The mid 2021 Synthetic Systems update is running as I write and should post this afternoon. Look for it under the Market Analysis menu item, Quarterly Charts, Synthetic Systems 2021.50. While we’re waiting, some background …
I began investing in the financial markets in 1993. Some time in 1994, after about a year of basically rudderless investing, I wanted to have something resembling a weather forecast, but for the financial markets. I began to develop some rough ideas of how it might work, and versions of what is now Synthetic Systems began to emerge. In early 2014, after about twenty years of concept, development, trial and error, the last major overhaul resulted in something resembling its current form.
Markets are notoriously difficult to forecast. By far the most consistently reliable forecasts are based on fundamentals and valuations and are really only effective on time frames of around a decade. Short term, meaning days and weeks, is far more hit and miss. Markets are a nearly perfect random walk over such time horizons. Some exceptionally skilled technicians can do a bit better than random in terms of handicapping odds, but more often than not one notable success isn’t followed by others.
What about time frames in between? Longer than a few weeks, but shorter than several years? Some gifted analysts combine fundamental and technical methods and original intuition to achieve striking results, but it’s rarely if ever possible to determine in advance who they will be. It’s generally only apparent after the fact who had that rare insight that proved uncannily accurate. And again, repeat performances are even more scarce.
Synthetic Systems is my effort to fill in that gap between the short term random walk and the fairly dependable long term. It is purely technical in its workings, but they bear no resemblance to the technical methods popular in finance. Its modeling has its roots not in traditional financial techniques but in the fundamentals and methods of engineering. No forecasting method is infallible, but Synthetic Systems is less fallible than any other systematic forecasting I’m aware of over its target time frame.
The charts posted here at Financology are also the most specific and unhedged forecasts I am aware of. Due to this very specificity, the most certain statement that can be made about its forecasts is that they will be incorrect in some respects. The charts literally plot a precise future course for each of five assets, each made up of hundreds of data points to four significant figures. The odds of getting them all exactly correct are akin to being struck by lightning while reading your winning lottery ticket.
Synthetic Systems’ performance in generally getting the course of the markets right, however, without hedges and conditions like “if this happens, then …”, is much better than random. For instance it correctly forecast the inflationary environment we’re now seeing as early as 2017 (see the Annual Charts page), long before the word “coronavirus” became a part of our current events vocabulary. On the other hand, it did not foresee last year’s corona crash itself … from Synthetic Systems’ perspective that was a genuine black swan. Notably, however, the markets did quickly reapproach their forecast course after the crash.
So what’s in the charts? They are generally plots of total real returns. This incorporates both prices changes and cash flows (interest and dividends).
TBills represents idealized Treasury securities of zero maturity. It’s very close to the shortest outstanding Treasury debt and overnight rates.
TBonds represents idealized Treasury securities of infinite maturity. While this might at first strike you as not particularly useful, it’s actually quite close to the total real return performance of some real world long term treasury funds, for example iShares’ TLT and Vanguard’s VGLT and EDV. The difference declines as rates rise. The
Stocks plot is pretty self explanatory. But don’t read anything into it that’s not there … did you see “US stocks”? All too often in the US we make the parochial assumption that the US stock market is the stock market. But here at Financology, “stocks”, without further qualification or context, just means stocks, the entire stock market, the whole world thing. It is very closely represented in real life by Vanguard’s VT. Finally,
Gold is gold and Copper is copper. Because there are no interest or dividends involved, total real return plots are identical to real spot gold and copper prices. Copper is included not merely for its own limited utility as an investment asset but also because it is broadly representative of industrial commodities as a class. Rising copper prices are associated with periods conventionally interpreted as “inflationary”.
Because it’s not absolute level, but changes in level (returns), that we’re interested in, there is no absolute on the vertical axis; the plots are vertically shifted as a whole as needed to fit a common window. The vertical axis is natural log, so that changes are proportions; for instance a change of 0.10 represents about 10%.
The original objective wasn’t, incidentally, to make forecasts of the cumulative returns for each asset, but to estimate as closely as possible the current smoothed rate of return, technically the cumulative return after application of a noncausal symmetric band pass differentiator. That remains my principal use for Synthetic Systems, and where it is most accurate. But the cumulative return data were a byproduct of finding the rate data, and due to their resemblance to common price charts, most readers find cumulative return plots more familiar and relatable. Readers who may be interested in seeing the rate plots, please let me know; if there is sufficient interest I may add a rate page at a future date.
There are two series of the charts at present. The annual charts are run at the beginning of each year, and cover the past twelve years and the next four. The quarterly charts are run at the beginning of each quarter, and cover the past twenty four quarters and the next eight. Each latest chart is accompanied by the four prior charts in the series, so readers can readily see for themselves how prior forecasts have worked out, by comparing time frames in the “Projected” portions of earlier charts with the same time frames in the “Actual” portion of later charts, as the future slips into the past.