Menu
Sister Sites

Options

Bookmark Thread

First Page First Unread Last Page Last Post

Print Thread

Similar Threads

Digital System Tuning & Optimization 4 replies

Revolutionary, trading system 2 replies

Grid system based on optimization and statistics 0 replies

Back testing and Optimization 3 replies

What's so revolutionary about pin bars? 12 replies

Revolutionary ideas regarding the market and system optimization

  • Post #1
  • Quote
  • First Post: Edited at 3:06am Apr 14, 2007 2:31am | Edited at 3:06am
Hey let me just say somethings that I've been thinking the last few months and see what everyone's idea is of them.

~ The market is made up of huge waves, shaped like sine or cosine waves, that last 2-9 days.

 ~These waves are slowly but constantly changing shape and size.

~Since these waves are constantly changing shape and size, fitting a system to several years worth of waves is completely counter-productive.

~It's much more productive to optimized a system to the current shape and size of the currencies wave.

~According to my research, the optimal length of time to optimize a system to the currency wave is just 4 months.

~This means that if compared with the market four months ago, the market now would look very different; and indeed it does.

~This is because 4 months is the time it takes the market wave to completely change shape.

~This is also why it's very very difficult to optimize a system over years of market data. It's like trying to find only one change of clothes to wear the whole year, for every season and event.

One thing to note is all this research was done on the hourly GBP/USD. But from glances at least other currencies, the principles seem to hold up. I believe the principle is there on other timeframes, it's just much harder to see.
Quoting aicccia
Hey let me just say somethings that I've been thinking the last few months and see what everyone's idea is of them.

~ The market is made up of huge waves, shaped like sine or cosine waves, that last 2-9 days.

 ~These waves are slowly but constantly changing shape and size.

~Since these waves are constantly changing shape and size, fitting a system to several years worth of waves is completely counter-productive.

~It's much more productive to optimized a system to the current shape and size of the currencies wave.

~According to my research, the optimal length of time to optimize a system to the currency wave is just 4 months.

~This means that if compared with the market four months ago, the market now would look very different; and indeed it does.

~This is because 4 months is the time it takes the market wave to completely change shape.

~This is also why it's very very difficult to optimize a system over years of market data. It's like trying drive a boat straight through waves, instead of sailing over and moving with them.

One thing to note is all this research was done on the hourly GBP/USD. But from glances at least other currencies, the principles seem to hold up. I believe the principle is there on other timeframes, it's just much harder to see.
No offense, but I'm used to powering through the waves in the fountain. I guess we all grow up differently.

However, I do agree that form-fitting a particular method is counter-productive. This is why I love discretionary techniques, take some money from the market pre-empting the procedings.
Never underestimate the powers of the Force.
Quoting The Jedi
No offense, but I'm used to powering through the waves in the fountain. I guess we all grow up differently.

However, I do agree that form-fitting a particular method is counter-productive. This is why I love discretionary techniques, take some money from the market pre-empting the procedings.
it was a bad analogy, I'm changing it.
Quoting aicccia

~This is also why it's very very difficult to optimize a system over years of market data. It's like trying to find only one change of clothes to wear the whole year, for every season and event.
I have to say that I also disagree on this new analogy and also on the general idea that is behind it.
First the analogy: I would say that weather change much more slowly compare to Forex (except maybe for some special events like hurricanes or similar things). Imagine that the temperature could move for +20 to -20 degrees Celsius in a matter of hours. What coat will you take. A coat for the "average condition" or just a coat for the current conditions? Me I will take the first choice.
Second the idea: optimising with a short recent period is in my opinion too much curve fitting. Phoenix is a good example of that. His creator Hendrick had the same opinion as you have now about fitting the EA on a short recent time frame, and after some time of testing he admitted that it was too much curve fitting and not robust enough.The idea is not so much to optimize a system over years of data but to have a robust system. For instance a system that will work very well under certain market conditions (like trending) and will resist (loose not too much) under opoosite conditions (ranging).
I believe that the market can behave the same way for several years. Intermediate trends change dramatically every couple of months, however, the overall trend major trend change once in a few years. Look at the major trend of most currency pairs, for the past couple of years, the USD major trend is down against almost all currencies, even against the Yen...All the small waves you are talking about are just intermediate corrections caused by hesitating shift between supply and demand, but this doesn't mean that a major shift in supply and demand occurred.

