Pattern recognition... | Forex Factory

Options

Permalinks
Print Thread

Similar Threads

Pattern recognition 18 replies

Autochartist - pattern recognition software 1 reply

chart pattern recognition 6 replies

Point & Figure pattern recognition 1 reply

Pattern Recognition indicator - Question 0 replies

Pattern recognition...

  • #1
  • First Post: Edited 5:15pm Jan 7, 2009 12:50pm | Edited 5:15pm
I have a handful of patterns I've tested during trading that seem to indicate certain trading opportunities. Problem is, classical indicator analysis of them is hit or miss, which is just too unreliable.

So I'm wondering if anyone has any resources on building a pattern recognition algo.
  • #2
  • Edited 3:01pm Jan 7, 2009 1:08pm | Edited 3:01pm
I'm not sure what you want for sure but this indicator might be something you can modify.
Also you can look at these on this forum;
http://www.forexfactory.com/showpost...6&postcount=10

http://www.forexfactory.com/showthread.php?t=19143

I may have missed the whole point.

UPDATE:
Now I know I missed the whole point. Good Luck.
  • #3
  • Jan 7, 2009 1:22pm Jan 7, 2009 1:22pm
MIT has done some research in this subject (http://datacenter.mit.edu/ Your going to have to search for it) Basically it concluded as have all of my own studies that the markets are "White Noise" and there are mathematically significant patterns whatsoever.

The underlying assumption being that you are looking for patterns to make profitable trading decisions.

Most studies follow the scientific approach. Hypothesis > Theory > Law

Usually the hypothesis are based on some "observations/function" of a computer learning model, yielding millions of scenarios but I have not been able to find anything that has even made it to a plausible theory.

This explains why my systems based on Random number generators yield similar results to filtered systems (based on indicators etc...)

Personally I think that your better off spending time focusing on what mathematical properties constrain the market and reverse engineering systems within the constraints.
Keep it simple stoopid....
  • #4
  • Jan 7, 2009 2:50pm Jan 7, 2009 2:50pm
Quote
I'm not sure what you want for sure but this indicator might be something you can modify.
Also you can look at these on this forum;

Thanks, unfortunately my patterns aren't candlestick or bar oriented.

Quote
MIT has done some research in this subject (http://datacenter.mit.edu/ Your going to have to search for it) Basically it concluded as have all of my own studies that the markets are "White Noise" and there are mathematically significant patterns whatsoever.

Except I know for a fact that there are, as simpler patterns test out quite well. I'm just looking for a way to test more complex patterns (I know what they are, and others claim high numbers with them, I just want to verify their results). One of the problems w/ "academics" trying to run market analysis, they don't know what they're looking for in the first place.

Anyway... after some thought on the subject I think I may have another way to approach this problem.
  • #5
  • Jan 7, 2009 4:04pm Jan 7, 2009 4:04pm
Quoting Rabid

Except I know for a fact that there are, as simpler patterns test out quite well. I'm just looking for a way to test more complex patterns (I know what they are, and others claim high numbers with them, I just want to verify their results). One of the problems w/ "academics" trying to run market analysis, they don't know what they're looking for in the first place.
The technical analysis fallacy. You see what you are looking for not whats mathematically significant. Your mind is playing tricks on you.

An example: Say your at a football game and your team catches the ball near the sidelines and it looks like he got both feet down, you call it a catch because thats what you want it to be; the fans of the other team see that the player stepped out of bounds because thats what they want to see. the Ref during a review determines that the player did not have possession and get both feet down and rules it an incomplete pass... You lose

Same idea applies with patterns here you see a pattern because thats what you want to see, but the REF(math) says that there is no pattern. Sometimes you may get the call to go your way but at the end of the day ... you lose

Quoting Rabid
Anyway... after some thought on the subject I think I may have another way to approach this problem.
Is this just a bragging statement or are you planning on discussing it in this forum.
Keep it simple stoopid....
  • #6
  • Jan 7, 2009 4:16pm Jan 7, 2009 4:16pm
Quote
The technical analysis fallacy. You see what you are looking for not whats mathematically significant. Your mind is playing tricks on you.

Sigh. Except the simpler patterns test quite well. Has nothing to do w/ my mind, it's simple statistics.

Quote
Same idea applies with patterns here you see a pattern because thats what you want to see, but the REF(math) says that there is no pattern. Sometimes you may get the call to go your way but at the end of the day ... you lose

Actually the math says there is a pattern. 65% accurate over 5 years of data. But whatever.

