Omicron Research Institute

QuanTek Econometrics Software

Model Portfolio

Most of the Graphs are too big to fit on the web page. So, we have provided a link to them and to other files such as Report files. To view these files in your web browser, just click the link. The files will open in their own tab in a separate web page. Then, to return to this web page, just click its tab in your browser. You can switch back and forth between tabs to compare the Graphs and Reports and then return to this web page.

Warning: The Price Projection represents the expected N-day prices.The error bars represent the expected N-day price range.  These are estimates only. It is entirely possible for prices to end up outside the expected price range at any time.

It should be emphasized that the Price Projection cannot predict every short-term fluctuation in prices. It is based on past prices only, and cannot anticipate exogenous events such as news events that influence prices. However, it seems to give predictions that are at least plausible, and over a long-term average (1024 days or 4 years), yield significant correlation with future returns, according to the correlation tests.

We will be setting up a Model Portfolio shortly, but first here are some preliminary results from the Portfolio Report and some screen shots of Graphs in QuanTek 3.9.

Dow Industrials (.DJI) (2018-12-07)

Here is the latest shot of the Dow Industrials (.DJI): (For some reason, all the data downloaded this week was missing data for Wednesday. Presumably this data will be filled in next week. At least QuanTek is robust against missing data points.)

Dow-1181207

This week the projected 32-day (annualized) return is 12.43%. The bounce from the previous week fizzled out and the market turned down again. Presumably this is due to unfavorable news events, such as predictions of lowered earnings expectations for next year, and the tariffs and impending trade war with China. Also, the expectation that the Federal Reserve may raise rates due to an overheated economy will have a serious effect. From a broader perspective, it appears that the market is gradually flattening out and will be essentially flat for the next year or two, or perhaps we will have a mild bear market. It could be more severe if the global economy undergoes a serious slowdown due to various (exogenous) events unfolding. But the 32-day Price Projection for the DJI seems to be in line with the historical return for this index.

The data are corrected for splits and dividends and goes back 2048 days. The yellow line is the 512-day smoothing curve. It should be noted that the cyan and magenta lines are 1x and 2x the average absolute deviation of the daily (log) prices (Bollinger Bands). So even the outer magenta lines only represent 2x the average daily range of the prices. For a real bear market, we would expect the index to range substantially far outside the range of the magenta lines, which they have not done so far. However, even a real bear market is usually a case of an over-extended market correcting back to its long-term trend line, as happened in 1987 for example.

Portfolio Report (2018-12-07)

The Optimal Portfolio we are tracking for this week is displayed in the following QuanTek Report file:

Report (2018-12-07).pdf

It looks like we lost some money this week -- ouch! The Optimal Portfolio looks a little different because we switched a few securities in and out of it. But we don't see any that are at especially favorable buy points this week, or sell short points either. The sell short candicates all seem to have dropped by quite a bit already, which makes them risky. However, probably a short sale or two would be a good idea, to protect the portfolio against these sell-offs that seem to keep occurring. No doubt it is just the market re-adjusting to lowered expectations for next year. So we may have to see how well we can do in a flat market for the next year or two...

We might be tempted to buy XOM, which has a 32-day expected return of 6.85%. This stock has a rather puny Sharpe Ratio of 13.99%. However, it will be noted that in the Optimal Portfolio, the Optimal Position is a short position of -571 shares. This means that the Optimal Portfolio calculation has determined that a short position gives a greater benefit by decreasing risk in the portfolio, outweighing the loss due to decreasing return. Another example of this is KSU. On the other hand, SILJ has a Sharpe Ratio of -0.27%, but a positive Optimal Position.

Since prices are down, I have decided to buy 800 shares of FB, due to its favorable Sharpe Ratio of 135.13%, at the close price on Friday. This stock (Facebook) has been falling sharply since July, no doubt due to its troubles regarding privacy of user data. But it seems to have reversed this week, while the broader market has fallen, so maybe now is a good buy point. This stock seems somewhat risky, but the price is right, and once again its favorable Sharpe Ratio of 135.13% is the best in the Optimal Portfolio.

Portfolio Securities (2018-12-07)

We have spent this week tweaking the Adaptive Filter for best performance on AAPL. Here are some screen shots of the QuanTek Splitter Windows for AAPL. The first two Splitter Windows represent three different Indicators, with low-pass smoothing and band-pass smoothing. The third Splitter Window shows the output of the Adaptive Filter directly and associated indicators:

Figures (2018-12-07).html

In the first two Splitter Windows, the three indicators shown are Relative Price, which is the price action relative to the 512-day smoothing curve, the Velocity, which is the returns or daily price differences, and the Volatility, which is the absolute deviation of the daily returns. All of the indicators are smoothed using N-day Wavelet Smoothing, and they are all causal (except the Range and Threshold lines themselves, which depend on taking an average over the whole indicator). The most important indicator in each case is the Relative Price. The Volatility indicator is relative to zero for Low-Pass Smoothing, but relative to its mean value for Band-Pass Smoothing.

