The future effect of banks using AI software | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Metatrader using 40% of the CPU all the time when using EA 6 replies

Software to test EAs for MT4 w/o using a demo? 0 replies

Using MAM software question? 0 replies

Anyone using subscription software for trading signals? 1 reply

Anybody using PDA software? 1 reply

Attachments: The future effect of banks using AI software
Exit Attachments

The future effect of banks using AI software

  • #1
  • First Post: Jul 17, 2007 12:48am Jul 17, 2007 12:48am
Back in the day (2005) ElectricSavant asked me what effect I thought artificial intelligence would have on forex. At the time I didn't really have an opinion.

Today is a different story. I am completing a M.S. in Computer Science, and just read an article about Grid Computing. It appears that Morgan Stanley is creating a huge grid for sharing algorithms, etc.

The article really got me thinking. Part of me thinks the markets will never change, because they are dominated by

a) bank investements based strictly on fundamentals
b) stop hunting (at least according to the sticky in the broker section)

Who's to say that human traders on the desks won't be replaced by software. Afterall, it is just one big chess match. Is my software smarter than yours? The behavior of Bazian networks, neural networks, and Random Forest Trees is smarter than a puny human brain.

My biggest fear is that:

PRICE ACTION PATTERNS WILL BE ALTERED.

I am hoping that they are not, as my business model is based on them. ---
  • #2
  • Jul 17, 2007 2:37am Jul 17, 2007 2:37am
There are too many things for me to comment on, but I will say this...even if that did happen, and the "patterns" were altered, brand new patterns would emerge to take the place of the old ones and we would just get used to trading those.

Also, price action patterns, if they are based on underlying market fundamentals that cannot be altered, such as struggle between supply and demand, will remain.
  • #3
  • Jul 17, 2007 3:09am Jul 17, 2007 3:09am
well, not for now but maybe in another 5-10 years.

AI being feed with historic data, but current events is unpredictable.
AI still, cannot overtake human in FX market.
  • #4
  • Jul 17, 2007 5:10am Jul 17, 2007 5:10am
Agreed. There will always be simply too many discretionary variables in my opinion. I haven't seen a single "black box" hedge fund return anywhere near what a human trader/trading group has been able to consistently make. I'm not knocking technology, I love it, but I can't see a computer making its own profitable decisions consistently. I only see a human programming it. Also, I doubt the central banks are ever going to relegate their decision making to AI...
  • #5
  • Jul 17, 2007 5:14am Jul 17, 2007 5:14am
Quoting tdion
Who's to say that human traders on the desks won't be replaced by software. Afterall, it is just one big chess match.
You know, as a chess aficionado I was one of those who swore that a machine would never outsmart a chess master, but we all know how that went with Big Blue and Kasparov. So I do believe you've got a point and trading might follow the same way.

Scary, isnt it?
  • #6
  • Jul 17, 2007 6:33am Jul 17, 2007 6:33am
Quoting pfdtrader22
Agreed. There will always be simply too many discretionary variables in my opinion. I haven't seen a single "black box" hedge fund return anywhere near what a human trader/trading group has been able to consistently make. I'm not knocking technology, I love it, but I can't see a computer making its own profitable decisions consistently. I only see a human programming it. Also, I doubt the central banks are ever going to relegate their decision making to AI...
AFAIK, black box trading is the heart of Rennaissance Technologies's income, and they are one of the biggest (and most profitable) hedge funds around.
http://en.wikipedia.org/wiki/Renaissance_Technologies

Note however their work is based on rigourous math & statistical analysis, which is a somewhat different kettle of fish than AI and other soft computing techniques.

The major effect that math based systems seem to have already have had is the reduction of daily variance (volatility), as a good deal of that came from the time lags involved in hedging different instruments, or simply executing trades. Whether that's a good thing, or in the long run opens up a huge risk factor remains an open question.
  • #7
  • Jul 17, 2007 8:39am Jul 17, 2007 8:39am
Quoting WTB
You know, as a chess aficionado I was one of those who swore that a machine would never outsmart a chess master, but we all know how that went with Big Blue and Kasparov. So I do believe you've got a point and trading might follow the same way.

Scary, isnt it?
Chess isnt comparable to Forex. In chess you have a limited number of variables( 2 players, 32 pieces, a 8x8 cardboard and the game rules) where the AI can easily compute the millions of playing scenarios to overcome a human player. In Forex you have an indefinite amount of variables that makes any mechanical AI impossible to have any clear advantage over human traders.