Remember, Warren Buffet 2 years ago, said that he is extremely bearish the USD and bullish the GBP and everybody used to laugh at him. Now, he is the one who is laughing at the whole world. He managed to see and spot the major shift in supply and demand for the GBP and the USD, he was looking at the big picture, while everyone else was looking for the intermediate small trend to scalp a few hundred pips. He traded the major trend, while everyone else traded the intermediates.


Thanks,

Nader
Quoting aicccia
~ The market is made up of huge waves, shaped like sine or cosine waves, that last 2-9 days.
I am very curious aicccia, do you have any evidence to substantiate that?



 ~Since these waves are constantly changing shape and size, fitting a system to several years worth of waves is completely counter-productive.
How many years of data did your study involve?


~According to my research, the optimal length of time to optimize a system to the currency wave is just 4 months.
How did you come to such a conclusion?

~This is because 4 months is the time it takes the market wave to completely change shape.
Very interesting! How did you find that out.

~This is also why it's very very difficult to optimize a system over years of market data.
That's what quant analysis is about...
Thanks.
Quoting jlpi
Imagine that the temperature could move for +20 to -20 degrees Celsius in a matter of hours. What coat will you take.
Obviously you never lived in New York.
Quoting irusoh
Obviously you never lived in New York.
LOL. In Oregon we have a saying; "Don't like the weather? Wait TEN minutes!" Today we woke up to sunshine, now at 9:30am the rain clouds are overhead.

On the idea of behavior, i tend to see things as Nader has writen. Even though i'm the guy "trying to scalp 100 pips", i have forced myself to take time to look at (eeeegaaads) weekly and even monthly charts to see the big picture, thinking the added knowledge will assist in intraday trades off my favorite 15m chart.

: ) Thom
Quoting jlpi
I have to say that I also disagree on this new analogy and also on the general idea that is behind it.
First the analogy: I would say that weather change much more slowly compare to Forex (except maybe for some special events like hurricanes or similar things). Imagine that the temperature could move for +20 to -20 degrees Celsius in a matter of hours. What coat will you take. A coat for the "average condition" or just a coat for the current conditions? Me I will take the first choice.
Second the idea: optimising with a short recent period is in my opinion too much curve fitting. Phoenix is a good example of that. His creator Hendrick had the same opinion as you have now about fitting the EA on a short recent time frame, and after some time of testing he admitted that it was too much curve fitting and not robust enough.The idea is not so much to optimize a system over years of data but to have a robust system. For instance a system that will work very well under certain market conditions (like trending) and will resist (loose not too much) under opoosite conditions (ranging).
I'm not optimizing it to the data, I'm optimizing it to the currency wave. You can tell when something is optimized too much if when you optimize it over half a year, get a,b and c settings, and then if a week later you optimize it over the last 6 months again, and get settings completely different. That is curve fittings. Now if when you optimize it again and your settings are only slightly different, that's not curve fitting, that's fitting the system to the currency wave. The slight change is because the currency wave has slightly changed.
You said what doesn't work but little about what does work.

Optimization is bugfixing in general. You can't take it seriously. It's also fun. seeing those 100% on 40+ trades and all those fake grails.

I start to believe in scalping and short term trading more and more every day.

But I like your thinking. All you said is true.
Quoting narafa
Remember, Warren Buffet 2 years ago, said that he is extremely bearish the USD and bullish the GBP and everybody used to laugh at him.
I'm more then sure that WB knew some data that we didn't knew. At that level you get all sorts of info plus a life time experience of trading to sort that info

Quote
~This is because 4 months is the time it takes the market wave to completely change shape.
This is intresting. My oppinion is that fundies creates major trends(direction) even if i'm a chartist.