Fig.1 AAPL-Indicators (Low-Pass Smoothing): These indicators are smoothed using Low-Pass Smoothing, similar to a moving average. The bottom indicator is Relative Price, and it serves as an indicator of overbought/oversold conditions. It can be seen that up to a couple of years ago, the indicator showed a Buy Signal, indicating an oversold condition. The past couple of years it has been relatively flat, however. The Range control sets the level at which Buy/Sell Signals are triggered.

Fig.2 AAPL-Indicators (Band-Pass Smoothing): These indicators are smoothed using Band-Pass Smoothing, similar to the difference of two moving averages. The bottom indicator is Relative Price, and it serves as an indicator of N-day inflection points, similar to an oscillator indicator. These are shown on the graphs as the N-day Buy/Sell Points. With the recent downturn, it can be seen that this indicator shows a strong Buy Point. The Range control also sets the level at which the Buy/Sell Points are triggered. Actually, we don't use these Buy/Sell Points, but they are convenient for lining up features on the graphs. They might be useful for N-day Swing Trading, however, if you are downloading data every day.

Fig.3 AAPL-Adaptive Filter Output: These three indicators pertain to the Adaptive Filter. At the bottom is the raw output of the filter. Green indicates a positive expected return and red indicates a negative expected return. This indicator forms the basis of the Long/Short Signals. The Threshold control sets the level at which the Long/Short Signals are triggered. It can be seen that up to about a year ago, the Adaptive Filter was giving a strong positive signal, which is a long indication. Lately, however, the indicator has been more or less flat, which seems to indicate that AAPL is no longer on such a strong up-trend. The middle indicator is the actual N-day 'future' returns which is the "desired output" of the Adaptive Filter. It can be seen that most of the short-term fluctuations are just stochastic noise and are not picked up in the filter output. This is also due to the fact that the LMS filter is a low-pass filter. The upper graph is the actual expected return due to the 2048-day long-term trend, extended into the future. It can be seen that the Adaptive Filter output resembles this long-term trend much more closely than it does the "desired output" that it is trying to mimick. However, it is not exactly the same, and under favorable conditions the Adaptive Filter can exceed the performance of the long-term trend. (The long-term trend is pretty good, however, over the long-term.)

Portfolio Report (2018-11-30)

The Optimal Portfolio we are tracking for this week is displayed in the following QuanTek Report file:

Report (2018-11-30).pdf

We have made some changes to the way the Optimal Portfolio is calculated. We now use the Long-Term (2048-day) Trend for the expected return instead of the N-day Adaptive Filter output. This will make the portfolio much more stable, and more profitable, since according to the correlation tests the Long-Term Trend is hard to beat (over the long-term). Now the Optimal Portfolio is completely independent of the Adaptive Filter, and will vary only slowly from week to week. Also, we have decided to go to a 32-day Time Horizon, instead of the previous 20-day Time Horizon.

It appears that the purchase of AAPL and MSFT last week was a good call, as the Model Portfolio has gained nearly 8.8% over the past week. The predicted "bounce" in AAPL and MSFT actually occurred as predicted. So the Price Projection along with the Graph indicators proved useful to find a good buy point for these two securities. The Price Projection never indicated a sell signal during the decline in AAPL over the past two weeks, but instead indicated a buy signal. So this turned out to be smart advice -- it would be a mistake to sell short AAPL during an ongoing bull market!

As for GE, note that the 2048-Day Return is 10.71%, in spite of the recent sharp downturn. For the past week, in accordance with the negative Price Projection, I could have made money by selling GE short. However, I feel it is risky to sell short a security that is already down by such a large amount, in an ongoing bull market. So that might be OK for a (risky!) short-term play, but not for a position in the Model Portfolio.

