like 'stocks that have a large price jump will revert back to a norm') and build on it. The code may become quite involved as one adds filters (to eliminate outliers perhaps) or execution controls (when does one exit a position?) but the basic hypothesis is typically quite simple. One can (and often does) combine several hypotheses together which again adds to the code. But again, the underlying ideas are simple.

AI is just a tool. Figure out what you want to build then choose one's tools accordingly. Many wonders of the world have been built with simple tools (consider the pyramids).

TL;DR: Was going to make a post saying people should avoid overfitting. Instead composed a philosophical commentary on society. Oops.

 

I would be more inclined to say that successful trading has a required level of simplicity.

Regarding: "Perhaps anything that doesn't use highly sophisticated AI is already arb'd out?"
That sentiment is absolutely a fallacy, although, I, too, was raised to view the world that way, and indeed I used to believe that.

AKA "Surely macroscopic science has been solved. There's no way I'll discover anything new without a billion-dollar particle accelerator." You know why particle accelerators were originally built? Particle research (that was much cheaper) was already underway studying cosmic particles entering the atmosphere, but among other things, they didn't like waiting for the random particles, so they built accelerators to speed up the research and control it. You could still do much particle research without them.

There is however some truth to the fact that thinking "commonly" - as opposed to with "simplicity" - can be an issue. If you think the same thoughts, do the same things, and use the same tools as everybody else, then no, you'd be hard-pressed to find an edge unless you have much more expensive versions of said tools.

Now, the market is complicated, don't get me wrong. And you won't ever be able to even predict 1% of its movement (someone would be able to take over the world if they could). But if the market were "efficient", then XIV/SVXY wouldn't accrue billions (80% was retail-invested/unspecified, https://www.reuters.com/article/us-credit-suisse-gp-notes/credit-suisse-volatility-fund-liquidated-after-market-selloff-idUSKBN1FQ256?il=0) full of buy-and-hold assets. Market seasonality wouldn't still exist even though it's been consistent going back 400 years (https://www.google.com/amp/s/seekingalpha.com/amp/article/1183461-seasonal-patterns-in-stock-markets-319-years-of-evidence). People wouldn't allow themselves to be led into bubbles created by central banks for hundreds of years (https://en.wikipedia.org/wiki/South_Sea_Company). People would remember how it all ended when fiat currency was invented (at least) 994 years ago; how their iron and copper savings were taken, and they were tricked into thinking it was worth something for 232 years until the jig was up and it all went bust (https://en.m.wikipedia.org/wiki/Jiaozi_(currency)).

 

As Credit Suisse so poignantly writes on page 197 of XIV's prospectus: "The long term expected value of your ETNs is zero. If you hold your ETNs as a long-term investment, it is likely you will lose all or a substantial portion of your investment."

AIs can't read prospectuses.... not yet...

@Ka A - nice commentary on markets and society!