Tuesday, October 6, 2009

GGT doesn't seem to be working ... why?

If you're trying to chase the market using GGT stocks and ETFs, you've undoubtedly been disappointed at the selections. I received a private message this evening :o) "complaining" that GGT wasn't working, and what were they doing wrong?

Simply put, if you're in this category, you're not doing anything wrong.

Let's make sure we understand how GGT works.

As of today, the GGT back-testing optimization period is two years. This means that when the squirrel is running in the cage, it is grabbing the last two years of price and volume data, then it tries to find the "best" EMAs and ROCs that make that sample set work over the past two years. "Best" is my term for a number of optimizations: I can optimize on peak equity, I can optimize on Calmar Ratio, or I can optimize on plateaus. It varies, and I cycle through the optimizations every 4-8 weeks, depending on what is coming next. The values you see updated in the ETF and stock files are the best-performing values in the context of a 1-year lookback.

So, optimization looks back at 2 years, the values in the spreadsheet are compared to existing values that are established over the last year. Whatever one does the best get's the most recent slot in the appropriate ETF or stock files that I post.

The EMA range is intentionally limited from 2 days to 25 days. This means that EMAs above 25d in length are not considered. I've toyed with letting this go to 34, but for now, it's 25. If you scan down columns R-W in the spreadsheets I post you'll not see anything shorter than a 2, nor will you see anything longer than a 25.

Of particular importance are the values in columns R, S, and T. The first two are price EMAs, and the value in column T is a volume EMA. If the first two values are relatively small -- 2, 3, 5, 6, whatever -- then this ETF or stock will respond quicker to changes in price. Alternatively, if these values are 11, 17, 20, etc., then this ETF or stock will respond quite slowly.

The values in columns U-W are important -- these are Rate of Change columns. These values too influence the Long/Cash calls; if they are small in value the ETF will respond a bit faster; if they are large it responds slower.

What we're seeing right now is that stocks and ETFs are actually somewhat bound between two ranges. The drop up until last week moved us down to the lower part of a channel, and the rally the past two days has moved us upward towards the ceiling of that channel. The period (or duration) of these oscillations is relatively short -- on the order of a week or two -- and expecting ETFs with a series of EMAs/ROCs to respond when they have values longer than 10 or so is a tall order.

With respect to the "average" values, the ETF file presently has an average value of 9.23 days, so we're going to see some delay. By the time some of the GGT ETFs change, the new trend is already half exhausted. By contrast, the average value of the stock database file is 7.91 days -- shorter, but not tremendously so.

As the administrator, I have two options. I can shorten the range permitted from 2-25 down to 2-18 or so, or I can stay the course. I've decided to drop it to 2-18 as a test, but it'll be several weeks before we see any output into the daily sheets since it takes about 3 minutes per stock/ETF. If I see more whipsaws in horizontal markets be aware that I'll move this back to 25...

You, as the user of the data, really do not need to do anything. If you're frustrated with GGT, then simply paper trade your favorite ETFs/stocks until you feel more comfortable. I posed some ideas about how to "ride the wave" this past weekend (see below).