Menu
Sister Sites

Options

Bookmark Thread

First Page First Unread Last Page Last Post

Print Thread

Similar Threads

BASKET TRADE Management: Using multiple trade management EA's on basket trades 9 replies

Alternative Trade Management/Money Management Methods 3 replies

Money Management..keeping the money... 39 replies

Money Management / Risk Management 24 replies

Help with Money Money Management Please! (Free EA) 4 replies

Attachments: Money Management 101
Exit Attachments

Money Management 101

Quote
Quoting mmeimandi

My point was about those system that are so tied up in rules over rules that have you say, how on earth this system is going to trade anything.

Thanks,

In memory of a once fluid man, cramped and distorted by the classical mess.

Inscription on mock tombstone in Bruce Lee's original Chinatown workout space.
Quoting mmeimandi
I am definitely going to write about position sizing in detail. In fact I believe no money management discussion is ever complete without saying something about Position Sizing.

Thanks,

Moh
Thanks.
  • Post #23
  • Quote
  • Edited 2:18pm Jun 20, 2007 8:29am | Edited 2:18pm
Ok I was talking about "Dependency" and the idea of exploiting it wisely in our system.

Good news is that there are methods to help us discover the dependency between trades. A simple method would be the “Runs Test” and another more efficient and elaborate method is to find the “Serial Correlation” between the trades. Personally I prefer the latter since it not only tells me that there is dependency but also I can find out if it is positive or negative.

Positive dependency basically means that winners beget winners and losers beget losers. So if I just saw a big win I will increase my position size because of the positive dependency between the trades.

Negative dependency is just as helpful. For example if I just suffered from a large loss, I can expect the next one to be a winner or at least a small loser.

I will dig into the details of Runs Test and Serial Correlation later.

One thing I want to clarify here, is that as I said above generally you have to assume that there is no dependency whatsoever. Only in those rare cases that the two methods show you a high “Confidency Limit” can you assume the dependency and exploit it. My philosophy is if there is dependency but you assume that there is not is much safer than the opposite. At least you don’t lose anything.

Capitalizing on dependency should never be used as the sole position sizing strategy in trading. It is good to know how you can discover and exploit dependency but it can never replace proven position sizing methods like “Fixed Fractional”, “Fixed Ratio”, and etc.

We will talk about these methods in upcoming posts. Make sure you stay tuned.

Thanks,

Moh
Hi Moh,

Say you find out there is no dependency between your trades how do you handle position sizing?

Thanks,
Al
Quoting aparsai
Hi Moh,

Say you find out there is no dependency between your trades how do you handle position sizing?

Thanks,
Al
Al,

Dependency is a very rare thing to happen between trades. That's why it shouldn't effect the position sizing strategy very often.

I personally recommend embedding dependency as a pre-check in Position Sizing strategy. Say we are using Fixed Fractional method or any other method for that matter. First step is to determine if there is dependency between trades or not. Next we calculate the size according to the rules of our method. At this point if we know that dependency exists from step one we can decide to alter the size by some percentage based on dependency being negative or positive and if there is no dependency then we just go ahead with the result of the calculations.

Thanks,

Moh
One of the methods that can be used to check the dependency between trades is Runs Test.

The runs test is essentially a matter of calculating how many standard deviations you are away from the mean of a distribution or better said how many standard deviations you are away from a random distribution of streaks of wins and losses. This measurement is known as Z score. For example, a Z score of 1.00 would mean that the data you are testing is within 1 standard deviation from the mean. In a trading sequence the Z score means how many standard deviations it is away from being completely random. The higher the Z score the less random the sequence is.

Next we convert the Z score into a confidence limit or degree of certainty.

It is recommended to use a minimum of 30 trades in computing the Z score. But the more you can dig in history the better.

Let’s clarify with an example. Consider the following sequence of trades:

-3,-1,0,0,+8,+6,+10,-4,0,0,0,+12,-9,+5,0,+1,-2,-11,+4,-2

Now we make some calculations:

The net profit is +14.
The total number of trades is 20.
The total number of wins is 7.
The total number of losses is 13. (Note that P&L of 0 is regarded as a loss)
The total number of runs is 9 (a run is anytime you encounter a sign change)

The equation for calculation the Z score is:

Z = (N*(R-.5)-X)/((X*(X-N))/(N-1))^(1/2)

Where:
N = The total number of trades in the sequence = 20
R = The total number of runs in the sequence = 9
X = 2 * W * L = 2 * 7 * 13 = 182
W = The total number of winning trades in the sequence = 7
L = The total number of losing trades in the sequence = 13


Therefore:

N * (R - .5) – X = 20 * (9 - .5) – 182 = -12

And:

X * (X – N))/(N – 1) = 182 * (182 – 20) / (20 – 1) = 1551.789473

And:

1551.789473 ^ (1/2) = 39.392759

Finally:

-12/39.392759 = -0.30462451 simply take the absolute value of this and you reach at:

Z = 0.30462451

A “Z” score of 1.64 means that there is 90% confidence limit. A “Z” score of 3.00(three standard deviations) means that there is 99.73% confidence limit. Since we are at such a relatively low number in our example, we can assume that there is no dependency between trades in this sequence.

