Could AI / Machine learning come up with the method the way you have ?

[guru]

the question is legitimate since technology advances fast. even if AI has not totally replaced manual traders, it has already replaced lots of market makers or traders on sell side. some hedge funds have already employed AI to make money. if more hedge firms can make money using AI, more people on the other side will lose, and professional traders could be one of them. how many professional traders before internet era has faded out when online trading becomes the mainstream?

It could be turned into a legitimate professional discussion, but it seemed to be posed out of fear. Though OP also talks about “patterns” and “factors” that are biased, opinion-driven, not published in research papers and therefore not even related to AI.
No one currently works on AI or technology to eliminate imagined non-existent edges that happen only to humans who constantly invent new “edges” after realizing they didn’t have an edge in the first place

 

[schizo]

It could be turned into a legitimate professional discussion, but it seemed to be posed out of fear. Though OP also talks about “patterns” and “factors” that are biased, opinion-driven, not published in research papers and therefore not even related to AI.
No one currently works on AI or technology to eliminate imagined non-existent edges that happen only to humans who constantly invent new “edges” after realizing they didn’t have an edge in the first place

I think a lot of people are mistaken in their belief that the machines will abide by the same logic as their human counterpart. When AlphaGo went up against Lee Sedol, everyone thought Lee would prevail. Even if AlphaGo wins, it would be a close call. Why? Because they all thought the machine would play in the same manner as humans have done for ages. However, as we all know, that's not what happened. The moves were unconventional, never-heard-of, and even unthinkable--simply downright "illogical". And yet, in the end, Lee lost by 4-to-1. It's safe to conclude that the machines will not play in the same league, or employ the same edge, as you or me.

 

[Specterx]

I think a lot of people are mistaken in their belief that the machines will abide by the same logic as their human counterpart. When AlphaGo went up against Lee Sedol, everyone thought Lee would prevail. Even if AlphaGo wins, it would be a close call. Why? Because they all thought the machine would play in the same manner as humans have done for ages. However, as we all know, that's not what happened. The moves were unconventional, never-heard-of, and even unthinkable--simply downright "illogical". And yet, in the end, Lee lost by 4-to-1. It's safe to conclude that the machines will not play in the same league, or employ the same edge, as you or me.

The financial markets are in no way comparable to fixed-rules, closed-system games. AlphaGo and similar systems work by, essentially, playing millions of games against themselves to "learn" the optimal path from any given game-state. This type of approach doesn't work in the financial markets because they aren't a closed system adhering to known, fixed rules. Applying AI techniques will probably shorten the life-cycle of profitable trading algos, but that's already a realm where retail traders can't hope to compete.

Retail trading will be killed off by a combination of other threats (FTTs, increasingly privatized markets, vol suppression by CBs etc) long before AI becomes a major factor.

 

[zdave83]

At some point with 100's or 1000's of well-funded, individually tuned, platforms competing against each other, it will be hard for any single one to dominate as their newly created 'edges' are discovered and absorbed. As long as markets continue to go up and down, successful traders are likely to be successful, and unsuccessful traders are likely to be unsuccessful.

 

[schizo]

The financial markets are in no way comparable to fixed-rules, closed-system games. AlphaGo and similar systems work by, essentially, playing millions of games against themselves to "learn" the optimal path from any given game-state. This type of approach doesn't work in the financial markets because they aren't a closed system adhering to known, fixed rules. Applying AI techniques will probably shorten the life-cycle of profitable trading algos, but that's already a realm where retail traders can't hope to compete.

Retail trading will be killed off by a combination of other threats (FTTs, increasingly privatized markets, vol suppression by CBs etc) long before AI becomes a major factor.

I believe the pit traders said the same thing about electronic trading before they were ultimately swept away in the dustbin of history. For too long, they brushed off electronic trading as unimportant, reiterating the fact that they have exclusive edge on trading, for no other reason than that (when you're behind a monitor) you can't see other traders' facial gestures or that you can't hear the squawk in the pit, blah blah blah. With the pace of technological changes picking up ever faster by the day, it's only a matter of time before manual traders will be displaced by the machine, just like any other manual laborers in different industries.

 

[dtrader98]

It comes down to whether or not rules can be coded. ML can replicate pretty much anything you want, given that the input/output decision/response space can be mapped to a data set, but unfortunately, not in a plug and play manner.

I still try to understand people who adamantly believe that it is impossible to backtest anything and historical data is useless. Clearly, they themselves must be making decisions based on things they've observed in the past. Even if they can not code it. Something like a squawk box, only can't be coded, if one lacks the raw data and knowledge to code it.

I agree with others, that things like GO, do not offer much in the way of beating markets. They work well in closed rules based systems with stable repeatable outcomes.

The same way that traders spend hours developing systems, AI developers too have to spend hours (if not years) mastering ML/AI, then figuring out how to apply it to developing systems. Once, again, the plug and play idea is just not there yet.

 

[guru]

I think it all goes both ways. I'm actually working on a system that is supposed to replace manual trading, so I'm trying not to be left behind and possibly even have such system compete with human traders. But there may be a million algos already out there, and I'm not sure if they're a threat to traders, or possibly those algos need traders so that actual trading activity happens in more “human like” form. Once that human form of trading disappears then both humans and machines may fail, left only with intricate HFT-like trading.
The question then becomes, how would've trading looked like if it was always done by computers while humans never traded in the first place.