Inference about bernoulli distribution of basic trades | Forex Factory

Options

Permalinks
Print Thread

Similar Threads

MACD indicator with distribution display for MT4 4 replies

Making Money and the Most Basic of All Basic Strategies 42 replies

Request: Price distribution indicator 0 replies

Accumulation/Distribution patterns.... 0 replies

The Magical Bell Curve of Standard Distribution 23 replies

Inference about bernoulli distribution of basic trades

  • #1
  • First Post: May 4, 2012 5:09am May 4, 2012 5:09am
Before delving further into this material.

Let us think of a simple experiment.

1) Enter market at random moments
2) Set TP+10 pips, SL-10 pips
3) Succes chance p is hitting the TP

In case of no spread random markets, we could expect the proportion of our trades to hit the TP to be 50%, hence p = .5.

Can anyone link me to a simple study like this with below hypothesis?

H0: p = .5 Ha: p =/ .5 with alpha confidence .05 and a sample size of 25 trades of above mentioned criteria?

Or how could I go about and do it myself?



Certainly for automated trading systems, this is the absolute basis. From here you could look:
- What if we only took trades with direction of 200MA?, make same hypothesis but now is p > .5?
- What is the influence of profit factor?
- What is the influence of spread relative to TP/SL?

ahh anyways, just woken up and this study idea came to mind.
  • #2
  • May 4, 2012 5:40am May 4, 2012 5:40am
This would be easily done with an EA. Beyond me though. With 10 pips of course it wouldn't be profitable because of spreads.

you could do a quick and dirty paper test on a 50 pip grid. (long and short at each 50 point)

Good luck
  • #3
  • May 4, 2012 1:43pm May 4, 2012 1:43pm
Your only problem is that markets are not truly random. So your equation just got a hell of a lot more complicated.
  • #4
  • May 4, 2012 2:29pm May 4, 2012 2:29pm
Well that is the thing we are trying to find out right, to which extent the market is not random. (how p = .5 deviates from our hypothesis H0)

I will find a list of 1min data on euro/dollar, take SRS of an open and see what get hits earlier, sl or tp. Ofcourse it's not really precise, and I ignore spread.

But it's a start.
  • #5
  • May 4, 2012 2:32pm May 4, 2012 2:32pm
http://www.dukascopy.com/swiss/engli...ed/historical/
  • #6
  • May 4, 2012 2:34pm May 4, 2012 2:34pm
https://en.wikipedia.org/wiki/Random_walk_hypothesis

might be of interest to you.

Your attack vector is not in if the market is random or not but how good that random generator is like i have written in my thread about indicators.

So with a random walk you are attacking it from the wrong side. You need to attack the source / random number generator which is the traders / alogs psychology. You will find plenty of attack vectors there.
  • #7
  • May 4, 2012 10:15pm May 4, 2012 10:15pm
Quoting FrankS
Before delving further into this material.

Let us think of a simple experiment.

1) Enter market at random moments
2) Set TP+10 pips, SL-10 pips
3) Succes chance p is hitting the TP

In case of no spread random markets, we could expect the proportion of our trades to hit the TP to be 50%, hence p = .5.

Can anyone link me to a simple study like this with below hypothesis?

H0: p = .5 Ha: p =/ .5 with alpha confidence .05 and a sample size of 25 trades of above mentioned criteria?

Or how could I go about and do it myself?



Certainly...
Not to put too fine a point on it, but you're asking a question with no provable answer. It would be like asking how many rain drops would hit the head of a pin during a storm. You can conjecture, but you will never truly know. Why even bother? It's so easy just to master market fundamentals and be on your way to a successful trading career.
To improve is to change. To perfect, is to change often.
  • #8
  • May 4, 2012 11:27pm May 4, 2012 11:27pm
I've tested in reverse pattern recognition with MA or any other correlated f. to the price with the random walk generator (so not the market) and you need a big big sample to get a true sample that will be random. My test was random price movement, gauss distribution of movement amplitude, central tendency f=0. So this is just a warning on just how big of a sample you need because a lot of stat analysis tools will tel this is not random within the 95% conf. True results are hard to find and you are most likely to fall between 0,4-0,6 range if you do hit anything...

you should also count that on M1, imo, price is often only deviating slightly from the 0,5 balance point of the equidistant move in both directions. your TP/SL size for test is a significant factor because price can be off balance for a move of 3 pips but not 10 pips. so a distribution can look like and it is diminishing:
3 pips => 0,7
4 pips =>0,6
5 pips =>0,55
...
10 pips => 0,5
  • #9
  • Last Post: May 5, 2012 6:37am May 5, 2012 6:37am
Quoting FrankS
Before delving further into this material.

Let us think of a simple experiment.

1) Enter market at random moments
2) Set TP+10 pips, SL-10 pips
3) Succes chance p is hitting the TP

In case of no spread random markets, we could expect the proportion of our trades to hit the TP to be 50%, hence p = .5.

Can anyone link me to a simple study like this with below hypothesis?

H0: p = .5 Ha: p =/ .5 with alpha confidence .05 and a sample size of 25 trades of above mentioned criteria?

Or how could I go about and do it myself?



Certainly...
Check these two articles : http://articles.mql4.com/631

http://articles.mql4.com/659
0 traders viewing now