Page 1 of 1

Trading Blox Equities Portfolio

Posted: Thu Jun 19, 2014 4:14 am
by AFJ Garner
I would be very interested to hear from Tim, Jake et al just how they devised this portfolio. Would you guys care to comment?

Incidentally, see this thread at Traders Place: ... -to-trade/

In particular:
I'm too investigating this issue at the moment. I have a solution at the moment that I use for my business, but I'm looking for a low cost solution to use for a new book. My current live solution is far too expensive for normal traders and I want something solid to recommend at retail prices.

My conclusion as well is that Norgate (Premium Data) seems to be the top choice. I'm not finished with my investigation yet, but so far they seem to have the strongest offering. The prices are almost silly and can't get much closer to free. That's not the interesting part though.

I had a long talk with one of the owners last week. They have some very intersting things in the pipeline and he seemed surprisingly knowledgable about what serious quant modellers need. Their website is crap and almost made me disregard them completely, but here's what I found out after speaking to them:

They cover pretty much all US stocks, except penny stocks which don't interest me anyhow.
They also cover a few other countries, but frankly for me the choice is US or world coverage. Adding Oz and a couple of more doesn't change anything for me.
They have full dividends history, including normal cash dividends (which their site says that they don't)
They cover delisted stocks.
They have historical index constituents, funny enough implemented the exact same way that I do myself. You'll have an indicator, where you provide inputs GOOG, NDX for instance, and the indicator will show you 1 on the days when Google was part of the Nasdaq 100 and 0 when it was not.
They are right now developing a plugin to RightEdge to get historical data and index memberships.

I volunteered to test out their new RightEdge plugin and I'll report back with my findings. So far, I'm positively surprised over this company.
Personally I prefer to design my own index rather than trade say Russell 1000 stocks. But there you go.

Posted: Thu Jun 19, 2014 4:29 am
by AFJ Garner
Incidentally, just for the sake of clarity, I am NOT recommending Right Edge. I have tried it and do not like it at all.

Posted: Thu Jun 19, 2014 9:27 am
by Jake Carriker
Here is what I did:

1. Pretend I am a stock trader sitting at my desk at a date prior to the start of my system's development period.

2. Develop a set of criteria for volume and price that indicates a stock makes at least a "blip on my radar screen". That is, I am not interested in penny stocks or those that never trade any appreciable volume.

3. Attempt to get data for every US or North American stock that has ever traded since the start of my system's development period, whether that stock still trades or not. This attempt is more or less successful depending on the completeness of the CSI database.

4. Starting prior to my system's development period, roll forward in time one bar at a time noting whether the cumulative activity to date of each stock in the universe meets my price and volume criteria. If a stock does meet the criteria to ping my radar the stock symbol and date it first registers are noted in a file. No stock can trade if it is not in this list, and no stock can trade on a date prior to being "noticed" by my fictional no-hindsight trader.

5. Note that once a stock makes the list, it never comes out, even if it becomes a penny stock or stops trading. Of course, if it stops trading, no more trades can be taken in that issue.

This process is an attempt to have a hindsight free set of tradables that matches what a reasonable trader might perceive the universe of candidates to be on any given historical date.

Posted: Thu Jun 19, 2014 12:08 pm
by AFJ Garner
Thanks Jake, very nice explanation. And the portfolio tests very well. Yes, volume and price: unless we have the rolling number of shares in issue or a list of leavers and joiners in a relevant index we can't pursue the more usual index type market cap route. Which may be no bad thing.

I have more or less done the same sort of thing as you describe but so far have around 4,000 listed and de-listed NYSE, NASDAQ and AMEX stocks in the portfolio.

The decision now is ETFs and foreign stocks: include/exclude.

Fascinating to test an otherwise good momentum system on the entire CSI database. Which, amazingly, I have been able to do, just, in TB. I tested the entire 25,000 delisted US stocks in one test, ditto active Nasdaq, active NYSE, active AMEX.

Very eye opening. And sobering.

Posted: Thu Jun 19, 2014 12:16 pm
by LeviF
AFJ Garner wrote:Very eye opening. And sobering.
Care to elaborate?

Posted: Thu Jun 19, 2014 12:46 pm
by AFJ Garner
LeviF wrote:
AFJ Garner wrote:Very eye opening. And sobering.
Care to elaborate?
Yep - the equity curve went from the top left 45 degrees down to the bottom right of the graph!

Not to be recommended.

The other "whole " portfolios were a disaster too.

You have to apply a filter: using historic index stocks is a market cap filter. What Jake has done applies a volume filter.

Over history, the vast majority of stocks ever listed crashed and burnt.

Posted: Thu Jun 19, 2014 12:46 pm
by AFJ Garner