This week it was a little difficult to find any securities that are at a good buy point. In fact, QuanTek is not indicating very many good buy points at present. I did finally decide to buy 800 shares of JPM at the close price on Friday (2018-11-30). This provides a little diversification for the portfolio. I was tempted to buy CSCO, as the Price Projection for this security looked favorable. However, it seems to be a little high-priced at present. In all probability the Price Projection will be vindicated. This seems like a good candidate for the buy-high, sell-higher strategy. However, if we have another tech downturn then buying at such a high price will turn out to be a mistake. I think I will wait until the next correction to buy this one. Also I was tempted to buy VGLT and ZROZ, which are bond ETFs. These seem to be favorably priced at present and showing a favorable return on the Price Projection. However, QuanTek does not know about the Federal Reserve raising interest rates, which most likely will hurt bond prices and therefore bond funds. As a matter of fact, these two funds have been on the decline for the past year or so. In fact, the rising interest rates will probably help the bank stock JPM, so I think it will be a good choice for the long term. The updated Model Portfolio is on the second page of the Portfolio Report PDF file for this week.

Now that we are using the Long-Term Trend for the expected return, the Portfolio Expected Return is now a much more reasonable 14.90%. This is because it no longer relies on the wild fluctuations of the N-day expected return from the Price Projection, with N ranging from 1 to 128 days. But the Price Projection is still very useful for timing the buy/sell points as well as for N-day short term trading.

Portfolio Securities (2018-11-30)

Here are some screen shots of the QuanTek Main Graph for various securities in the Optimal Portfolio:

Figures (2018-11-30).html

Fig.1 JPM-JP Morgan Chase & Co. (Scale 4): This was our pick for the week. It appears to have a stable behavior and not overvalued. The Price Projection looks favorable. This is a bank stock, and should benefit from rising interest rates over the next few years.

Fig.2 CSCO-Cisco Systems, Inc. (Scale 4): This looks like a healthy stock, but a bit overvalued. Note the very favorable Price Projection and Buy Points. But we prefer to wait for the next downturn for a better price.

Fig.3 AMZN-Amazon.com Inc. (Scale 4): This is a high-flying tech stock, one of the FANG stocks, but seems to be on a downturn from a recent overbought episode back in Ausust and September. The negative Price Projection is also worrisome. This should be a Buy if it can correct back down to a more reasonable price.

Fig.4 VGLT-Vanguard Long-Term Treasury (Scale 4): This is a bond ETF. It looks very reasonable at first. But it appears to have declinded in price over the past year due to rising interest rates, which should cause it to decline further as rates rise. QuanTek has no way of knowing about Federal Reserve interest rate policies, so it predicts a return to the mean mechanism and positive Price Projection.

Fig.5 ZROZ-Pimco 25+ Zero Coupon (Scale 4): This is a bond ETF. It looks very reasonable at first. But it appears to have declinded in price over the past year due to rising interest rates, which should cause it to decline further as rates rise. QuanTek has no way of knowing about Federal Reserve interest rate policies, so it predicts a return to the mean mechanism and positive Price Projection.

Fig.6 TGC-Tengasco Inc. (Scale 4): This is an oil/gas company, and it serves as an example of a very risky security. Note the extremely wide Bollinger Bands. You might not notice how risky this stock is, if you were to see a normal graph of it in isolation. The QuanTek graphs are all on the same (logarithmic) scale, so you can immediately notice the wide Bollinger Bands and wild behavior compared to other stocks. The most recent price was $1.02, so this qualifies as a borderline "penny stock". Typically, the lower the price, the greater the percentage swings in price, and the greater the risk. This is immediately visible on the logarithmic scale, which shows percentage changes. Note also the corresponding very wide Error Bars on the future Price Projection. You may not want to own this stock unless you are a real risk-taker!

Portfolio Report (2018-11-23)

The Optimal Portfolio we are tracking for this week is displayed in the following QuanTek Report file:

Report (2018-11-23).pdf

Some of the securities are again down for the past couple of weeks, and QuanTek is still predicting a "bounce" in some cases such as AAPL. That is why the projected Portfolio Expected Return is an apparently amazing 238.36% (annualized) over the next 20 days (high, though less than last week). (Note also that this Portfolio Expected Return depends on the Price Projection being correct.) If you look at the recent past in the graphs, it can be seen that a similar "bounce" has taken place when the prices dip as they have been doing lately. At present, except for GE, the securities are still within a normal trading range. So the 20-day (4-week) prediction of a "bounce" still has a good chance of being correct. Short-sellers of GE over the past week would have done well. This security is way outside its normal trading range and has departed drastically from its trend line. But its recommended position in the Optimal Portfolio has been reduced to -25.76% of the portfolio, not quite as drastic as before. As for AAPL, we expect it to reverse course and continue its upward trend sooner or later.

I decided it was time to buy some securities for the Model Portfolio. Over the weekend I bought 400 shares of AAPL and 800 shares of MSFT at the close price on Friday (2018-11-23). So now you can see the Model Portfolio on the second page of the Report file. I went on the basis of the graph itself and the Expected Return from the Price Projection, but not really on the Buy/Sell Signals or Points. Now QuanTek will keep track of the gains/losses in the portfolio automatically. Note that the basis price is recorded so that it can be easily compared to the current price for each security in the portfolio, at any time.

