Fractals, ZZ & the Pissing Dog

LittleCaro,

The algo is choosing/selecting the pairs, not me...
Except for bug-fixing I'm currently not changing anything.
At the end of the test period I'll do some statistic analysis.

G.

GG, I will throw help your way whenever I can. I think I am armed with a lot more forex ammo than you; so I'm more than happy to lend what I have to you, in hopes that you are more productive with it than I am.

I've dug up the original spreadsheet that was the basis for the Rabbit Traders.

The original Rabbit Trader question was: "How do I turn $5000 ==> $5M?"

The two simple spreadsheets break down the two methods:

1) @ 22% earnings per month; $5K becomes $5M in 36 months.
2) Doubling your money, it would take 10 iterations to go from $5K to $5M. At doubling ever 2 months, you'd do so in 20 months; or less than 2 years. It seems insanely aggressive I know; but it was the original goal of the Rabbit Traders and I have not veered off from it.

You let me know what items you feel you are lacking in your war chest, and I probably have a document that covers that topic. I have 800+ Forex docs and over 1Tb of material. Name it, here to help. This is one of the few threads I feel is worthy of my effort to help.

Just Few word about my journey... and Im not strongly familiar with deep stats and math, But I saw that % winning on a 1:1 RR is the way to start with..

When I started testing the pigz I did test various scenarious. (statistical work on the variables I can put into the EA). Have several scenarious...more that I can handle

After more than 400 trades of one EA pig I could say that some directions of the research have potential. At least 60% of winnings was what I was searching for.

I concentrated on 1:1 RR. I got the best percantage winnings out of that. For me that is the most important statistical fact.

With these, Im moving on from this to try to filter out the trades with the highest probability to maximize the fixed TP and minimize the fixed SL from 1:1 winning percantage. Its very hard to do some statistical work on this part, to try to play a variable SL and TP. This is beyond me at this point.

Plus, I still work on finding new ways to lift the percantage winningz at this point.

What I can recommend is that one when it is forward testing assures the highest possible amount of MT4 and EAs on computers and have as much data as possible out of this. Im testing the most non reasonable scenarious which sometimes give one a direction starter for a new thing.

GG. I think the main problem you are experiancing is:
Avg. winning trade: 1%
Avg. loosing trade: -3,1%

Winning rate is excellent at this point over 70%, but one needs this win rate on 1:1 RR at least- cushion for further market changes.

Keep on!

http://www.admiralmarkets.ru/mqlabs/12.11.2009-zigzag-i-statistika

Zigzag and statistics

From a set of the indicators which are perfectly working for stories and few than useful in real trade, most "beautiful" is ZigZag (A zigzag, further ZZ). A little still what indicator can brag of the exact indication of an entry point to within one point. However, as it was already noted, all this "beauty" shows exact entrances only on history. At real trade the last peak (extremum) of ZZ all the time is in movement and the point останова can be much farther than that value on which this peak appeared for the first time. And only at emergence of new peak, the previous peak is fixed and remains such on history.

Nevertheless, we will try to learn to predict the termination formation of the current peak without expectation of formation new. For this purpose we will adopt two ways of an assessment.

The first way - conservative and even a little banal. It consists in use of the additional indicator as which in our researches MACD very often acts. Essence of application of MACD in that after emergence of new peak of ZZ to wait for MACD turn. For example, new peak of ZZ - bottom. Then we should wait for the moment when the main MACD line (histogram) becomes above the alarm line. This confirmation of purchase. To the contrary, if new peak of ZZ - top, we wait when the main MACD line becomes lower than the alarm. This confirmation of sale.

The second way - statistical. As at our disposal there is a fine tool for allocation of separate phases of the market, it is possible to make calculation of the elementary statistical characteristics: the maximum and minimum movements, and also average value of all considered movements. Between the next peaks of ZZ we will call one such movement "-legged" ZZ. So, touching a quantity of feet of ZZ (for example, 10 last), we can precisely decide on at least three sizes: maximum length of a foot, minimum length of a foot and average length of a foot. We compare length of the last foot to the average received size and if length of the current foot already became more average size further, we make the transaction in the opposite direction from the last peak. Thus, having the minimum value of length of a foot, we use it as the transaction purpose, as the most probable distance which there can pass the price. At the same time, the maximum length of a foot it is used as transaction stop as the least probable size of movement of the price against the transaction.

Uniting two listed methods, we receive system where opening of the transaction happens at coincidence of two criteria - passings at the price of dynamically changing average distance and formation of a signal of the standard MT4 indicator .


In drawing it is visible that the short transaction was open with quite big delay when the price passed already more than a half of the return movement. Too sluggish MACD became "responsible" for it in this case. Nevertheless, it didn't prevent to take profit, though the market and "помотал a nervishka" good kickback. The long transaction was open practically without delay. And in this case MACD just helped as differently (if to take only the average size of feet of ZZ in attention) the transaction would be open much above. Approximately at the level of closing of the short transaction. Here thus these two criteria help each other with trade, insuring against premature entrances.

Obeying normal human aspiration to squeeze out a maximum of any situation, we will make one more addition to algorithm. Apparently in fig. 1, despite all our attempts to approach transaction opening to ZZ peak, it turns out badly. Most often the price still fails against a transaction course, it is good that infrequently reaches stop level. There is a question - why we have to give these points of failures to the market? Whether it was more logical to take away them? Quite logically.

