Some useful definitions of neural networks in forex and quant

(Last Updated On: September 17, 2010)

Some useful definitions of neural networks in forex and quant
This was a discussion found on Linked In:

OK, some definition type discussion:

A) Useful:

1) Does the approach produce profits greater than the prior approach that was in use before the current method was implemented? If so it is more useful than the prior method.

2) Does the approach allow better operations? If one can follow and trade more market instruments than the prior method then that is useful as returns are maximized when capital is usefully invested 100% of the time as opposed to sitting in the account waiting for the next trade.

3) Does it support totally automated operation? If so this is useful as it eliminates human intervention and allows trading 24/7

These are a few things I’d call “useful”. What do you define as useful?

B) Real money usage

1) Everyone needs to speak for themselves. Mine have. I assume yours have as well.

C) Long term solutions

Could you define long term for us? I don’t know what that means.

My approach is to remodel based on a rolling window as new market data is created. So the solution is in the process as opposed to some fixed “long term” solution if that is what you meant.

To me long term is predicting market activity more than 12 bars into the future. The best performance is in the 3 to 6 bar range. Bars can be 1 min 1 hour, 1 day. I’ve never tried it but it’s likely the tone could model weekly, monthly or annual bars as well.

D) Feedback Loop.

Could you define a bit? Are you saying that because markets react to what just happened not much can be predicted? I say it this way: Not everything can be predicted, which is fine. One doesn’t need to be in a given market instrument 100% of the time. The useful task is when a prediction is made it is has a high level of probability of being true….what is the Profit Factor on the predictions made and trades taken.

There are consistent elements in the market that if found do lead to stable long predictive performance. By that I mean little variation in performance metrics for over 33% of the time window used in the training set. These are difficult to find and a simple solution is just to remodel fairly often.

E) You said “If my way, wouldn’t it be nice IF someone had experience USING their mathematical theories”

I guess you assume that you and I are the only people trading on this thread? Possible I guess but unlikely

F) You said “They could then describe broadly how their methods fit into the broad scheme of things (as traders).

Could you define “broad scheme of things”? I don’t know what you mean beyond:

1) One has some capital
2) One has a method to decide when and how to invest that capital
3) One does so and one makes money or looses it.
4) How slow or fast does this happen? (Net ROC)
As to “useful”, I would posit only one definition matters (making more money). Even the purpose of risk reduction is driven by the desire to NOT lose money.

As to “real” experience, I’m not intending to be “nosey” just practical. Unless and until one knows what kind of TOTAL yield one is getting, one can’t determine how much of that TOTAL yield to ascribe to a particular technique being used.

As to “long term”, I would answer by citing examples from history – Nadoff, Enron, 911, Gulf spill, Obama’s election. No fixed term is “safe” – long or short. That’s why the slogan for my trading is “IT (x) HAPPENS”. I cleaned up my old farming inclinations.

As to “feedback loop”, depending on what time frame you choose, one has to be agnostic as to causes and effects and the stability of such relationships. I am reminded of the definition of “assUme”. I am speaking from a stock trading (company level) point of view (micro-economic rather than macroeconomic). Macro relationships may TEND to be more long lived.

As to “trading experience”, I hope and would assume more than just you and I have trading experience. In my experience, experience is the “fire” that forges the trading “steel” one needs.

I would still classify myself as no more than a (67 year old) student. I’m always willing to learn how to improve my trading and make more money. Guess I’m shameless (even though society chooses to classify me an “old fart”). By the way, I plan to live till 120 or later depending on how much genetic engineering they come up with to replace or grow parts.

As to “broad scheme of things”, what I meant was: MY trading is based on a modular analytical approach which I call EMSCAN (kind of stole from CANSLIM). EMSCAN stands for Economic, Market, Sector, Company Analytical Nexus/Nisus. EMSCAN (as an analytical tool) then functions within (is used by) a broader stock picking and trading system (SPATS). SPATS job is to manage the trading process, make and establish broad policy (and portfolio) decisions, and report on results. As such, I was interested in which aspect of these processes NN (etc.) is being used AND with what the marginal results (increases in profits).

NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!
Don't miss out!

You will received instantly the download links.

Invalid email address
Give it a try. You can unsubscribe at any time.


Check NEW site on stock forex and ETF analysis and automation

Scroll to Top