A negative “Z” score implies positive dependency meaning that wins follow wins and losses follow losses. Same way a positive “Z” score implies negative dependency; that means wins are followed by losses and losses are followed by wins.

Statisticians recommend a confidence limit in excess of 99% in order to assume dependency. Rarely if ever you will find a system that shows confidence limits in excess of 95.45%(2 standard deviations).

Thanks,

Moh
Hi,

I was wondering, if one could use Runs Test to measure the dependency between pairs?

PipItUp
Hey Moh,

Is there a spread sheet availalbe for the Runs test that you can enter your data and it calculates the Z factor?

Thanks,
Al
Quoting PipItUp
Hi,

I was wondering, if one could use Runs Test to measure the dependency between pairs?

PipItUp

PipItUp,

I don't think that it is a good idea to use the "Runs Test" for measuring the dependency between pairs. The context is different.

To measure the relationship between currency pairs you should be looking at the correlation between pairs. Serial Correlation could be used to measure the dependency amongst trades as well as pairs. In fact not only can you study the relationship of currency pairs but also you can check the relationship of any market systems using the "Serial Correlation" method.

Thanks,

Moh
Quoting aparsai
Hey Moh,

Is there a spread sheet availalbe for the Runs test that you can enter your data and it calculates the Z factor?

Thanks,
Al

Hi Al,

Actually I don't have such a spreadsheet. In fact when it comes to measuring the dependency between trades I prefer using another method which is "Serial Correlation". I am going to dedicate a complete post to this method.

I know that there are lots of traders that use Excel to track and maintain their trading record but personally I prefer using a database to do that. In my opinion while Excel is a very powerful and easy to use tool it is not appropirate to maintain data. It is suitable for analyzing data but not keeping it and when it comes to getting reports, measuring the performance, integration with portfolio management softwares and looking at data from different perspectives I believe using a database management system would give me much more flexibility and power. Of course that is just my opinion.

All that said I think I give that spreadsheet idea a shot. I guess it would be useful for everyone.

Thanks,

Moh
Quoting mmeimandi
Al,

You are absolutely right. If using parameters gives your system more alternative ways to enter, exit, etc. by all means go ahead and use them.

My point was about those system that are so tied up in rules over rules that have you say, how on earth this system is going to trade anything.

Thanks,

Moh
I find this very true...
infact its due to these facts that traders start trading over 4 currencies, just so that they could find 'the setup', a more worthy note is to keep tabs on pair correlation instead and a simple entry and exit rules...plus less price analysis.
Quoting mmeimandi
Hi Al,

Actually I don't have such a spreadsheet. In fact when it comes to measuring the dependency between trades I prefer using another method which is "Serial Correlation". I am going to dedicate a complete post to this method.

I know that there are lots of traders that use Excel to track and maintain their trading record but personally I prefer using a database to do that. In my opinion while Excel is a very powerful and easy to use tool it is not appropirate to maintain data. It is suitable for analyzing data but not keeping it and when it comes to getting reports, measuring the performance, integration with portfolio management softwares and looking at data from different perspectives I believe using a database management system would give me much more flexibility and power. Of course that is just my opinion.

All that said I think I give that spreadsheet idea a shot. I guess it would be useful for everyone.

Thanks,

Moh
Thanks Moh.
Quoting jest1081
I find this very true...
infact its due to these facts that traders start trading over 4 currencies, just so that they could find 'the setup', a more worthy note is to keep tabs on pair correlation instead and a simple entry and exit rules...plus less price analysis.

jest1081,

Thanks for the comments. I think it is obvious that I am totally with you on this

Thanks,

Moh
Using the “Runs Test” we are trying to determine if the sequence of wins and losses exhibit dependency or not. One thing that might not be obvious is that the “Runs Test” accounts for the sequence of wins and losses but not their size.

Serial Correlation looks at the sizes of wins and losses from a different perspective. This technique utilizes the linear correlation coefficient, r, sometime called Pearson’s r to quantify the dependency relationship.

Here is the formula for finding “r” between two sequences, X and Y:

r = (∑(Xa - X‾ ) * (Ya - Y‾ )) / ((∑(Xa - X‾ ) ^ 2) ^ (1/2) * (∑(Ya - Y‾ ) ^ 2 (1/2))

A bar over a variable means the arithmetic mean of the variable.

I know the formula looks kind of scary so let’s go through an example. We use the same sequence of trades that we used for “Runs Test”.

-3,-1,0,0,+8,+6,+10,-4,0,0,0,+12,-9,+5,0,+1,-2,-11,+4,-2

As you can see the formula needs X and Y. Basically correlation is measured between two processes. So here we use the trade P&L’s as the X value and we use the same sequence for Y only this time we skew them by 1 trade.

First we average the X’s and the Y’s.

X = -3,-1,0,0,+8,+6,+10,-4,0,0,0,+12,-9,+5,0,+1,-2,-11,+4,-2
Y = -3,-1,0,0,+8,+6,+10,-4,0,0,0,+12,-9,+5,0,+1,-2,-11,+4,-2

We only average those X’s and Y’s that have corresponding X or Y value. Therefore:

X‾ = 0.8
Y‾ = 0.7

Now we find the difference between each X and the average X and each Y and the average Y.