For this purpose we will need additional transactions which will open in an unprofitable zone of initial (primary) position. We will limit the maximum quantity of possible transactions on one signal to favourite number of storytellers - 3. Then from level of opening of primary transaction to its stop we can divide all distance into three parts. At the level of the first third the first additional warrant (the secondary warrant), and at the level of the second third - the second additional warrant (the tertiary warrant) will be located. At the level of three thirds we have the general stop. As a result, even at the most adverse succession of events when the price after all will reach stop, instead of one stop we will receive two (full stop from primary position + two thirds from secondary + one third from tertiary). On the other hand, we have an opportunity to close all cumulative transaction earlier, than was planned initially as when opening additional transactions without achievement of stop we get additional profit on new positions. Therefore at operation of each additional warrant the general profit will be recalculated and be transferred to such level where the general profit will be equal to the originally planned.

work 600


Gold H4


But it is rather an exception, than the rule.
Is over what to work
+++++++++++++++
author Scriptong (Igor Gerasko)

Hi Poruchik,

this looks very interesting. Is there just an indicator which shows the entry and exit points for this EA? I don't like using EA's but like to know how and when it get triggert.

twoblink,

Thanks for that.
Later on I'll explain the inner working of the EA, and ask for your advice or written material on the subject.
Thanks for your support.

G.

21vs7,

My problem right now is B U G S...
On top of it you can add the fact MT4/EA was stuck for 3 trading days with open trades.
Besides causing massive loses, it didn't open new trades to compensate for it.
When I came back and restarted MT4 - it updated itself to 600 build... causing more headache.

This system/EA was never tested live nor optimized in backtests.
The main part of it is WHAT to trade among 28 currency pairs and with what system.
This part was never tested before - and it has a lot of holes and bugs.
Due to the nature of such system - it's very hard to confirm that the selection of pair is the correct one.

G.

poruchik,

This is very good article.
What I did is using statistical approch using and looking at fractals as the building blocks of the ZZ.
With that method I try to guess the next swing. You can see from posted charts that it is quite accurate.
But your approach is very interesting - let me study it more and I'll post my remarks later on.

Many thanks for your input and code.

G.

I understand.
First phase of testing.
Cant talk about bugs, Im not a programmer. Hard to say anything in this direction...just that one needs that all results can be statistically done in excel.

Dont worry about picking the right pairs. All pairs are right!
I trade 21 of them. They should all be in synergy...

That is why its so important that you test several of EAs - several, means 20+ with different options and figure out what one does on the 1:1 RR. I then moved on into right direction, where Im trying to see what advanced tweeks do...

Filtering them out from all, now that is the right story.
I use index based filtering.

21vs7,

My approach in this EA is not profit - but maximizing the potential of every invested $$$, while minimizing the risk.
There is a difference in that respect among currency pairs depended on each individual currency strength, day & time of day, etc.

As to R:R ratio, I tend to agree if your Win/Loss ratio is above 70%.

G.

I understand. And I do the same.

Currency Index gives me all the entries and all the filtering. I dont mind day & time of day.
Testing all around scenarious in this view will give one a bigger picture.

Forward testing is a waste of time, if one does not drives it on a heavy wagon - Hundreds of strategies can be tested in 6m time to get a right direction to tweek later on..

Cheers

poruchik,

Can you please translate the remarks within the EA to English?
Thanks in advance !!

G.

don't download the adviser - 12 mistakes

need recode

poruchik,

Many thanks!!

G.

21vs7,

What currency index/strength indies are you using?

G.

My own.

My good friend A programmer did create all my ideas..Top art! (cant do a line of code, could do a line of c***)
I use only 21 pairs. 7 Currencies. For this the name 21vs7. Started 6 years back..

OK.

Have several speeds of indexes. On all TFs.
Every speed on every TF has its own DNA, as you would put it. This edges one could use as potential direction and potential entries.

The problem I had half a year ago, was that I couldnt backtest this beast. Cus this EAs dont look at the eurusd alone, they only look at EUR and USD. And you cant backtest EURUSD without all 6 EUR pairs and six USD pairs. That is why forward testing is crucial.

Had killed over 100 bad EAs and only seeing what is getting better and better results. Then you see the bigger picture, what matters.

ATM Im testing around 200 EAs. Several are 10.000 pips in green, which is ridiculous. But im only looking for one with 90+ percent atm.

21vs7,

Roughly the same here.
I'm currently using Currency Strength indie that I developed years back and it's still evolving.
Backtesting is indeed a problem for those kind of indies.

G.

Cool.
My friend developed really good structure. Meaning...one could try everyfooking thing not needed to be a coder. So I just played.
Also the importance of excel to play out the statistics out of hundred of EAs with magic nmb.

Yes, with no backtesting, you should not test only one EA. Try at least 10 with different settings. One gets a feelling then what is not working and what could work.

Here is new version with bugs fixed from build 509.
There was a compare using = instead of ==.
Next step is to move to my laptop with new build 600 and check again.
The Doc file looks like code. What is the purpose?
Robert