Saturday, March 21, 2020

Dow Theory Update for March 21st: Secular trend for US stocks turned bearish on March 21st, 2020

Primary and secondary trends unchanged for US stocks, US interest rates, precious metals and their ETFs miners unchanged


Under Schannep’s Dow Theory

The primary trend is bearish since March 9th (penetration of the lows of the current secondary reaction) or even February 25th (alternative exit based on the penetration of the lows  of the previously completed secondary reaction), as explained here.

 Of course, the secondary trend is also bearish.

Under the Rhea/Classical Dow Theory

The primary and secondary trend turned bearish on March 9th (once again: the lows of the last completed secondary reaction violated) as explained here.

Change in secular trend (appraised using Schannep's Dow Theory applied to weekly bars)

If we appraise the trend for US stocks by using weekly bars, what I’d label the “secular” trend turned bearish yesterday. Readers of this blog know that I use weekly bars to appraise the higher order trend. More about the use of weekly bars in order to appraise the “secular” trend here.

Now let’s look at the entrails of the current “secular bear signal”. Here is the chart:

The starting point was a secular strong bullish trend that was signaled on 1/08/2016. 

On 09/21/2018 the INDU and the S&P 500 make their highest weekly closing highs. The Transports did so on 9/14/2018. From those weekly highs, the INDU and S&P 500 declined for 13 weeks. The Transports declined for 14 weeks. Hence, the average time of the three indices was more than 8 and the time requirement for a secondary reaction was met. As to the decline, the minimum movement I require is 6.7%. Why this figure?

Very easy.

Under Schannep’s Dow Theory the minimum movement stands at 3%. The minimum time for a movement to be meaningful is 2 days (the two bars rally/pullback following the secondary reaction lows/highs). If we use weeklies, two bars imply 10 trading days. If we divide 10 between 2, we get 5. Search in the web “Brownian motion” and you’ll understand why I calculate the square root of 5 which equals 2.236. This is the volatility adjustment (2 days is not the same as 10 days, two weeks). 2.236 x 3 (the minimum volatility for 2 days) equals 6.7%, which is the minimum volatility I use when appraising the trend with weekly bars.

As you can see in the spreadsheet below all three indices declined more than 6.7%.

Furthermore, the lows of the secondary reaction were followed by a rally exceeding 6.7%, so the setup for a secular bear market was completed. Such secondary reaction (and the setup for the secular bear market) was nullified when the INDU and S&P 500 bettered their secondary reaction highs. However, even though the secondary reaction was ended, it remains a valid exit point, as I repeat ad nauseam in this blog. So, last Friday March 20th, 2020, when the weekly bar was completed, the lows of the last completed secondary reaction were penetrated by the S&P 500 and Industrials (the Transports had done so last week), and, accordingly, we declare the "secular" trend as bearish.

The good news is that, with a little bit of luck, this "secular" bear does not have lot's of follow through. I tested what would have happened if, during the period 1998-2020 tested, I had gone short when the secular trend turned bearish and covered when it turned bullish. 

Here you have the three closed trades that would have been taken (the current one: enter short on March 20th, 2020, has not been covered yet, so it cannot be included):

Trades taken in going short when secular trend turned bearish and covering when it turned bullish
 Out of three "ended" secular bear markets. Two resulted in significant further decline following the secular bear market signal. The short corresponding to the 2000-2002 bear market netted out a profit of 16.81%, which implies that the further decline following the secular bear market signal had to be much larger than -16.81% (as by definition we cannot cover at the market bottom).

The short corresponding to the 2008-2009 bear market resulted in a profit of 28.01%, which implies that the further decline following the secular bear market signal had to be much larger than -28.01% (as by definition we cannot cover at the market bottom and the profit of 28.01% was made in spite of our covering at a quite high distance from the bear market lows. Please mind that we'd have covered on 07/27/2009, a relatively distant date from the market bottom of 03/09/2009).

However, the secular bear market signal of 08/01/2016 did not have any significant further decline. Shorting resulted in a whipsaw (having to cover at a much higher level) as the trade was a loser of -13.87%.

If the current secular bear market is going to be like the one of 2016, then we have already suffered the worst part. However, if 2020 is a repeat of 2000-2002 or 2008-2009 then we should brace ourselves for more declines.

From the three short "secular" trades we can draw one conclusion: The Dow Theory, irrespective of time frame (daily or weekly bars) tends to be most of the time precise at timely signaling changes in trends. I really was surprised to see that shorts would have been on average profitable. On daily bars this wouldn't have surprised, though. 

While I don’t trade according to the secular trend, it serves me to give me a much longer term perspective about the current condition of the market. Furthermore, even if we time our trades as per the trend determined by using daily bars, my preliminary research seems to show that superimposing a “Dow Theory-determined” secular trend filter seems to improve results (or more exactly: weeding out suboptimal trades and reducing drawdowns, even though there is no guarantee of increased overall performance, as in some instances too many trades, some of them good ones –in the beginning of new trends- may be weeded out). In other words, don’t enter a trade determined by the daily bars if the “secular” trend is bearish.

