Statistical Analysis

 

hi, i am planning to embark on a project that would attempt to build correlation between how an indicator behaves just prior a breakout of a certain value.

For example, how would a bollinger band (or macd or RSI..it can be anything) behave in t-1, t-2, t-3, when i know that within t+15, the price wld have broken out in a certain direction by 30pips

I have already tried this out with a custom indicator...and there's no meaningful correlation, (there goes that particular custom indicator) and now intend to run this excercise with other indicators

What i hope to discover is a statistically-backed set of conditions that leads to a certain outcome, for instance, how would a bollinger band behave just prior to a breakout that would net minimum of 30 pips in the next 15 candles. then by varying the outcome conditions, (meaning 15 pips in 10 candles, etc), i can create a contour map.

(note, you can't directly use this to trade as you would need to be able to know the future first, before deciding to collect the data points. But a statistical analysis of historical data might point to an evidence-backed use of some indicators)

I'm wondering if there is any discussion already out there on this particular topic....

 
wolvl:
I'm wondering if there is any discussion already out there on this particular topic....

Probably but not on this forum. When you registered you missed the purpose of this forum: Any discussions except of concerning MetaQuotes Language 4 and auto trading are forbidden.

 
wolvl:

hi, i am planning to embark on a project that would attempt to build correlation between how an indicator behaves just prior a breakout of a certain value.

For example, how would a bollinger band (or macd or RSI..it can be anything) behave in t-1, t-2, t-3, when i know that within t+15, the price wld have broken out in a certain direction by 30pips

I have already tried this out with a custom indicator...and there's no meaningful correlation, (there goes that particular custom indicator) and now intend to run this excercise with other indicators

What i hope to discover is a statistically-backed set of conditions that leads to a certain outcome, for instance, how would a bollinger band behave just prior to a breakout that would net minimum of 30 pips in the next 15 candles. then by varying the outcome conditions, (meaning 15 pips in 10 candles, etc), i can create a contour map.

(note, you can't directly use this to trade as you would need to be able to know the future first, before deciding to collect the data points. But a statistical analysis of historical data might point to an evidence-backed use of some indicators)

I'm wondering if there is any discussion already out there on this particular topic....

Hi wolvl,

Interesting. I did some statistical for harmonic pattern several times and I'm gonna have it again actually. Just google harmonic pattern and you'll find some interesting statistical analysis. However, no, I never heard such discussion on internet, but maybe I'm wrong.

just teasing WHRoeder, we could create some mql4 program to collect the data, you know ;D.

:D

 
WHRoeder:

Probably but not on this forum. When you registered you missed the purpose of this forum: Any discussions except of concerning MetaQuotes Language 4 and auto trading are forbidden.


onewithzachy is right... mql4 allows me to write such programs to get the necessary data...

guess i'll limit the discussion to when some of the codes (and not the results) dun come out as expected :P

onewithzachy - thanks, i'll search for harmonic pattern

Reason: