Are 3 year track records meaningful? This is the blog post I referenced in the quote above. This post doesn't give a warm and fuzzy feeling about diversified programs that have Sharpe ratios < 0.5. After watching the video recommended by bassramy ...

Diversified Trend Trading Approach

Hi Copernicus, With regards to the paper, I had similar thoughts to yours. On the subject of overfitting, I too thought that it would be prevalent among the users, and it isn't surprising that there is a correlation between the amount of backtests ...

Diversified Trend Trading Approach

j4d, since you mentioned Quantopian I wanted to cite an interesting paper just published by some of its principals that approaches the large number of strategies backtested by users and the subsequent out of sample performance of those strategies. ...

Diversified Trend Trading Approach

In theory this is true - accuracy on the training data should go up the larger the number used. But you will also get overfitting this way no matter what tool you use (GP, NN, regression, ELM). So basically this is more a critique on a flawed model ...

Machine Learning with algoTraderJo

It is VB6 - the old classic, perfect VB.

Accumulative Profits Grid System

Thanks Remon, now you're talking my language!!! Great concept and beautiful code. I appreciate your sharing it very much! I did some tests with EUR/JPY 10 pip data converted to 0,1 (down,up) moves. Testing 1000 bars and using all possible parameter ...

Accumulative Profits Grid System

I did a quick prototype of the strategy and came up with the following that is fairly true to the rules in video #2 as I understand them. Note: no trading costs were deducted and it was assumed that each fill was on the grid perfectly rather than ...

Accumulative Profits Grid System

Some of you might find this (fairly detailed) blog post interesting: Machine learning for financial prediction

Machine Learning with algoTraderJo

The contract size of M6E is 12,500 so $30,000 in the account is plenty - for 2 lots your leverage is 25/30 or less than 1. The NAV of the account (includes floating gains/losses) will determine whether you are above the 25K level for PDT, so if you ...

Interactive Brokers - experiences and opinions

I also haven't had much luck with linear models with this data. I think the short swings could probably be successfully modeled with a linear model, but not the longer swings. The data also isn't very autocorrelation friendly, probably due to the ...

New historical Forex models

What makes you think an ARIMA(2,0,0) model is appropriate for this data?

New historical Forex models

Yes I have ranges and bar counts for each swing, so calculating standard deviation is no problem. For the SL are you thinking set it at 2-3 deviations? As GoldTheHun alluded, there is also the objective measure of comparison to prior swing. If the ...

New historical Forex models

To clarify this further: p[-1] >= p[-length(p)] Where p[-1] is the current prediction and p[-length(p)] represents the previous prediction made by the algorithm. It's R (your favorite language) not math syntax.

New historical Forex models

I feel for you but don't think I or anyone else can save you from disaster. Only the person who put you in disaster can do that. Many of us (including me) have been there and done that. There is life after this, never forget. If manual trading isn't ...

One Way to Develop a Winning System

Hi, why not just post here in the thread and if it is of interest I'll respond here. If it really is private I likely won't be able to help anyway. PM is off due to lack of time.

One Way to Develop a Winning System

p is the predicted result and o is the observed result. Yes, not predicting up/down because that's simply alternating. Predicting whether the next swing will be larger / smaller than the current swing in absolute range. There may be some difficulty ...

New historical Forex models

Yes there is always hope, but what about faith and charity? This thread was started in 2007 and while I think the fundamentals for single system creation are still valid, I'm not convinced anymore that a single strategy can give a trader consistent ...

One Way to Develop a Winning System

I guess I changed the definition of the problem from regression to classification in measurement. I noticed that the magnitude of the predictions was not that accurate (lots of error) but the direction was fairly good. The regression attempts to ...

New historical Forex models

GoldTheHun, thanks for posting the picture and this description. Yes, even though the formula has normal distribution characteristics, when using regression the best prediction accuracy I achieved was 67%. This probably isn't enough to overcome ...

New historical Forex models

I've been meaning to reply on this great thread but haven't had time. I wish I had some deep, new insights on the subject of diversification but the subject has already been covered in depth in this thread and in a very logical manner with lots of ...

Diversified Trend Trading Approach