I can only see two ways that an AI can be profitable.
1. An AI taking advantage of a market inefficiency that can be mechanically exploitable.
2. An AI that takes in consideration and can compute human variables (such as net positions of retail traders)
  • #8
  • Jul 17, 2007 9:03am Jul 17, 2007 9:03am
Quoting WTB
You know, as a chess aficionado I was one of those who swore that a machine would never outsmart a chess master, but we all know how that went with Big Blue and Kasparov. So I do believe you've got a point and trading might follow the same way.

Scary, isnt it?
Nope, you are dead wrong. An AI can NEVER outsmart a HUMAN trader in FOREX trading. There's a lot of variable in FOREX than in CHESS.

An US War ship sent to Gulf and US dollar down by 30 pips. Do you think AI can pick that up? Hell no.
  • #9
  • Jul 17, 2007 9:09am Jul 17, 2007 9:09am
How many variables can there be that are not computable? interest rates? oil prices? fundamental data (employment data, trade balance, etc)? technical levels of support/resistance? international affairs (wars, terrorism, etc)? obvious levels of stop loss placement?... all that can easily be computed. So why not?
  • #10
  • Jul 17, 2007 9:40am Jul 17, 2007 9:40am
Quoting WTB
How many variables can there be that are not computable? interest rates? oil prices? fundamental data (employment data, trade balance, etc)? technical levels of support/resistance? international affairs (wars, terrorism, etc)? obvious levels of stop loss placement?... all that can easily be computed. So why not?
Tons can go wrong. Good introductions to the problem are in Nassim Taleb's books "Fooled by Randomness" and "The Black Swan"; for wonderful examples of how mathematically perfect systems go wrong, try "A Demon of Our Own Design", by Richard Bookstaber. Both had more or less successful careers in finance, so they aren't just talking theory here.

A major problem is what Rumsfelt (love him or hate him) referred to as the "unknown unknowns" - those events that seem so predictable in hindsight but no one got them right in foresight.

On a more general level, markets and various data exist in a feedback environment; one of the best intro's to this is Soros' "The Alchemy of Finance" (in which he refers to it as "reflexive"). An environment like that can be subject to cascading events triggered by almost imperceptible shifts, which makes robust longer term prediction very dubious.
  • #11
  • Jul 17, 2007 9:42am Jul 17, 2007 9:42am
Re games - last time I checked, computer smarts were only good enough to play at the mid-level in more complex games, like Go. Chess is just too easy
  • #12
  • Jul 17, 2007 9:51am Jul 17, 2007 9:51am
Lol , debating the existence of a profitable AI is much like debating the existence of the Holy Grail.

For me it's far more easy to profit using my own brain than to develop such thing.
What I can only say is that the success rate among those Holy Grail and profitable AI seekers are ALMOST zero, specially for retail traders whom they lack the necessary resources, methodologies and skills to develop such thing.
  • #13
  • Edited 10:35am Jul 17, 2007 10:19am | Edited 10:35am
Trading is a business..

until computers can efficiently operate and run entire enterprises on their own, I do not believe computers can do the same in any financial market.

I think there will always be some sort of necessary relationship between machine and man when it comes to successful investing and trading.
  • #14
  • Jul 17, 2007 10:33am Jul 17, 2007 10:33am
Long Term Capital Management?
Patience + Humility + Study = Success
  • #15
  • Jul 17, 2007 10:35am Jul 17, 2007 10:35am
Quoting DutchTrader
Long Term Capital Management?
huh?
  • #16
  • Jul 17, 2007 10:45am Jul 17, 2007 10:45am
Quoting DutchTrader
Long Term Capital Management?
Excellent example of where math goes wrong (BTW, an excellent postmortem of LTCM's approach in Bookstader's book).
  • #17
  • Jul 17, 2007 11:01am Jul 17, 2007 11:01am
Quoting tdion
Back in the day (2005) ElectricSavant asked me what effect I thought artificial intelligence would have on forex. At the time I didn't really have an opinion.

Today is a different story. I am completing a M.S. in Computer Science, and just read an article about Grid Computing. It appears that Morgan Stanley is creating a huge grid for sharing algorithms, etc.

The article really got me thinking. Part of me thinks the markets will never change, because they are dominated by

a) bank investements based strictly on fundamentals
b) stop hunting (at least according to the sticky in the broker section)

Who's to say that human traders on the desks won't be replaced by software. Afterall, it is just one big chess match. Is my software smarter than yours? The behavior of Bazian networks, neural networks, and Random Forest Trees is smarter than a puny human brain.

My biggest fear is that:

PRICE ACTION PATTERNS WILL BE ALTERED.

