Yes it is very easy to be fooled by randomness. Have you got any random walk charts you could post in candlestick/bar format, I would be very interested in viewing these?
As I mentioned earlier, I was performing Fuzz testing on my neural network. I found several bugs, and work for the future. The investment of time into PI has already paid off
neural networks do not work in financial trading, but I thought you knew that. Not one single successful quant hedge fund (including Rentech) applies neural networks to its trading signal and order generation. Welcome to the world of data snooping bias. What makes you believe that you with an apparent background on the IT side rather than financial side beat the slim odds to successfully apply neural networks to quant trading when even those with decades of experience in micro market dynamics, risk management, and those with expert knowledge of mathematics and statistics failed? Well, do as you like but I am absolutely certain that you will leave this endeavour disappointed and without tangible results, other than losses after applying transaction costs.
Thanks for your input volpunter. I can see exactly who you think that I am, and while interesting as an empathy experiment, unfortunately, you assume too little of others. I'm glad that all the kings horses, and all the kings men, couldn't put two and two together. It certainly means that my profitability will continue up until the singularity.
When I used empathy to understand your point of view, it seems to have had a knock-on effect, here's my perspective of you, correct my assumptions: You are a discretionary trader, likely for 12-15 years. You are afraid that the use of computers will eventually erode your manual edge. You have looked on the internet, and having found that people have had problems with neural networks in the past, smugly confirmed your bias. You can now sit back and relax, with little fear that your system will erode. What's more, it seems you run sneak attacks on others from this website. Attack the personality of another while retreating yourself. I'm not sure why, but it is interesting to watch. Can you post some more so that I can run statistical analysis on your posting? (Oh wait, 10,000 posts is a good sample size)
Good point. I guess I should have said pseudorandom, as you can generate the same sequence time and time again. The hypothesis is that given a PI sequence of (n) digits, you can't calculate the next digit. Here is a nice article, in which they also graph PI in a different manner.