I like your ideas, but I don't know why they lead my to see a Neuronal Network...
"Abandon all hope, you who enter here" La Divina Commedia, Dante Alighieri
Quoting aicccia
it was a bad analogy, I'm changing it.
Oh mate, I didn't mean it as a cutting comment. I was just being cheeky.

Also, to everyone else who's scrutinizing this man's every word. I get the sincere notion that his conclusion is one of those that you get not from research and experimentation, but from the experience of so many other failed experiments and research projects. And for that, I can really connect with where he's coming from. I think we all have that one moment, that one epiphany to ourselves where it all just clicks for us. What clicked for me surely isn't anywhere near the same as what all fit together for Aicccia (I know, because he told us his epiphany and I know it's not mine), but this is what works for him and you just have to find your own instead of trying to form fit someone else's to be your own.

You want to succeed in trading? You have to look at yourself and your personality and develop (not find, actually develop) a methodology and trading style that fits you like a glove that you want to work for and have work for you. Yes, you need to take the time out to read what other people have said, but don't try and deflate their opinions. Instead, pick it apart and look at what makes sense to you and try and figure out why what doesn't fit with you doesn't fit so you can know for your self.

You'll get alot further trying to understand why you should credit someone else than discredit them.
Never underestimate the powers of the Force.
Well these ideas just slowly formed after spending several months looking at long 60-100preriod+ MAs on the hourly charts. So I developed a trading system that relied on their existence and their principle to work. And the system works great, so I'm taking that as evidence that I'm on to something. I was just wondering if anyone had surmised anything similar about the markets.
That's interesting. I have always wondered if it would be worth using a dynamic ARMA when trading (I remember having to construct it when studying advanced econometrics at uni)... has anybody tried this? It would be interesting to find out.
Quoting CHECKstar
That's interesting. I have always wondered if it would be worth using a dynamic ARMA when trading (I remember having to construct it when studying advanced econometrics at uni)... has anybody tried this? It would be interesting to find out.
What is a dynamic ARMA?
ARMA; Auto Regressive Moving Average. It is a forecasting technique used in time series data.
Quoting aicccia
I'm not optimizing it to the data, I'm optimizing it to the currency wave. You can tell when something is optimized too much if when you optimize it over half a year, get a,b and c settings, and then if a week later you optimize it over the last 6 months again, and get settings completely different. That is curve fittings. Now if when you optimize it again and your settings are only slightly different, that's not curve fitting, that's fitting the system to the currency wave. The slight change is because the currency wave has slightly changed.
If your system is robust and if you just slightly change your settings, you should only get slightly different results I think. So is it really significant?
Quoting CHECKstar
That's interesting. I have always wondered if it would be worth using a dynamic ARMA when trading (I remember having to construct it when studying advanced econometrics at uni)... has anybody tried this? It would be interesting to find out.
Yes of course some people tried it as it is a common modelisation technic.
I even remember having seen results in a book (but I don't remember which book... maybe from Perry Kaufman).
I don't remember exactly the results but in my opinion it could be too much curve fitting if you take high orders ARMA and not so well adapted to non linear events like the big news.
Quoting jlpi
If your system is robust and if you just slightly change your settings, you should only get slightly different results I think. So is it really significant?
I'm saying, if a system is really just curve-fitting randomness, then changing the period just slightly that you optimize it over will give you completely different settings.

If it's robust, you can optimize the system over several different periods and still get similar settings.

That it how you tell if a system is curve fitting. To say a system is curvefitted simply because the data sample was months instead of years is inaccurate.
Quoting aicccia
If it's robust, you can optimize the system over several different periods and still get similar settings.
Yes exactly.
That's why it is in my opinion not useful to try to fine tune your system very frequently.
0 traders viewing now