6th Degree Poly Help! - page 5

New comment
 
I still think you will find the line is not even close to representing its actual position on each bar when that bar was the current bar.
 
dennisj2:


My testing methodology has always been to thoroughly execute as many backtests as it takes until I am confident. Then, I run both a demo and a live account side-by-side and compare the results. As soon as I am ready with backtest results, I'll post them here -  if you are interested - I'll keep you posted every step of the go live. And, if any one else is interested, I'll post the read only password to the demo account here (is this allowed?).  I expect the first week to be slow - the EA collects a lot of data before actively engaging the market - however, once the initial data collection is complete - it actively trades and will almost always have open positions based on the overall trend.

My EA now does a meager 5% a week. In the event I take a loss, I successfully dollar-cost-average out around 80% of the time. With this new enhancement, I am expecting at least 12-15% per week, if not more.


Basic backtests are for debugging. But, you can get a true backtest that will give realistic results if you are using tick data from your broker and tick data suit.
Also, Alpari has identical demo and live, as far as i have seen. Perhaps also other brokers, but i can only confirm this for Alpari.  So EAs can be verified rather well these days.

I am highly interested in results, so please keep me in loop. Fell free to send me PM if necessary. You can also create a signal from demo account. 

Numbers sound great, of course they depend on single transaction risk and RRR.

Remember, the poly6 is not an entry/exit indicator - but a trend analyser; I use Fibonacci and pattern recognition off of the poly6 to determine the strength of the trend and a completely separate trigger for entry/exits. This outlines the issue with my prior attempts - using an MA whether smoothed, logarithmic, exponential - gave too many directional changes. Poly6 smooths this data out and removes the noise. Put simply, the MAs have now been degaussed - (grazi-gaussed!).

I am interested to hear more about this, of course. I am aware that you are using it as trend analyser, i had it in mind when i was testing it out. But as i have other approach i probably didn't adapt to it well.

Regarding MA, SMA gives best results for me, on OPEN, CLOSE and WEIGHTED applied price, depending on usage case. Although i am using derivative values as input.

 

This is the 3rd degree poly. I modified i_regr to show the true progression of the line on live prices by making it place objects at the zero index.

I also coded a moving version of the algorithm. It's line is following the same path as the i_regr objects.

poly

Files:
i_regr_obj_1.mq4  10 kb
moving_i_regr_3.mq4  11 kb
 
OK, now this curve can be compared to MA to see the point of all this
 
graziani:
OK, now this curve can be compared to MA to see the point of all this


I had to modify the algorithm a little bit to create the moving version I think it is correct I tested it against the objects on 3rd and 6th degree it looks about right but anyone is welcome to check it for accuracy.


I was not able to replicate the smooth deviations using the std deviation algorithm, I think because he creates the entire line at once. If i code std deviation to the moving version it will look something like Bollinger Bands. We might have to do it a different way.

 
SDC:


I had to modify the algorithm a little bit to create the moving version I think it is correct I tested it against the objects on 3rd and 6th degree it looks about right but anyone is welcome to check it for accuracy.


I was not able to replicate the smooth deviations using the std deviation algorithm, I think because he creates the entire line at once. If i code std deviation to the moving version it will look something like Bollinger Bands. We might have to do it a different way.


This indicator is brilliant - the math was the hard part - check this out guys. On the yen, for last week - unfortunately, all of the other pairs I trade show two down-trending lines ("Sell Sell Sell!") Here's the indicator attached. Try this one on for size. Just look at how smooth and noise free the lines are. Makes coding the EA so much easier...

Files:
regrma-v1.ex4  15 kb
 

aha, your smoothing it with two MAs, just like e.g. MACD

unfortunately i am out of time this days, but for the weekend i guess i will catch some time for experiments. 

 
dennisj2:


This indicator is brilliant - the math was the hard part - check this out guys. On the yen, for last week - unfortunately, all of the other pairs I trade show two down-trending lines ("Sell Sell Sell!") Here's the indicator attached. Try this one on for size. Just look at how smooth and noise free the lines are. Makes coding the EA so much easier...


You should post mq4 not ex4.

OK guys I have modified the original algorithm again in a different way this time, I think this is a useful analytical tool to determine the validity of the poly line's historical signals. This version has an extra input parameter "Pos"  It will draw the line exactly as it was when the bar number entered into the Pos parameter was the current bar.

Original i_regr: Green

New Positional_i_regr: Blue



and here with the purple moving demonstrates it is accurate to the blue positional


Files:
positional_i_regr.mq4  12 kb
 
z
SDC:
 

New Positional_i_regr: Blue

and here with the purple moving demonstrates it is accurate to the blue positional


The blue line is positional? Does this mean real time? Let me take a look at what you got -
 
dennisj2:
z
The blue line is positional? Does this mean real time? Let me take a look at what you got -

Yes you can place it at any section of the chart by bar number to see how it would have looked at that time, it will continue to update in real time from that position on the chart.
123456
New comment
Reason: