Disliked{quote} Are you using sqlite wrapper / mysql wrapper to connect to your database? there is new feature at mt4 build 670 called Webrequest, are you using it to connect to database via webserver (ex. Using Wamp)?Ignored
ECN && STP && Scalping && Hedging - Broker Review - Oct 2009 24 replies
Alert on Fractals (MT) 2 replies
Zigzag & Fractals 10 replies
EA wanted for sound alert depending on fractals condition 2 replies
Fractals Edge Fx 3 replies
Disliked{quote} elen,For the historical building of the database, you should ......CarlIgnored
DislikedI have been away working on other things but have turned my mind again to this thread. It seems that people are having issues with the next bar prediction bar of the concept, so I thought I would start fresh also and work my way through the concept. Maybe some things I find will help others or generate new ideas. To start with, I have gone away from using time bars (eg 1min, 5min etc) and moved to using range bars. I create the range bars myself from raw tick data. From the range bars I create a big list of all the range bars, with up bars having...Ignored
Disliked{quote} That is pretty interesting and I will take a closer look at it after monday when I finally got time. yay. In your introduction you talked about the next bar prediction but you continue with talking about the direction of fractals. Just a few questions: How do you determine if a longer fractal sequence is more likely than the once we currently have as this is really the information we are after. I haven't looked at the actual probabilities but I assume that the probability of getting a 1 fractal after a 0 fractal is much higher than getting...Ignored
Disliked{quote} Lets get this stuff rolling. I found pretty good success in using svm or ann's to predict the direction of the next bars. I really like to procrastinate studying to I quickly reconfigured my neural network and surprisingly got a hit rate of 388 to 112 = 77.6% on out of sample tests which looks astonishingly good. I haven't had time to look into the data but if that is true I can stop going to university after that semester ... For the beginning http://intelligenttradingtech.blogspot.de/2010/02/practical-implementation-of-neural_04.html...Ignored
Disliked{quote} Thanks a lot Carl. After i Read and Read this thread and the old one Again and Again, Now i GOT it .... the Main POINT is to remember that Zigzag Leg must Have a Fractal at its End. This is important, as in Real Time it is Not Exist yet, but our algo Must include this in its Calculation, i think this is what YoU try to say about "Start with a Fractal and End with a Fractal" ..... The Start of a Zigzag is actually also the End of the Zigzag (the Previous Zigzag, which has a Fractal), so it MEAN that We just need to pay attention to the End...Ignored
DislikedI have just built a little chart app to visualise what my algo was doing (I don't use MT4). In the screen shot below, I have learnt the dna string for the range bar fractals from 2012 to 2014. The screen shot shows 1 week in June 2014. Green bars are up bars and red bars are down bars. The green and red arrows show the actual fractals (as used in the algo), they are draw in after the next bar has completed (because you always need an additional bar to determine if you have a fractal). The Purple bars show the predicted direction of the next fractal....Ignored
DislikedOk I'm onto the zigzag stuff now and am hoping that somebody can confirm my understanding of the concept. I have build my database of past zigzags that for each past zigzag I hold the following: direction, length and the fractals contained in the zigzag In the out of sample data (real time) I watch for fractals as they form. Let me give you an example: I have confirmed zigzag heading upward, the current length of the zigzag is 22 bars. I look in my historic collection of zigzags for all matches where the direction is the same and the zigzag length...Ignored
DislikedIf I understand you correctly you want to replace the amount of bars with "a height based value"? ...... why don't we just take the "base" value pips.Ignored
DislikedBut aren't we kind of doing that at the moment? We take the average of x bars and if a bar is 2 times as big as the average we (at least thats what I am doing) count it as 2 bars. This is a transformation from the horizontal bar count (time) to a vertical approach the pip count.Ignored