Quote
Is this just a bragging statement or are you planning on discussing it in this forum.

Actually it's a combination of things. I haven't done extensive work on the other approach yet so I don't know how well it works, hence not much sense in posting it. Along the same line, while I appreciate the feedback, I don't plan to get into a debate over technical analysis.

My approach on this is very profitable, I'm just looking for a way to verify some results that others have came up with as a way to expand into other setups. Most of the time I have no trouble coding a test, but these are anything but traditional setups. Until I can verify them for myself I won't be trading them or wasting anyone's time posting them. No sense in adding more pollution to the trading systems sub-forum.
  • #7
  • Jan 7, 2009 4:23pm Jan 7, 2009 4:23pm
Quoting Rabid
Sigh. Except the simpler patterns test quite well. Has nothing to do w/ my mind, it's simple statistics.
Testing well does not = significance...



Quoting Rabid
Actually the math says there is a pattern. 65% accurate over 5 years of data. But whatever.
HUH... I never thought that 65% = pattern... A coin toss is 50%. I think that any success you have in your trading has more to do with money management than a pattern.



Quoting Rabid
Actually it's a combination of things. I haven't done extensive work on the other approach yet so I don't know how well it works, hence not much sense in posting it. Along the same line, while I appreciate the feedback, I don't plan to get into a debate over technical analysis.

My approach on this is very profitable, I'm just looking for a way to verify some results that others have came up with as a way to expand into other setups. Most of the time I have no trouble coding a test, but these are anything but traditional setups. Until I can verify...
I thought the point of this thread was to discuss pattern rec...BTW I have a 2 headed quarter that lands head side up 100% of the time.
Keep it simple stoopid....
  • #8
  • Jan 7, 2009 5:15pm Jan 7, 2009 5:15pm
Quote
Testing well does not = significance...

Depends on how well you test things.

Quote
HUH... I never thought that 65% = pattern... A coin toss is 50%. I think that any success you have in your trading has more to do with money management than a pattern.

LOL. Really? Money management provides absolutely no edge at all. If you're working with a pattern that has negative expectancy I don't care how well you manage your money or how small your trades are, eventually you'll go broke. You may go broke slower than the rest, but the destination is the same. Either you have an edge or you don't, and money management is not an edge.

Anything greater than 50% +/- margin of error is a pattern, provided it's over a large enough time frame, large enough sample size and the distribution of the pattern is correct.

Quote
I thought the point of this thread was to discuss pattern rec...

No, the point was to find resources that could lead to the development of a complex pattern recognition algorithm so I could apply that algo in testing to determine whether or not the numbers I'm being told have any merit. It's not to discuss the merits of technical analysis. If you don't believe in TA then you don't have any reason to look for patterns, so what's the point?

If I find something good in that arena I'll certainly share it. But after some thought I think there may be another way to approach the problem that doesn't require the same level of complexity.

Quote
BTW I have a 2 headed quarter that lands head side up 100% of the time.

Too bad the market doesn't have a similar analogy.
  • #9
  • Last Post: Jan 8, 2009 3:42am Jan 8, 2009 3:42am
Quoting Rabid
LOL. Really? Money management provides absolutely no edge at all. If you're working with a pattern that has negative expectancy I don't care how well you manage your money or how small your trades are, eventually you'll go broke. You may go broke slower than the rest, but the destination is the same. Either you have an edge or you don't, and money management is not an edge.

Anything greater than 50% +/- margin of error is a pattern, provided it's over a large enough time frame, large enough sample size and the distribution of the pattern is correct....
Nicely put; anybody who's read my rants over several months will know that I agree 101% with everything that you're saying here.

However, it's occurred to me that, while I use the term "MM" to reflect the the size and method (fixed frac, fixed ratio, ....., Martingale, whatever) that a trader uses to determine his initial trade risk, other folk use "MM" to also include R:R, stoploss placement and advancement, and other exit-related criteria.

Are we agreed that expectancy is determined (solely) by entry and exit? If so, and others are including their exit strategy under the "MM" umbrella, then by their definition "MM" has a bearing on expectancy, and could therefore be deemed (if statistically proven) to contribute to any overall edge.

In other words, I make a clear distinction between "trade management" (exit) and "money management" (position sizing), but others possibly don't.
0 traders viewing now