I am hoping that they are not, as my business model is based on them. ---

Hoboy. Okay, considering the fact that all of the following could be complete and utter hogwash and nonsense, here is MY PERSONAL opinion:

1. I don't think banks trade on "fundamentals" per se.
My personal and firm belief. All they care about is "if we buy this, does it make interest for us? Good. Buy it!" (i.e. they want the interest differential on their side, that's it)
That's about as fundamental as I see them getting. The next part is - some of the jocks at the trading desks are living in tick-space, so they trade the price, not the news. If they do take the news into account(Trader888 mentions this), it's still going to show in how price moves, so in turn that becomes charts and price.

2. I'm not going to go into stop hunting since I have no real / true insight into this and haven't dealt with real "order books" / level 2 etc.

Now, to the case at hand.
Having programmed a bit myself, studied a semester of computer science and being knowledgable in the whole tech department a bit myself etc, etc, etc. I view things like this:
We already have automated trading systems as it is. They hook onto trending somethings and tag along e.g.
I've also got a theory of these systems using "price grids" which can look a bit odd on a chart but are imho a feasible theory.
These algorithms trading prices would imho work off charts and mathematical input, not fuzzy logic induced fundamentals or something.

Since a) this is already in place and happening for quite a while now and b) all this can do if more and more use it is EMPHASIZE moves, not diminish them (assuming that folks prefer having winning systems as opposed to losing ones), I am thinking this can only be beneficial.
The systems in turn will emit tradeable patterns or simply make old known patterns more exaggerated / visible and carry price one way or another.

I mean seriously, if you have an inside bar and the way to trade it is to straddle both sides and stay in as profit is made, how is anyone going to change THAT? By inventing something that whiplashes price around a bar for 50 weeks or what? And how would that be profitable for the system owner? Exaggerating here, but I hope you get my point.

Really, there's not going to be a change for the simple reason that price is price.
The algorithm has to have a buy/sell criteria simply put, and it will have to use a price in the market for that. It "sees" the same real world price as us.
Hence there will still be trendlines, s/r, breakouts, channels, formations and so forth.
They might be slightly altered, skewed etc, but aren't they now as well? Due to news, changing volume and so forth.

I don't see a reason why anyone would ever have a reason to panic.
In order for ANYONE to profit off price it has to move. All WE need to do is latch on. Close / exit in profit. Repeat.

Why be scared of something that can only hope to mimic us in the first place?
And if it supercedes us then welcome to Evolution, we just have to try harder, one way or the other

I'm not scared.
Trust price. Know yourself.
  • #18
  • Jul 17, 2007 11:13am Jul 17, 2007 11:13am
Quoting zamans98
Nope, you are dead wrong. An AI can NEVER outsmart a HUMAN trader in FOREX trading. There's a lot of variable in FOREX than in CHESS.

An US War ship sent to Gulf and US dollar down by 30 pips. Do you think AI can pick that up? Hell no.
You should check up on this article for example:
http://www.ft.com/cms/s/97d023d0-2ed...0779e2340.html
  • #19
  • Jul 17, 2007 11:17am Jul 17, 2007 11:17am
Quoting HalifaxCB
AFAIK, black box trading is the heart of Rennaissance Technologies's income, and they are one of the biggest (and most profitable) hedge funds around.
http://en.wikipedia.org/wiki/Renaissance_Technologies

Note however their work is based on rigourous math & statistical analysis, which is a somewhat different kettle of fish than AI and other soft computing techniques.

The major effect that math based systems seem to have already have had is the reduction of daily variance (volatility), as a good deal of that came from the time lags involved in hedging different instruments, or simply executing trades. Whether that's a good thing, or in the long run opens up a huge risk factor remains an open question.
I knew this was coming LOL! I was also waiting for someone to bring up DE Shaw who is managing even more assets (29bil at last read) than James Simons himself (6.4 - 9 bil medallion fund, however Simons returns vastly outstrip Shaw's). Simons is someone that I admire more than anyone in this industry, but when I compare between 15% (his 2006 return) and John Arnold's consistent 200% + returns for more than 5 consecutive years, (except for one which was 178%, but this year it was over 350%!) I just tend to think that AI by itself isn't going to outstrip our decision making. I also (check the sig) greatly admire Soros/Druckenmiller and I'd like to see AI make calls/decisions like some of those made by that duo. In any case, I love tech (I'll say it again!), but I just don't feel the AI argument to be realistic. Also, the chess comparison is about as far off as possible, even though I respect your thoughts (trading and otherwise) WTB. You too Halifax. You guys always make me think
  • #20
  • Jul 17, 2007 11:17am Jul 17, 2007 11:17am
Quoting BlackMage
You should check up on this article for example:
http://www.ft.com/cms/s/97d023d0-2ed...0779e2340.html
How is that related to FX?
Patience + Humility + Study = Success
0 traders viewing now