Yes your argument makes sense, so perhaps the reason is deeper than just scaling. After reading The Number 1 Novice Quant Mistake (not my blog) it appears that using prices instead of returns also may lead to a misinterpretation of the underlying ...

Machine Learning with algoTraderJo

algoTraderJo, many thanks for all your comments and insights! This has been a very fulfilling thread for me. Have a great holiday and see you back next year! I'm looking forward to the next chapter!

Machine Learning with algoTraderJo

PipMeUp, in addition to what algoTraderJo stated on the subject, please see Feature Scaling for a bit more background and some additional ways to scale features. Doing the analysis on returns is probably the easiest and possibly the safest. Also, ...

Machine Learning with algoTraderJo

stt, I'm a huge fan of Marcos Lopez de Prado. I've read many of his papers but don't think I've seen this one, so thanks for the citation! The paper seems to consider a 'large amount of data' as time, as shown in Figure 2 rather than either bars or ...

Machine Learning with algoTraderJo

algoTraderJo: Many thanks for all your comments. This has helped me clarify several issues I've been working with over the past few years. I've used Markowitz portfolio optimization in the past. My results may have suffered because 1) I wasn't using ...

Machine Learning with algoTraderJo

So now I got around to testing EU with lm and then princomp in R for comparison purposes as a follow up of the knn model I previously created. I haven't used exactly the same settings because my daily data is different, so it won't match anyway. The ...

Machine Learning with algoTraderJo

Here is my attempt to make a rough reproduction of algoTraderJo's post referenced in the quote above, using R (the simple knn model with a window length of 70 for training). In this simulation, I don't use a stop loss, so this is an always-in ...

Machine Learning with algoTraderJo

Hi algoTraderJo, I have a question about how you prepared the data for training, specifically the class label or actual outcome that is used in training. As you prepare your data, are you assuming that future price (20 bars) hits the +/- ATR20 * 0.5 ...

Machine Learning with algoTraderJo

In addition to the excellent and highly recommended course by Andrew Ng on Machine Learning at Coursera cited by stt, the course Statistical Learning, (also offered by Stanford) was extremely good and offers an introduction to supervised learning ...

Machine Learning with algoTraderJo

I admit, I am puzzled by this response. I had hopes that we could have an intelligent exchange of ideas in this thread. I was mistaken. I won't make that mistake again. Best regards, FXEZ

The Great Pumpkin

I'm late to the debate but just wanted to add that if you perform time series analysis using the mean and standard deviation, you are implicitly assuming that the distribution is normal. This is the case with parametric statistical measures. The ...

The Great Pumpkin

Hi jimsterk, thanks for posting your code. It has helped me debug my code. I have worked via a slightly different algorithm a comparison with your numbers. My data isn't the same as your data so there will be differences. Here is what I got for EU ...

The similarity system

I'm very interested in figuring this out, happy to communicate privately (your PM is off).

The similarity system

I have coded up my interpretation of this "Decreasing-increasing sequence in set theorem", problem 1, theorem 1 in R. I am not getting the same results (93.75%). I'm getting ~83% depending on how many trials I use. Here is my code. Can anyone see ...

The similarity system

You were wondering why the number of trades is so low. Are you aware that when you have a TP or SL of 200, based on $1000 starting account and 0.1 lot size, that it is equivalent to a 2000 pip TP or SL? Also, your entries are NOT random. They are ...

Random Trading EA with success

Hi ForexQuant, long time no see! Is it Trend Spotting?

Looking for an old thread: measuring the strength of a trend

@canada, To get that information there is no getting around wading into the inner logic and calculations of the zig zag. Please see the Zig Zag is not Zig Zag thread for one of the best in-depth looks at zig zag that I've seen anywhere, including ...

Can't understand how the zig zag indicator works!

Pipologoist, See url Sterling Ratio. Edit: Sorry about that, link updated.

Alternative Reward to Risk Ratio for Trading System Statistics

Madmoney, when I started out I was having trouble with generalizability and I didn't have any way to deal with patterns that aren't in SET2, so more data helped reduce unfound patterns. But now that I've improved my pattern matching (logic1), this ...

Fractals, ZZ & the Pissing Dog

Madmoney, I'm using 2011-2013 at the moment (3 years) for my in sample period. I have high quality data outside MT4 that I use to create the database. I'm working on adaptability now.

Fractals, ZZ & the Pissing Dog