Yesterday I mentioned testing and tweaking a core upgrade to Synthetic Systems, and that it should be ready by mid year. That process is still under way, but in light of the activity in the financial markets I want to give you a preview. Although I’m already more confident in its forecasts, it is not yet final; the main reason for previewing it is to highlight the areas in which the current forecast is most subject to uncertainty.
So here is the current SS chart, with the new beta version (let’s call it SSS) immediately below it for comparison. The most noteworthy differences are that SSS doesn’t see the strong summer rally in stocks and commodities that SS does, and that SSS thinks the bond market has already bottomed. That these have already been the areas of SS I think most questionable is purely coincidental … SSS is purely mathematical and has no provision to input my opinion of the fundamentals.
I’ll give a fuller introduction of SSS when the testing and tweaking is complete and it formally comes on line at mid year. To correspond with the current latest quarterly SS update, 2022.25, SSS was also run as of that date with the same data available at the time.
The additional S by the way is for “Symmetric”, because it replaces a formerly antisymmetric system with a symmetric one.
19 thoughts on “Systems Beta”
I’m curious about whether you feel energy and agriculture are well represented by your commodity proxy-copper. I certainly see copper tied to industrial metals, but I think there are fundamental reasons energy and ag might rise even in conditions that hit industrial materials.
In the short run, I would call the connection loose. Copper of course represents copper better than anything else, but copper itself isn’t a big part of most portfolios. You can’t own much directly, it’s volatile, and I don’t want give readers the impression they should be trying to trade copper. I think a broad commodity ETF is appropriate for many though. And for lack of a better word, I use “industrial commodities” to distinguish from monetary commodities such as gold.
Longer term, however, the correlations are surprisingly tight. We discussed this in detail in the comments under Outlook for the 2020s. Note especially the cited correlation coefficients and for visual impact, the twenty year chart of copper and oil.
Specific commodities and commodity sectors are better approached via fundamentals. An investor specifically bullish on energy could buy a commodity futures fund linked to the energy-heavy GSCI like COMT or via an energy stock fund like VDE. For agricultural commodities one could buy fertilizer producers or farmland stocks. DBE and DBA are also good options. In such cases the Copper plot only gives an indication of the prevailing wind. It’s not the only determinant of performance, but it’s better if it’s at your back.
A less obvious implication of the Copper plot is its strong anticorrelation with bonds. In less complex portfolios it may be preferable to simply reduce bond exposure in response to a rising Copper plot.
Bear in mind that Systems’ bailiwick is the macroeconomic and financial system. The implication of the long rising trend of the Copper plot is commodity price inflation and by extension, consumer price inflation. That’s what I predicted at the beginning of 2020 and that’s what we’ve been seeing. The implication of the long declining trend ahead is the opposite. Although it seems likely on fundamental grounds that energy and ag will outperform copper in the coming months, it’s unlikely they would be exempt from decline once the copper trend turns decisively downward. I expect to underweight all industrial commodity related positions in that kind of environment.
I’m assuming that you are the one who has programmed SS & the new beta, SSS, correct? I spent many years as a programmer and would like to know, purely out of interest, what language is SS/SSS programmed.
And, I’m assuming it is a Windows based program, is it?
Thanks, Peter. Yes Systems is my own handiwork. I started working on it back in 1993, so it’s nearly thirty years now. (Although it took nearly fifteen of those years before I started finding something that would actually work!).
With this background maybe it’s not so surprising that I’m running it on Excel 5 (circa 1995). On Windows 7. Alas later versions of Excel added a lot of clutter and actually lost some key features for serious number crunching. Also keeping the same platform helped me to get really familiar with it without being distracted by updates. Theoretically it could be made faster by being rewritten in a lower level programming language, but the time required for calculation is still quite a bit less than it would take me to do that;-).
You might find it interesting that explicit market forecasting wasn’t the original objective. Rather the goal was to estimate a smoothed current rate of return. At some point I realized it was just a couple operations from producing forecast charts … they were practically just a byproduct. SSS is the first version ever to go straight to the forecasts.
