Chinook's EUR/USD (E/$)Mumblings

[chinook]

Quote from Andy62279:

chinook,

do you have enough data to know the system is profitable? I hear 300 entries is good enough to know if a system is profitable or not. For me, 2 yrs of data is barely over 300 entries... but I went the extra mile and did 3 yrs of data.

Took me several months to complete...

I have about 100 entries for each data set. And I have two datasets. However, the system also performs well on the Yen and is stable on other instruments. It's a very simple system and wins about 40% of the time with 2.3/1 Win/Loss ratio on Euro futures. I think it'll at worst break even. And I can test new ideas as I go along.

I see this system exactly like a casino game. But in this case, I have an edge (positive expectancy) so in the long run it should pay off.

 

[Andy62279]

Quote from gartley222:

Is there anywhere i can DL EUR/USD data? I wanted to take a look on a daily chart going back a few months. Fores is really only good for short term.

i replied to your request at the other thread

 

[Andy62279]

Quote from chinook:

I have about 100 entries for each data set. And I have two datasets. However, the system also performs well on the Yen and is stable on other instruments. It's a very simple system and wins about 40% of the time with 2.3/1 Win/Loss ratio on Euro futures. I think it'll at worst break even. And I can test new ideas as I go along.

I see this system exactly like a casino game. But in this case, I have an edge (positive expectancy) so in the long run it should pay off.

My first system traded in direction of the trend in a higher timeframe. It had problems being out of sync sometimes. When the higher timeframe was probable of retracement, I had many false signals where I was stopped out, but I'd have a good signal after that. I just had too many unnecessary losses. Using higher TF can be a good tool, but I wasn't using it the right way.

That's what I like about mechanical systems. No need to think while trading, so you can think about new ideas.

 

[chinook]

Quote from Andy62279:

My first system traded in direction of the trend in a higher timeframe. It had problems being out of sync sometimes. When the higher timeframe was probable of retracement, I had many false signals where I was stopped out, but I'd have a good signal after that. I just had too many unnecessary losses. Using higher TF can be a good tool, but I wasn't using it the right way.

That's what I like about mechanical systems. No need to think while trading, so you can think about new ideas.

That's why I too decided to go mechanical. I had a good plan on paper but most of the time tried to tinker with it as the day or trade progressed. Now after testing my ideas, I know what they can do. In the future, instead of micro managing a trade, I can manage the systems running on a portfolio of instruments.

What did you use for backtesting and which criterions did you use to evaluate its performance?

 

[Andy62279]

Quote from chinook:

That's why I too decided to go mechanical. I had a good plan on paper but most of the time tried to tinker with it as the day or trade progressed. Now after testing my ideas, I know what they can do. In the future, instead of micro managing a trade, I can manage the systems running on a portfolio of instruments.

What did you use for backtesting and which criterions did you use to evaluate its performance?

I have tradestation, but I don't know how to program... I backtest by hand. A long tedious process, but I get to train my eyes backtesting this way.

I look at the same stuff as you. I've heard of people achieving 80% accuracy with the setup I am using. I only achieved 64% Also, IMO accuracy isn't very important. I am the type of trader that can have 50 pip unrealized and get a breakeven trade and not care. I aim for 100+ 200 + even 300 + pip trades. So reward/risk ratio is very important for me. With two years of data I figured out the performance. reward/risk was 2.85 but as I started testing 2004 volatility really picked up, which is a good thing.

Another figure that is important to me is max consecutive losses. This helps me figure out my money management settings. The system's max cons loss is 5, but I risk more than most people's comfort level. This is fine because I have good reward/risk ratio.

expectancy, reward/risk ratio, and cons max drawdown are important numbers for me. I don't remember what the expectancy was. It has very good performance, but I also had a lot of help along the way.

time to run errands. Gotta go to DMV.. someone hit my car again Someone backed into my car a few months ago, and last friday some lady spun out of control and hit a bunch of cars.

