LinkBack URL
About LinkBacksBut what if we get another year with the same PA as 2005? Flip again?Originally Posted by doczero
But what if we get another year with the same PA as 2005? Flip again?
Hi Narn,
Nope - whatever the system - the last test is passing all historical data. Basically we are looking for consistency. Currently this system has produced 1 good year, 1 bad / mediocre year. If 2007 is bad then I will flip the entry and keep forward testing. The thing is whatever we settle on must work year in year out.
So if we cant get consistency we throw the entry away and test a new idea. In my first posts I mentioned I did not know if this would work or not. The entry must pick up on the desired price action before it happens - the system has to be setup behind the entry to make the reward / risk and accuracy into a positive expectancy.
This definitely is a lengthy process - however I haven't risked a dollar until I know the average outcome over a large sample size. Its a good insight into how long it takes to find something that works and be confident in the system.
keep the comments coming - additionally if you have some ideas / direction to add - throw it in the mixYour experience is most welcome.
Ok, So i've just pushed the system with 25 SL / 100 TP through 2007.
The results are a lot better , This gives us an expectancy of $8.07 or 0.31R expectancy.
The current Issue I have with this system is that it does not comply with my personal goals. I was looking for something around 50% win ratio - additionally the largest consecutive losses is 18. There is no way my psyche could take an 18 loss streak before I started doubting the system.
So what does this mean? My next step is going to reverse the entry. I will be trialing as per before - first of all multiple stop / profit targets. Then I'll walk forwards the one with highest expectancy.
As technically this is a new system (ie different action on entry) I will be starting from year 2005 again.
Lets see how we goI'll keep you guys posted!
Very nice doczero
One 'thing' I use in my own trading plan is to always trade with the bigger trend. The entry system is based on a 30m chart. Maybe throw in a daily or 240m chart and only trade with the trend on that timeframe. You will have less trades but it should improve your win/loss ratio.
Expect less performance when the trend is reversing.
Lets give it a shot!
So what do you suggest for the definition of a trend on the higher time frame, running a 30day MA and if price is above uptrend?
I think i'll throw in the 4 hr chart, will be of interest.
You could use the same setup as on the 30m chart. 2MA's that need to be stacked in the same order.
There are hundred ways for trend determination. All have their strengths and weaknesses. I use trendlines and support/resistance to figure out the (possible) trend. But this is hard to code. Therefore I never was able to create a fully automated system.
really good post doczero!Ok, So now we have some Rules - we can get onto the fun part: TESTING!
So how do we do that you ask? Well first thing we need to grab is some software. I use Forex Tester 2 as its pretty simple to use and I like the interface. However there is many pieces of software on the market for this (hehe search the forums.. might be something hiding there)
Next we load it with data. 10 Years is a good starting point, and only 1 pair at this point in time. I've got 10years of EURUSD.
Now the next step is critical - Choose 1 year to test (In-Sample Data). So all our system building / alterations will be done in this year. The reason why we choose only 1 year is because it is so easy for us to over optimize (while looking for the holy grail- this is called curve fitting) and we end up with a system that only works in that time period. Im going to choose year 2005
Once we have done that period (I actually like aiming for 100 trades) and we have finished altering and modifying until it works, we have to test the next time period (known as Out-of-sample Data). No changes can be made in out of sample data, we just run it fully and get to the end of the year. If you are happy with the results and can test the next period ahead. However if you are not happy go back to the middle of 2005 and test/alter/modify from there to the middle of 2006.
This process is called: Walk-Forwards Testing , and can be summed up by this picture (just ignore weeks):
Once we get the system to a stage where it has passed multiple out-of-sample tests to your satisfaction - you can finally either 1) trade the system with reduced capital at risk, or 2) paper trade until.
Why not start with full $$? Mainly because the testing procedure is fairly mechanical and does not put pressure on your psyche. You dont know how you will react while trading the system and its a good thing to make sure YOU can trade this system.
Anyway - I'll be starting the testing phase shortly.
Walk-Forwards Testing is one of the few valid strategies for developing algorithmic trading.
-- Is this useful information ? -- PLEASE SAY THANKS --
Ok, So i've changed the setup and thrown in the 4 hour chart to track longer term trend. As discussed above the trend definition is usong the 15 bar MA / 30 bar MA. When 15 is above 30, long only, with 15 below 30 short trades only.
It looks like this:
Its funny, I have tested 2005, and i've found by adding in the filter - I haven't reduced my amount of trades.
The results are as follows (using 25 SL / 100 TP)
Once again this illustrates how difficult trading truely is. Imagine trying to get a system to work while using real money. Its crazy how many people jump into the market , read a few books, and imagine they can make money. Some have luck (but only for a bit) and the rest sink. Its the dedicated that learn how to swim (hopefully they havent sunk too many accounts in the learning process).
I've gotta jet for now - but im gonna think this over some more.
2005 was a positive year, could you test with 2006 and 2007? The first results where not so good, let's see if adding the long term trend filter improves this.
I'm just giving ideas and you do all the hard work. If I need to STFU let me know.
Hi Narn,
STFU! ... lol kidding! The input is appreciated- It helps show the community what it takes to get an idea working. I will be testing forwards- im currently toying if I should change the TP / SL. Im unsure what I'll decide at the moment but i'll post some results for 2006 / 2007 in a few days.
Posting Permissions 
Reply With Quote