X - X‾ = -1.8,-0.8,-0.8, 7.2, 5.2, 9.2,-4.8,-0.8,-0.8,-0.8, 11.2,-9.8, 4.2,-0.8, 0.2,-2.8,-11.8, 3.2,-2.8

And

Y - Y‾ = -3.7,-1.7,-0.7,-0.7, 7.3, 5.3, 9.3,-4.7,-0.7,-0.7,-0.7, 11.3,-9.7, 4.3,-0.7, 0.3,-2.7,-11.7, 3.3

Now to calculate the numerator, multiply the answers from step 2:

(X - X‾) * (Y - Y‾) = 6.66, 1.36, 0.56,-5.04, 37.96, 48.76,-44.64, 3.76, 0.56, 0.56,-7.84,-110.74,-40.74,-3.44,-0.14,-0.84, 31.86,-37.44,-9.24
Total up all of the answers to step 3 and you will have your numerator:

= -128.06

Next sum up the squared X differences and squared Y differences and take the square root of them.

∑ (Xa - X‾) ^ 2 = 597.96, ∑ (Ya - Y‾) ^ 2 = 604.91

AND

(∑ (Xa - X‾) ^ 2) ^ (1/2) = 24.45322, (∑ (Ya - Y‾) ^ 2 (1/2) = 24.59492

Now multiply the results together to find the denominator

= 601.425

And finally divide the numerator and denominator to find the “r”:

r = -128.06 / 601.425 = -0.21293

The value for “r” will always be between +1.00 and -1.00 and value of 0 means no correlation.

To be honest I didn’t expect to reach such a high linear correlation coefficient but still it is not enough to assume any dependency. High positive correlation (at least .25) suggests that big wins are seldom followed by big losses. Negative correlation (below -.25) imply that big losses tend to be followed by big wins and vice versa.

Well this is it about serial correlation but it is nowhere close to be complete. It is recommended to break your data into two or more parts. If you see dependency in the first part, then see if that dependency also exists in the second part. This usually leads you to a more accurate result.

Thanks,

Moh
Attached Image(s) (click to enlarge)
Click to Enlarge Name: NegCor.jpg Size: 30 KB
Click to Enlarge Name: PosCor.jpg Size: 30 KB
Quoting mmeimandi
PipItUp,

I don't think that it is a good idea to use the "Runs Test" for measuring the dependency between pairs. The context is different.

To measure the relationship between currency pairs you should be looking at the correlation between pairs. Serial Correlation could be used to measure the dependency amongst trades as well as pairs. In fact not only can you study the relationship of currency pairs but also you can check the relationship of any market systems using the "Serial Correlation" method.

Thanks,

Moh
Hi Moh,

Thanks for all the info. So I assume I can use your new post about "Serial Correlation" to check the dependency between pairs. I read some where that using the linear coefficient I can achive this.

Im gonna make sure to bookmark this thread it is really informative.
My input on this thread is that i dont recommend a trade that does not have at leaST a 2:1 profit:lose ratio.

chances are u will only win 50% or less of your trades
100% of traders are losers. Just that some win more than they lose!
Quoting Mr demark
My input on this thread is that i dont recommend a trade that does not have at leaST a 2:1 profit:lose ratio.

chances are u will only win 50% or less of your trades
Over the long run with proper MM it is proven most traders will become profitable. But something ticked me and i have heard this from two people in this forum already, those who are above average are in a different class alltogether. Its as if they know when and how to beat the red lights? Pattern recognition is probably the most important thing in trading charts?

How important is analysis in our trading? Traders game is to make money in the market, while thje ANALyst throw alot of crap on the TV, to become really profitable in this business does it constitute a huge sum of analysis or all we really need are just some basic entry triggers and exit rules? Please share your views and concerns.

regards
I think we need to post examples of trades with proper money management. Without this, many will be lost. And yes, i believe a proper MM will beat the economic analysis. Technical analysis coupled with MM on the other hand, will produce excellent result
100% of traders are losers. Just that some win more than they lose!
  • Post #39
  • Quote
  • Edited Jun 29, 2007 12:30am Jun 28, 2007 11:25pm | Edited Jun 29, 2007 12:30am
A next thing, seeing a setup and not taking a trade because the R:R is not enough. Like for example u see a nice setup but the required SL in pips is 45 but there is a support level in the direction of the trade 40 pips away.

i wouldn't take it....why? R:R ratio not greater than 2:1. that is a simple probability based trade

permission to post an external link that demonstrates this
http://www.fibonaccibook.com/?page=4 go down to #3
100% of traders are losers. Just that some win more than they lose!
Quote
As a trader I generally assume that dependency does not exist in the marketplace for the majority of market systems. When trading we will usually be operating in an environment where the outcome of the next trade is not predicated upon the outcome(s) of prior trade(s).
hello mmei mandy,

while the outcome of the news is predicted by the outcome of previous news, why not the trade? is it purely analysis?
0 traders viewing now