Andy

 

[Andy62279]

oh one more criteria when I test. I test it in up, down, and sideways/choppy market conditions. If the method doesn't make money in unfavorable market conditions (sideways) then it's not worth my time.

 

[chinook]

I see, you're doing it all fashioned way. Some say that this gives you a better feel for your system. TS language is not hard to learn but it has some weird nuances. Best way to learn it is to look at the canned indicators/systems EL code.

I also don't care about the absolute accuracy. Important quantity is expectancy. I find out that if the accuracy goes up Win/Loss goes down. Actually, I feel more comfortable with lower accuracy and higher Win/Loss.

Good luck with your car. Are you at an icy city?

 

[illiquid]

Quote from chinook:

Just kidding-- I'm gone mechanical.

Sounds great, definitely keep us posted on how it progresses.

How closely do you need to monitor your system? When you see it completed, does it scale-in, adjust stops, take partials etc completely on its own, or do you need to observe and adjust manually? The idea of complete automation is probably impossible for my trading but I can definitely appreciate the advantages.

 

[chinook]

Quote from illiquid:

Sounds great, definitely keep us posted on how it progresses.

How closely do you need to monitor your system? When you see it completed, does it scale-in, adjust stops, take partials etc completely on its own, or do you need to observe and adjust manually? The idea of complete automation is probably impossible for my trading but I can definitely appreciate the advantages.

For now, I need to monitor it closely because I need to enter the orders manually--I use Bracket Trader so OCA order is ready to go anyway, I just need to adjust the trailing stop once in a while. But it only trades during the US session, so I don't have baby sit it all day and night long. After a week or so, I might semi-automate it by making Bracket Trader read a text file and send OCA orders automatically. Then I just need to manually trail the stop.

Probably in about two months, I can go full automation with the newer software versions. Then it should be able to enter, adjust stops and scale in. Partial executions might be a problem but I think I'll have Tradestation EL functions available by then to communicate with TWS.

The tested results look promising but markets change. I just need EURUSD to have similar daily volatility. Since the system is quite intuitive and not tweaked to weird parameters, I think it'll be stable and do better than breakeven.

 

[Andy62279]

Quote from chinook:

I see, you're doing it all fashioned way. Some say that this gives you a better feel for your system. TS language is not hard to learn but it has some weird nuances. Best way to learn it is to look at the canned indicators/systems EL code.

I also don't care about the absolute accuracy. Important quantity is expectancy. I find out that if the accuracy goes up Win/Loss goes down. Actually, I feel more comfortable with lower accuracy and higher Win/Loss.

Good luck with your car. Are you at an icy city?

Sorry for the wait... I didn't even go to the dmv today. Will do that tomorrow instead.

Yes, I do it the old fashioned way. It took me a month to understand and recognize the setup. Once I understood it then I had to come up with an entry method. I tested a few different entry methods, and testing by hand took a very long time. I have a few documents about TS easy language. That will be my next task to begin studying.

I have also found similar results as you. I could have achieved around 80% accuracy, but a lot of the big moves occured from less reliable setups. Hmm, that was an easy decision.

No it is not icy where I live. I live in the suburbs of portland, oregon. Last winter there was a huge winter storm, but in most cases it just rains here. The lady that caused the accident was going at least 50 on a 25 zone... during rush hour :eek: From witness testimonys, she was impatient from waiting so she went around traffic. Once she hit the train tracks she lost grip of her rear end (mercedes are rear wheel drive) and she did not let go of the accelerator during the whole spin. She bounced off a telephone pole and started headed towards the opposing lane. I turned into that road at that time. I did not have time to react. Her front end rammed into a driver side of a car, and that car hit me.

My friends think I should sue her... but I don't believe in that crap. Yea she messed up from being impatient, but it's not like she intentionally wanted to hurt anyone. I'm not physically hurt... I just want my car fixed. Damage is not too bad. I will have a rental car and imagine all the hand brake corners I'll be doing with that car, haha