Pattern Based Strategy Design

[RedRat]

Quote from frostengine:

The indicator values are passed as raw values to a specialized NN that learns underlying patterns with those values that at the moment are being trained to predict if the price will be higher or lower 5 bars from now.

For example:

AddParam(RSI(3,1)[0],"RSI3 ");
AddParam(RSI(7,1)[0],"RSI7 ");
AddParam(RSI(14,1)[0],"RSI14 ");

Means I am adding the values for RSI(3), RSI(7), and RSI(14) as inputs.

Then this prediction can be used to build strategies around.

Ok thanks,

There is a misunderstanding at my side because of I thought patterns are something you can formalize. You cannot formalize NN, it is a black box, you don't know how it mixes the input parameters. I worked with NN a lot, I used different inputs, no RSI or other indicators, I dislike them.

 

[ammo]

Quote from frostengine:

I have struggled for years with developing profitable strategies.

have you used spx vs djt for an indication of turns,bottoms tops along with reactions to market profile s/r,couldnt program a vcr but i can watch these and find those answers

 

[frostengine]

I have just posted the DLL that runs this first pattern. This installation/setup is a work in process. Give it a shot and please provide feedback on how to make the setup easier and/or if it doesn't work for some reason.

<a href="http://thestrategictrader.com/dll-installation">DLL Installation</a>

 

[ssrrkk]

Quote from frostengine:

The first attempt at using this method can be viewed in this post: <a href="http://thestrategictrader.com/technical-analysis/trading-strategy/rsi-pattern-strategy/722">RSI Pattern</a>. There were over 220 statistically significant patterns generated.

You mention that you found 220 statistically significant patterns. How many have you tested to arrive at those? You do know that the more hypotheses you test, the p-value needs to be revised upward (less significant). The easiest is the bonferroni correction: simply multiply the p-value by n or divide your threshold by n = number of hypotheses. If you had tested 10,000 hypotheses, that means a p-value of 0.05 is no longer significant -- rather you need to use a threshold of 0.05/10000 or 0.000005. This is because the more you test, the more likely you will find a random profitable pattern by chance that will not be profitable in the future (i.e., is not mechanism based).

 

[jstox]

Quote from frostengine:

I have struggled for years with developing profitable strategies. Many times I have found what I thought was the "holy grail" only to crash back down after further testing/live trading. My recent threads being no different.

hi frostengine happy new year.

Ya ever wonder what the algo of the the trader is that can move the YM. Probably very simple. Send the market up 50, down 50, and go back where it started. Gunned the stops whenever they randomly turn on the machine.

Ya really think neural networks can figure out the programming statement random(x)? x is the time-of-day it decides to take your money.

Seriously, I'm arrogant but not arrogant enough to think the facial recognition and thumbprint guys haven't visited this market, the futures are drying up. I suspect a lot of system failures are taking place.

 

[frostengine]

Anyone try to use the DLL? Test the pattern against other instruments or time frames? Any interest at all?

 

[Wide Tailz]

Quote from frostengine:

Good suggestions. This current strategy has ONLY a time-stop. However, soon I will need to start building a more logical exit. Will explore the use of continuing a time stop.

Exits and position sizing are far more important than entry.

 

[Occam]

Quote from Wide Tailz:

Exits and position sizing are far more important than entry.

I diagree, if your objective is to outperform the market. If you're just churning your account for entertainment purposes or are diversifying as part of a retirement strategy, then maybe it makes sense, but this is a trading forum, so I'm going to assume otherwise.

“What’s nice about investing is you don’t have to swing at pitches. You can watch pitches come in one inch above or one inch below your navel and you don’t have to swing. No umpire is going to call you out. You can wait for the pitch you want.”

Wait for the pitch you want. The better the entry, the less significant the exit. As for sizing, just do something that makes sense -- don't blow up. It's really not that hard, if you've got positive-expected-value trades and don't lever too much (or don't lever at all, if you can reasonably avoid it).

 

[lindq]

Quote from frostengine:

Anyone try to use the DLL? Test the pattern against other instruments or time frames? Any interest at all?

I don't believe that you've addressed a very valid issue raised by abbatia earlier in this thread.

>>>Unless you understand why a certain pattern might make sense (and unless you understand this IN ADVANCE ideally), it is very dangerous to focus on the "n" pattens that work after discarding "N" (where N >> n) patterns that didn't. <<<

In my experience systems development is best begun by first asking YOURSELF...not your computer...what is the logic of what you are trying to accomplish?

If you believe that the movement of any instrument is truly random, then you haven't a chance in hell of making money, no matter how much computing power you throw at it. Random means random. Patterns are an illusion.

If, on the other hand, you believe that there is a human force of emotion and reason (or lack thereof), then you should first determine what that is. Then put your systems development to work on it.

It appears to me than in a real sense you are putting the cart before the horse. Run the stupid computer, don't let it run you.

Thus, if you were to first state a basic objective of your research founded on some sense of logic or observation...then share your testing results...well, that would be interesting. Otherwise it's just spitting out garbage from your software. IMHO.

 

[RedRat]

Quote from frostengine:

Anyone try to use the DLL? Test the pattern against other instruments or time frames? Any interest at all?

I do not run external DLL or EXEcutable files. Let us discuss the idea, not implementation, for implementation I use FANN library.

IMHO RSI indicator does not work at all, why do you use it as the input parameter for NN? Have you tried EMA?