Peter, it appears you posted a reply to JK intended to reply to the above. Look for your comment below.
just reviewed our exchange in the earlier thread you referenced.
i have the [dangerous] idea that “it’s different this time” for both energy and agriculture.
re energy, capex has been very low, both because of the capital destruction in the shale “boom,” and because esg mandates have squeezed the availability of capital to energy firms. the issues around russia are merely the icing on this cake – energy investment and supply was constrained prior to the russian invasion of ukraine.
further, natural gas in particular is in a globally awkward position, with european gas selling for about 5x u.s. gas’ price. eventually this can be partially arbitraged away less the cost of cooling, transporting and regassifying lng. all 3 of these steps are significantly capacity constrained, and each requires billions in investment and years of construction to loosen those constraints. thus it seems that an etf like unl will do well for the next 5 or more years, or until sometime after putin is dead and significant and unlikely political changes occur in russia.
similarly, nitrogen-based fertilizers which depend on natural gas as an input can be another arbitrage mechanism. cf and ntr have been doing well and will likely continue to do so. these combine the energy theme with the ag one.
in the meantime, limited supplies of fertilizer, higher prices for the diesel that powers ag equipment, and supply chain problems for equipment components all conspire to raise food prices. rja is a low volatility play.
Good analysis, JK. There are multiple issues involved. My aim here is to try and give readers as clear a picture as I can of what Systems does and doesn’t do. There’s also the general question of the correlation between physical commodity prices. There’s also what investment landscape looks like over the next few quarters and over the rest of the decade. Addressing these separately will help sort us out.
First remember Systems is not a replacement for fundamental analysis. It doesn’t override any of your fundamental arguments. What it does do is fill in some color on the economic and financial context and timing. I’m sure you’ll agree that whatever the long term merit of your outlook, these commodities will not go up in a straight line. There will be periods where they pull ahead of their underlying trend and others where they get ahead of themselves and underperform their trend. With this in mind, Systems is best looked at as an overlay to the fundamentals to shed some light on when these things are most likely to take place.
Correlations are not absolute or perfect. (Especially when it comes to making predictions about the future;-) Take for example my twenty year chart of copper and oil. The cross correlation is stark. Yet over the course of those twenty years (recall this is a log chart), copper nearly doubled in terms of oil, or looked at the other way, oil fell almost by half in terms of copper. Suffice it to say that even over two decades correlation did not become perfect.
Yet there is always risk in extrapolating fundamentals too far. There is a widespread tendency to read asset performance due to financial factors into fundamentals. You may remember back in 2008 we had a vigorous debate in Finster’s Comments on iTulip in the thread “The Last Time We Had Peak Oil”. Oil was approaching $147 a barrel, with forecasts of $200, $400 and higher popping up like weeds. Investors were convinced the world was running out of oil. I argued at length and depth that the case was grossly exaggerated, that oil was soaring because it had entered an inflationary vortex. I won few converts, but a mere six months later, oil prices had fallen to $30 in a deflationary crash. I’m not disputing your fundamental analysis, but arguing that there is a very real risk of overestimating the price impact of fundamentals and that the financial environment holds much more sway over commodity prices than most analysts realize.
Thanks for the update regarding what program SS/SSS are written. I can understand why you stayed with Excel 5 & Windows 7. Excel has become more and more bloated over the years.
I do find it interesting that explicit market forecasting wasn’t the original objective. Well done.
15 years before you got something that worked; now, that’s commitment. Of course, I assume you were still working full-time back then, hence most likely didn’t have a lot of time to devote to it.
Correction – It’s Excel 95 – I just checked. My understanding is that the main change from Excel 5 was 32 bit processing for Windows 95. It ran bug free on Windows 2000, but there are focus problems with Windows 7. I’ve gotten so used to them though that they’re still a lot less problem than newer versions.
I have Excel 8 running on a six core 3.33 GHz Mac Pro that would be much faster than my old Windows PC but it’s just not as elegant as Excel 95 and I’d have to find add-ons for some of the serious number crunching features it lacks. I also use several VBA routines and custom functions that would have to be rewritten. Making matters even more complicated is that I don’t connect the Mac to the Internet. I use it mostly as the core of a recording studio and don’t want to clutter it up with junk that seems to inevitably accumulate with an internet connection. No security bloatware just for starters. Just as fast as the day I first booted it up. But getting new data into it would be cumbersome.
