Online Stock Trading

This article is part III of a three-part series.

Here is Part I

Here is Part II

Now comes the really fun part. Now we’re going to take all the information we have and translate it into something useful: now we’re going to pick some stocks. I think it’s important to keep in mind what it is we’re doing here. We’re surveying a market where many of the normal rules don’t apply, and by examining it, we’re trying to answer this one question: which rules do apply? We’re divorcing our instincts and wearing our scientists’ uniforms right now. So far, we’ve been looking at what the market has actually been doing, not what we think it should be doing, but scientifically, what it’s really been up to.

After having looked at various graphs which show us what the market is doing, or more precisely, how fundamental stock metrics vary across the various yields the market has had, let’s summarize the data a slightly different way:

What I’ve done here is split up all the useable stock yields into “groups”, each representing 10% of the market. Group 1 is the 10% worst performing stocks, group 10 is the stocks in the top 10%. The various statistics (I’ve used 11 for sake of simplicity) are the average values for those stocks in each respective group. For instance, the average yield of the top 10% of stocks was 40.04% since the credit crunch, and they also have an average market cap of 8.9 billion dollars.

Now then, just because the numbers there are the averages doesn’t mean any stocks actually fit those criteria. For that we’re going to have to test them to find out. Below the groups I’ve set up initial ranges we can use to test with (initially I picked plus or minus 10%). We’re going to use this to do some rigorous back testing. Next, I’ve polled the July dataset to see if any stocks actually met the criteria. What we’re doing is, after having looked at the various groupings of yields, and seen which statistics can be associated with the highest yields, we’re reversing course and going about it the other way now. Just because the average statistics for the various buckets are what they are, doesn’t necessarily mean anything.

It turns out that no stocks met all 11 traits in the ranges we set up (+/- 10%) for group 10. In fact, with the tight ranges we picked, the most number of traits that any stock met was only four. In other words, no stock out there had more than 4 fundamental statistics which fell within plus or minus 10% of the averages of top 10% performing stocks. Six stocks made it:

On balance, these stocks have gained 9.23% since the end of July. Not too bad. Considering that the S & P 500 has lost over three percent:

Now what we can do is relax the tight ranges we set up initially, to see if we can get some stocks that have met more than 4 of the statistical averages from the 10% best performing stocks. If we relax the range to +/- 15% of the averages for the best performing stocks, we get 4 stocks that now meet 5 of the traits, and many more that meet only 4:

The average yield for those 4 stocks is still very high. Let’s relax again, this time to +/- 20%:

While we don’t get any stocks that have more than 5 traits within 20% of the average of the best “grouping” we have more stocks that have those 5, 12 now. What’s good about this is that the diversity gained reduces risk, and, even so, the average yield went up significantly. Now the average yield is almost 12% in the worst 5 months we’ve had in a long while. Outstanding. In theory, we could have placed 8% stop loss limits on all of them and not endured the losses of 3 of the 5 that have fallen. Now that we’ve done all this “past looking” let’s look to the future. It will be interesting to see if any of this will be validated going forward.

What I’ve done is taken historical data (July through December), took averages of that data, then picked stocks that met the averages of that data. Now let’s do some similar screening for the stock data as it was at the end of December, and see what happens. The good thing is that we can have some initial results immediately (there was a lag between my collecting and analyzing the data and doing all this writing). Let’s look at how our ranges are now:

When we take these ranges and apply them to the market data from December 22nd, we actually get 1 stock that meets 6 of the traits, and 13 stocks that meet 5:

Now, I’ve taken these stocks at their December 22nd prices and put them in a Yahoo portfolio so we can track them on the fly:

It’s definitely holding its ground. In the same timeframe the S & P 500 has lost 1.08%, whereas ours have lost 0.79%, so they’re outperforming the S & P 500 by 0.29% in just a week. You can come visit this portfolio picture as often as you’d like to see how it performs (I’ll update the picture every evening).

For those who want to download the graphs I used from Parts I & II you can get them in an Excel file here, or as the actual picture files here.