Portfolio Securities (2018-11-23)

Here are some screen shots of the QuanTek Main Graph for various securities in the Optimal Portfolio:

Figures (2018-11-23).html

Fig.1 GE-General Electric Co. (Scale 4): Here is an example of a stock with a very abnormal breakdown. This shot is on Scale 4, because the higher scales would be too big to display. The price action for GE has broken down badly, having declined by over 50% in the month of October alone. The 512-day smoothing curves are curved up because, after the end of the past data, these curves are trying to make it back to the 2048-day long-term trend curve, which is far above. Notice that the Price Projection has a negative slope, indicating trend persistence, unlike the case of .DJI in which the slope was positive, indicating return to the mean.

Fig.2 GE-General Electric Co. (Scale 2): Here is GE again, this tine on Scale 2. On this scale, the 2048-day (robust) long-term trend line is displayed instead of the 512-day smoothing curve. This shows more clearly how the prices have broken down far below their long-term trend line, starting back in the summer of 2017. When the break-down first started, a Buy signal was given, according to the return to the mean mechanism. As the breakdown progressed, a Sell signal was then given, indicating a trend persistence mechanism, in this case a downward trend. This GE stock is evidently a case of what Peter Lynch would call "diworsification" -- the company got so big and diversified that it could no longer function efficiently. How sad...

Fig.3 GE-General Electric Co. (Scale 1): Here is GE again, this tine on Scale 1. This is the lowest scale, and (if you have a big enough screen) you can see the whole 2048-day (8 year) price action all at once in one view. In this view the yellow line is a (robust) 2048-day trend line, and it is clear once again how GE has broken down and fallen away from its long-term trend. Before the breakdown, the return on the long-term trend was 10.83%. This view also emphasizes the importance of using corrected data, corrected for splits and dividends. Only the corrected data shows the true price action, and allows the Adaptive filter to function. In this view, each data day is 1 pixel, so you need 2048 horizontal pixels to see the whole price action at once.

Fig.4 AAPL-Apple Inc. (Scale 4): This is a graph of AAPL showing its recent breakdown, which started near the beginning of November. From the graph, AAPL appeared somewhat overvalued in the months August through October, which may account for its breakdown in November. The QuanTek Price Projection predicts a "bounce" with an annualized gain over the 20-day time horizon of 206.77%. This would be a bounce similar to those that occurred in early February and early May, from similar levels. At any rate we don't know of anything fundamentally wrong with the company, so this might be an excellent buy opportunity. At some point the prices must return to the strong upward long-term trend -- there is no apparent reason for them not to.

Fig.5 AAPL-Apple Inc. (Scale 8): This is the same graph of AAP, but on a higher Scale 8. Notice one of the main features of the QuanTek graphs, that they preserve the slopes of the graphs between different scales and different securities. This makes it easy to compare at a glance the rate of return between the graphs. This scale shows some Buy/Sell Signals, represented by green/red rectangles at the recommended daily buy/sell prices. These Buy/Sell Signals are indicated when the Relative Price (low-pass) indicator is below/above the Range setting, and the Projected Future Returns (output of Adaptive Filter) are positive/negative. You would think there would be a buy signal at the present time, but in this case the Relative Price is actually less than the Range setting. Setting the Range lower would yield more Buy/Sell Signals. This seems to indicate that the "breakdown" in AAPL is not really as bad as it looks.

Fig.6 AAPL-Apple Inc. (Scale 16): Finally, here is AAPL again on the highest Scale 16. This is a really big graph! (I hope it fits on your screen!) This graph shows the Long/Short Signals. These are indicated when the Projected Future Return (output of Adaptive Filter) is above/below the Threshold setting. These show the actual output of the Price Projection, superimposed on the graph. The double green/red triangles show the actual projected price, while the small green/red rectangles show the corresponding buy/sell prices, located at the Close price. (If N is the Time Horizon setting, in this case 20 days, then the projected prices are located N days after the corresponding buy/sell prices.) It can be seen that the projected prices from the red sell prices on the left of the graph seem to track the actual prices amazingly well. This is because the price pattern actually repeats in the graph! However, on the right the projected prices from the green buy prices seem to miss the mark and are too high. So the Price Projection missed the downturn starting around the beginning of November. This downturn must be a result of some exogenous event, such as some of the current news events. Nevertheless, overall a buy signal on AAPL seems a lot more prudent than a sell signal.

Go back to Home Page