No wonder technology isn’t yielding more productivity gains. So much potential is eaten up by update and security overhead.
I’ve never really worked on this full time though. Just whenever an idea occurs to me I may put in a lot of hours developing it. Minor upgrades are pretty common but don’t take much time. The bigger changes come only every few years. A notable exception is as it turns out an idea for an even more basic architectural change popped up last Friday night and SSS may already be obsolete. So another round of testing is under way. Both SSS and the newer version are markedly more accurate than SS but the newer version is much more compact and takes a fraction of the time to run.
all good points, bill.
btw, do you plan to do revised annual runs with the new core?
Right now the plan is to use SSS for each update as it’s done. All along I’ve been keeping past runs original even through updates to the model. That’s why some of the older annual charts still show a smooth line for Bills … they were done in dollar terms in older versions of SS.
But time permitting I hope to add a new page that shows past forecasts using the new architecture. This would include posting charts as of the same date but one showing each forecast directly aligned with the actual outcome. You can already compare forecasts and outcomes on the present page, but they’re time shifted. The new page would make it easier. The tradeoff is that that would be limited to forecasts at least a year or two old because the actual outcome for more recent forecasts is not yet available;-).
question re interpretation of your charts:
it appears that we’ve already begun the equity sell-off predicted by both ss and sss to begin at the end of q3. does that imply to you that the other asset classes are similarly ahead of schedule? i.e. have bonds and gold bottomed? have commodities already peaked? should all the curves be shifted 1.5 quarters to the left?
Excellent question. I wish I had an equally straightforward and clear answer. I think in part it’s yes. But there is a substantial caveat. The lion’s share of the equity declines to date have been nominal only. In other words, what we’ve seen wouldn’t be indicated by a falling Stocks plot, but by a rising Bills plot. That was the gist of the point I wanted to make in Bear Market Update 20220507. SS has actually been startlingly accurate up to that point. As of then, the main departure from forecast was a subtle increase in the rise in the blue line. Even that recently the Stocks plot itself had remained strikingly level. This also added to nominal declines in gold and offset some nominal gains in copper.
Will that continue to be the case? I don’t think so. As mentioned in this post, early indications are that the bond market may have already bottomed. If so, the gains in the dollar should pause; indeed both SS and SSS have a brief retreat penciled in. But I don’t want to belabor the point because that’s training a microscope on a time frame well below Systems’ resolution.
But if the forecasts as a whole are off by as much as 1.5 quarters, I would call them just plain wrong. Yes you could make an adjustment for that, but it would raise still more questions. Is the discrepancy in the blue line? In stocks? Or everything? I just think that with the extra-economic jokers in the deck it’s not possible to draw any firm conclusions.
Given the uncertainties my tack generally is to fall back on the timeless principles of diversification and to not go very far out on any one limb. I continue to hold elevated cash and think bonds once again merit at least their normal allocation. Similar with gold and other commodities. My main underweight continues to be US equities. I plan to do a more in depth review at mid year when the new SSS is fully on line and meanwhile continue the running market commentary.
As always, please feel free to weigh in with any questions or insights!
i’ll be interested to see the annual sss 2022 chart, esp. to see if it shows the rather dire 2H24 in the ss annual.
I think we can arrange that. Just set it up … give it an hour or so to process and check back in. But do you mean 2025 H2? 2024 H2 looks relatively unremarkable to me.
Bear in mind though that this is about at the opposite end of the short term spectrum for Systems … four years is about where it leaves off. Otherwise I could just run an “annual” forecast only once every four years! Also that this will be far from final until that testing and tweaking is complete…
Again take with an extra grain of salt – there are lots of updates over the four year span – but the big picture to me looks like a Fed engaged in a quixotic quest, struggling alternately with trying to fight inflation and trying to prop up the markets.
looks like powell might get his wish to be the reincarnation of volcker with a double dip recession, 3q21 to 3q24 and then 2h25 until heaven knows when. be careful what you wish for.