Statistics and Forex

Hello traders
It is my first thread here and i'm interested in statistics however its not my main study . But let's think simply any price has certain probability to go to any price above it or below it . Some people said that any price has the same probability to go from any price up or down for the same distance but i think that is wrong because market movement is not separate . It is related to each other so we will get the probability for any price to go or down for certain distance by studying the history of this price . You can see that on EURNZD M5 chart when price touched 1.6140 it went 3 times to 1.6175 and one time to 1.61050 for the last 2 days and when price was 1.6140 today it went directly to 1.6175 .

What are you trying to say?

firstly sorry for my english Look at the picture below if the price moved from point A to point B 7 times and moved from A to c two times only what do you expect when price comes to point A next time ???

Nothing

Why ? Can you explain your point of view ?

Wow, tough audience, today. I think Mhmdfx is making a very good point, and making it rather well ...

You cant "Assume" that it will , all prices are Not created equal.
From an AMVT perspective you have 7 "price time events" (tpos) at B , 9 "price time events" at A, and 2 "price time events" at C. You have price "being used" more at A&B (at higher prices)
Sellers trade at higher prices, and you have price "being used" at higher prices more often. Therefore, the market offered more "price time opportunities" for sellers than it did for buyers ..

I think that perspective deals with price as separate events but this is not true its connected to each other. Any price will go to other price and so on

Thank YOu

You're spot on. FX Options are regularly priced on black scholes. It's all about normal distribution/bell curve.
See: https://en.wikipedia.org/wiki/Normal_distribution

That's quite right mhmdfx

In determining probabilities of price movement you need to address movement away from the mean. Obviously there is less probability to take an excursion from the mean of say 2 standard deviations away than there is of taking say 1 standard deviation of movement away from the mean. So you need to ask yourself, where is the mean which is a moving feast for the timeframe you are working on?

Here is how I apply a bit of stats to my trend trading to at least underpin the decision making with some statistical basis as opposed to simply trade off brain farts. I hope it helps.

In statistics we use sample size as a basis to determine the confidence of our assumptions so a mean taken from a price series of say 200 seperate data points has more meaning than say a mean taken from say 50 data points. This basis can therefore be applied independent of timeframe and relates to the sample size and it's relevence to the total data population of that timeframe. In general simply as a starting point (an this applies to if I were trading M5), I only consider sample sizes of >=200 in my analysis as you have to at least start somewhere and the more sample you have the better it is going to represent the timeframes price data. Obviously as you move up to longer timeframes you need to choose a sample size that is sufficiently representative and something you are happy with. For example look at the history to determine trends of significance that you would like to trade and determine a standard sample size you are happy with).

You can use a linear regression channel or standard deviation channel across your data set to determine the 'line of best fit' that is most representative of the mean price of that data series.

For example in applying this to trends, the first thing you want to find out for a trend is what is the trend momentum and average slope of 'drift' of the price series. The example below takes your chart and applies an SDC channel to the data between major swing highs and major swing lows. On this timeframe we are going to treat for the purposes of ths exercise a trend as that which is defined by the price series data between a commencement point (swing high/ swing low) and an end point (where a touch is made of the outer SDC channel).



According to this trend interpretation we can see that Trend 1 commenced at the swing high on 24/11/2015 at 13:20 GMT and ended where price touched the outer SDC channel on 26/11/2015 at 12:15 GMT (green cross). The entire price series was contained in a standard deviation channel of 1.30. (note that the SDC is calculated off a moving average of the price series and is not a true reflection of how a SDC is calculated but for the purposes of trading it is sufficient).

The mid line of the channel (blue) is the line of best fit representing the mean of the price data series and represents the overall linear drift of the price series. Price moved away from and towards the mean along the course of the trend. You can insert quartiles into the price series to determine zones of overbought and oversold conditions if you want to trade mean reversion retracements or you can simply trade the trend using bearkout or retracement entries recognising the overall direction of the price drift.

The slope of the Line of Best Fit is representative of the overall momentum of the trend. In both trends defined on the chart you can see that momentum is quite good for both trends defined and you can use the line slope function in MT4 to quantify this slope if you wish to rank the momentum against alternatives. It is also helpful to use the SDC width as a basis for ranking and a narrower SDC indicates a less volatile trend where price better hugs the line of best fit. Both these ranking measures are useful if you trade across a portfolio of correlated instruments and you need to restrict your opportunities to a few choices amongst several.

The sample size of Trend 1 was 565 bars of data so in terms of the confidence surrounding this data ( where n>=200) then it is sufficiently representative to suggest that the data was meaningful as opposed to simply a random representation......however you are never going to be certain regarding this assumption. In general however more sample size is better than less sample size in it's deduction potential. Also the important question to note is where you took your entry in this price series. Obviously after say 200 bars, and entry would most likely have delivered a profit but an entry at say 400 bars of data would have led to a loss. Note however that if you use a trailing stop from entry and risk say 0.5% on entry then using a trailing stop along the outer SDC ensures that at each tick interval your loss is reduced from entry. This approach therefore manages trade risk very well.

Clearly according to the M5 timeframe, price is now in an uptrend defined by the SDC of trend 2 and we now have a data series of 375 bars between the swing low and the most recent swing high so we are more confident of the assumptions surrounding the SDC and its projection.

Anyway, you probably can then define a trading solution within this analysis to suit your personality. I simply like the trend following method using a simple breakout strategy and a trailing stop at the upper/lower SDC and only risk per trade up to 0.50% of trade capital on entry.....however you may want to use retracement entries into the direction of the trend, and some others get a lot of joy scalping from the overbought/oversold zones in addition to trend trading opportunities.

Go Study IQR !
It is the best way to trade forex with statistics

Good work Copernicus . I agree with you that more sample size is better than less sample size so the more data you have about any point the more safe entry
I like your idea about using the mean. But now I think I want indicator that can calculate the history of any point

The indicator idea could be a breakthrough; It could reveal a lot about price action and the probable future market direction. Try to explain your indicator idea to one of FF programmers and let's see what they can do.

Unfortunately i don't know any programmer here

Okay, let me see what I can do...I'll try to contact a programmer to turn the idea into an indicator. I'll make sure to post any updates here to your thread...Cheers!