Julian’s Journal

[JulianCohen]

The ranking I was referring to was mainly for Mean Reversion Systems that are using the Custom Backtester Code which uses Rank to filter the signals.

Momentum systems by their very nature really have to rely on some measure of momentum (Derrr…) and that does limit you a little. I use ROC but there are Linear regression lines, ADX strength, maybe a steepness measurement for the Moving average (I just made that one up by the way…let me know if it works

I guess you have to think outside the box a bit. Sometimes reading stuff makes you think of something you can try. I read a lot of blogs, but ignore 95% of what I read, but every now and then on a podcast or maybe Cesar’s blog, or JB Marwood, I get an idea that’s worth testing.

[JulianCohen]

here’s an idea for you Mike….I have it on my list to test this week.

Try combining a number of variables and assigning them a numerical score. You can then change the score depending upon how much weight you want to attach to the variable, for example

if RSI(3) is over 70, score 100 point. If ROC(20) is over 5, score 200 point. If Close > MA(C,200) score 100 points etc.

Then you can rank according to the total score. Might be an idea to create a sliding scale of some sort within each variable, and then you will need to assign a method of ranking stocks with the same total score, but this seems like an idea worth spending countless hours on

[MichaelRodwell]

Hi Julian

Thanks for the suggestions!

I was just checking in to let you know I got some great progress using ADX. Glen P had mentioned it to me before but I couldnt figure out a way to use it. What I did was use a combination of ROC (short and long) with ADX to confirm strength. The main bang for buck I got was over 1% reduction in my drawdown. I’ll post my latest stats in my journal.

Thanks!

[JulianCohen]

Mar ’19

Short-Term Systems

US MR: 0.03%
US MOC: -3.4%
New US MOC: -4.22%

Long Term Systems

S&P 500 Momentum: 1.63%
NASDAQ Momentum: 0.56%
Long Term NASDAQ: 4.47%
US WTT: -0.66%

ASX Growth -0.19%

Total Account: -0.46%

meh!!!

[Anonymous]

Hi Julian,

In reading some of the other journals I remember you commenting that you have had success using the universe filter to improve your systems. I was wondering if you would be willing to expand on that a bit more without giving away any secret sauce! I have played around with the universe filter with really no improvement to my systems.

A couple specific questions would be the type of system that you found the universe filter help, MOC, mean reversion, rotational etc. Most of my testing was on MOC systems and anytime I reduced the exposure by excluding sectors I was weakening the system. Also most of my testing was trying to exclude sectors that I thought would be more unpredictable, such as biotech, natural resources etc… all with no improvement to my systems.

Anyway, thought I would pick your brain a bit if you are willing to share.

Thanks!

Dustin

[JulianCohen]

Happy to share Dustin, but I can’t remember where I said that ha ha.

I don’t use a universe filter any different to the ones in the course. I use a sector filter on one of my swing systems. Could it be that do you think?

[Anonymous]

Hi Julian, thanks. Yes that’s probably what I remember reading. It sounds like you exclude certain sectors only on your mean reversion system?

[JulianCohen]

yes I saw the idea on a blog post and tested the code on my systems but it was only the Swing system that it seemed to have a nice effect on. So that could well be optimisation in play as it didn’t have much effect on anything else I have. The code is below if you want to try it.

Basically the RSI of the sector has to be over the minimum you set in order for the Sector to be considered

Code:

_SECTION_BEGIN(“Sector Filter”);
// Controls for sector filter
lenSectorRSI = Param(“Sector Filter RSI Length”,90,1,200,50);
minSectorRSI = Param(“Sector Filter RSI Minimum”,50,10,90,30);
enableSectorFilter = 1;

// Map the sector ID for this symbol to an index that represents the sector
sectID = sectorID();
switch(sectID)
{
case 1: tkSector = “$SP1500M”; break; // Basic Materials
case 2: tkSector = “$SP1500D”; break; // Consumer Discretionary
case 3: tkSector = “$SP1500R”; break; // Real Estate
case 4: tkSector = “$SP1500E”; break; // Energy
case 5: tkSector = “$SP1500F”; break; // Financials
case 6: tkSector = “$SP1500A”; break; // Healthcare
case 7: tkSector = “$SP1500I”; break; // Industrials
case 8: tkSector = “$SP1500T”; break; // Information Technology
case 9: tkSector = “$SP1500L”; break; // Communication Services
case 10: tkSector = “$SP1500U”; break; // Utilities
case 11: tkSector = “$SP1500S”; break; // Consumer Staples
default: tkSector = “$SP1500″; break;
}

// Determine sector health using RSI of the sector index
pSector = Foreign(tkSector,”C”);
rsiSector = RSIa(pSector,lenSectorRSI);
_SECTION_END();

Then in your conditions of entry code add this

Code:

Cond7 = IIf(enableSectorFilter, !IsNull(rsiSector) AND rsiSector > minSectorRSI, True);

I’m not promising the code is correct so if you find an error I’d lurve to hear about it

[Anonymous]

Thanks a lot Julian. This is an interesting bit of code. I will play around with it and see what I can find!

[ThibautTenaille]

Hi Julian, thanks for posting the code. This is a great concept. I have a doubt as I tried plotting the below code to check the signals:

Plot(Close,””, colorBlack, styleBar);
PlotShapes(shapeHollowSmallUpTriangle*Cond7, colorGreen, 0,L,-100);

Assuming that the plotting code is correct ! the signals appears to be correct when using the default $SP1500 but if I change to $SP1500M the signals don’t match. I am not sure what is missing as the idea was to show signal for each sector when above the RSI>50.

[JulianCohen]

I’ll have a look Thibault. Craig do you know if you can use the *Cond7 as Thibault is trying in this case?

[Stephen James]

An exploration is probably the better way to see what is going on here.

Cond7 will just return zero for each symbol as it is switching between several sector symbols for the condition.

Part of the code is missing and that is where is it combined with the RSI value for each stock symbol in PositionScore:
PositionScore = RSI(60) + rsisector;

So it takes the RSI of the sector symbol if above the minimum and adds it to the stock RSI value to gain a ranking.

Add some exploration columns for Cond7, the stock RSI and PositionScore to see it all together and confirm with RSI values from charts of each sector and symbol.

[JulianCohen]

Sorry Chaps….I missed the most important bit

[JulianCohen]

I had a look back at my code and I had missed the positionscore code in mine too. When I retested it the results were not as good as my existing system.

So…sorry to have wasted your time if you tested it, and thanks to Dustin for bringing it up and Craig for his invaluable help.

[JulianCohen]

Anyone have this pleasure last night? VC (Visteon Corp) or as I now call it, Fisteon….

[Author]

Posts