Page 1 of 1

Selecting a Portfolio

Posted: Tue Jun 10, 2014 8:47 pm
by Ghost11365
I was looking to get some direction/thoughts on creating an equities portfolio geared toward a long term trend following strategy.
I'm assuming there is some sort of rhyme and reason to selecting stocks to include in such a basket...or is there? Does it make most
sense to just use the S&P 500 constituents or should I be selecting say the top 10 companies, per sector based on market cap? Any thoughts
or experience anyone has with trend following relating to equities would be greatly appreciated.

Posted: Wed Jun 11, 2014 9:25 am
by sluggo
The usual thought process is to decide how you will modify the portfolio in the future, loooong after backtesting is completed, and real-life, real-money trading is well underway. Then build your backtesting portfolio using that same methodology.

For example, many people believe that some day in the future, Tesla Motors will be added to the S&P 500. Let's suppose that on March 1st 2016, Consolidated Amalgamated Heavy Industries is dropped from the S&P 500 and Tesla Motors is added. How will you trade? Will you continue to trade Cons.Ind and to not trade Tesla.Mot? Will you stop trading Cons.Ind and start trading Tesla.Mot? Will you trade both? Neither?

If you decide to add Tesla Motors to your portfolio on March 2nd 2016, what will you do if your technical indicators say that a long trade was hypothetically entered back in 2015? Will you enter the long trade on March 2nd (months and months after the original entry signal)? Will you ignore this first trade but take all subsequent entry signals? Will you enter the long trade on March 2nd, but only at half-size? etc. Think about it.

When you've decided how you will trade in the future, simulate trading that way in the past.

Posted: Wed Jun 11, 2014 9:52 am
by Ghost11365
Hi Sluggo - appreciate the reply. Today my approach is I've created the s&p 500 using all constituents from the past ten years with a filter that says ignore security X if it was before Y date (any date prior to bring added to S&P. But is using the S&P a good/practice/logical approach? Is it better to limit my portfolio to a smaller population? Thanks for feedback and thoughts.

Posted: Wed Jun 11, 2014 10:07 am
by fab1usa1
Just my two cents, but if the S&P 500 appeals to you then why not develop a system that trades SPY or ES? That way you can concentrate on your entry and exit logic rather than focusing on concerns of what stock comprised the index at any given time. Once you develop a winning system having characteristics that you can stomach then focus on position sizing to reach your goals.

Posted: Wed Jun 11, 2014 10:31 am
by AFJ Garner
Backtesting current S&P 500 constituents is not a great idea. A better idea would be to test all constituents past and present, listed or de-listed. And as Sluggo says decide if you add and then drop stocks as they come in and then drop out of the index.

Another point to consider is position sizing. The S&P 500 is an index based on market cap. Using similar indexing methodology would give similar performance to the S&P which itself is a type of mechanical trend following system.

I like indexing but prefer a different methodology to the S&P 500. Using past and present constituents you can devise different indexing methodologies which can dial up and down performance. One system I have coded produces around 33% CAGR in backtesting for a high vol and high DD. Risk can be toned down to produce a volatility similar to the S&P 500 but for a better reward.

Posted: Wed Jun 11, 2014 10:31 am
by AFJ Garner

Posted: Sun Jun 15, 2014 4:34 am
by AFJ Garner
What I do is to assess what it is I am aiming to achieve. Why should I choose the S&P 500? The S&P 400? The Russell 1000?

What am I aiming at? If I want a universe of stocks to trade what should the nature of those stocks be? Does the methodology of the S&P 500 suit my aims? If not, would the Russell 1000 or the Willshire 5000 suit me better?

I start by looking at the index methodology. Is the S&P 500 strictly quantitative? If not, how about the Russell 1000?

Do I like their index methodology or would I like to create my own?

You will find it a far from trivial task to trade only those stocks which formed a part of the chosen index AT THE TIME THE TRADE IS TAKEN.

If that is what you want to do, you will need some hard coded method of telling your test engine whether a given stock was part of a given index at the given time.

You may also like to take a look at what you data provider has in his goody bag and whether his offering suffices.

Do I want to trade only stocks which are incorporated in the US? Or which are head-quartered in the US? Or do I want to include ADRs. The list of questions continues.

Here is what CSI provides:

Code: Select all

	No of Stocks
Description	By Exchange
Albrta Count	427
AMEX Count	2190
FINDEX Count	1275
INDEX Count	2057
LSE Count	8199
MSE Count	550
MUTUAL Count	44570
NASDAQ(CM) Count	566
NASDAQ(GM) Count	798
NASDAQ(GS) Count	1456
NYSE Count	9723
OTC Count	18454
OTC BB Count	529
OTC Markets Count	580
OTC PKC Count	1611
OTC PKL Count	209
OTC PKN Count	522
OTC PNK Count	4347
OTC QBV Count	993
OTC QXE Count	78
OTC QXI Count	39
OTC QXP Count	8
OTC QXR Count	7
SWX Count	1070
TSX Count	5022
VSE Count	4567
Grand Count	109847

Posted: Sun Jun 15, 2014 5:00 am
by AFJ Garner
You may also like to ask the Trading Blox staff how they went about choosing the portfolio they provide as part of their equities package.

Which, incidentally, is downloadable direct from CSI.