I’m also preliminary concluding that the secular trend filter comes in particularly handy when dealing with noisy markets. In my last post I wrote about a noise indicator I’m developing and how I can classify markets according to their noise. Thus, gold is noisier than US stock indices. And Chinese indices (i.e. large cap) are noisier than US stocks. In other words, if I am trading US stock indices according to daily bars (that is the “normal” way to apply the Dow Theory, be it Schannep’s or the “classical”) it seems unnecessary superimposing a secular trend filter. There is enough “signal” in US stocks indices so it is not necessary to add an additional filter (the less rules, the better; the simpler our trading systems, the more robust and dependable when looking forward).

On the other hand, with noisier markets such as gold, I seem to preliminary discern that rejecting long trades when the secular trend is bearish improves return while reducing drawdowns. It might even be the only way to trade noisy markets if one aims for decent profits.

And what about trading US stock indices only according to the secular trend? Well, I don’t recommend it because trading along the secular trend entails too ample stops, so you can potentially (or “actually” as in March 2020) experience drawdowns exceeding 30%. For my personal risk preference this is way too risky. Two losing big trades in a row could be quite unpleasant. As far as I tested from 1990 to date, this has never happened, but I feel uncomfortable trading with stops (the lows of the current or last completed secondary reaction) that are too wide.

I must say, though, in favor of trading stocks along the secular trend the following:

·        From 1998 to date there has never been a losing trade. Even the current one, exited more than 30% from the top, has been a modest winning trade.

·        The profit factor (total amount won divided per total amount loss on a closed trade basis) has been infinity (that is there were no losing trades).

·        Hence, on a closed trade basis, even now we have never had a drawdown (albeit we have seen unrealized profits “vaporize” in just three weeks)

·        If we appraise performance from 08/31/1998 to 03/20/2020, CAGR would have been 5.10% which is slightly inferior to the CAGR made on identical period when using daily bars (CAGR of 5.88% not published yet on this blog).

·        However, it would have been piece of cake to trade, just 4 trades in more than 21 years.

·        Even with the current drawdown, the Dow Theory applied to weekly bars would have handsomely beaten buy and hold (a total return of 196.97% versus 102.48%) whereas drawdowns (in 2000-2002 and 2008-2009) would have been smaller. Even now in March 2020 it could well happen that the drawdown of the Dow Theory when applied to weekly bars  will be smaller than that of buy and hold. We still don’t know.

Therefore, it maybe makes sense to trade US stock indices along weekly bars if they are part of a portfolio (i.e. if stocks allocation were just 25%, and, i.e. a 30% loss would just entail a 7.5% loss at portfolio level). I insist, though, than on a closed trade basis all trades were winners. The accuracy in detecting winning trades shows us that the longer the time frame the more signal and less noise. However, more accuracy does not necessarily translate into more return or even less drawdown (when measured as "vaporized" unrealized profits)

Here you have some screenshots taken from TradeStation® performance report:

Four winning trades in more than 21 years. No losers
Total return outperformed buy and hold

Accumulative monthly return. In 2000-2002 and 2008-2009 drawdowns were kept at bay

Detail of each trade taken and (right side) percentage won on each trade

Equity curve on a closed trade basis: No drawdown


The primary and secondary trend were signaled as bullish on 02/19/2020, as explained here

In just four days SLV declined drastically and almost violated the last recorded primary bear market lows. GLD declined but on a much more muted basis. Hence, the decline does not meet the time requirement for a secondary reaction.

On March 11th, SLV violated the lows of the last primary bear market. Rhea (page 77 of his book, Fraser Edition 1993) recognized as a valid exist point the closing lows of the last primary bear market (red horizontal lines on the charts below). GLD remains at a safe distance, and hence no primary bear market was signaled. More about the alternative entry and exit signals under the Dow Theory here.


The primary and secondary trend was signaled as bearish on March 11th, 2020 as explained here


Depending on the way one appraises the secondary reaction that led to the setup that resulted in the primary bull market signal, the primary bull market was signalled either on 11/19/2018 or 12/18/2018. Rhea wrote that the definition of secondary reaction is not carved in stone. The signal of 11/19/2018 was obtained by being satisfied with just 14 trading days for TLT and 15 days for IEF. The signal of 12/18/2018 was obtained by being strict and demanding on a confirmed basis at least 15 trading days on both ETFs. It’s up to each investor to decide what to do (i.e. to commit to each signal 50% of one’s equity or go fully invested with just one signal).

On 11/08/2019 as secondary reaction was signaled, as explained here.

On 02/21/2020 TLT bettered its last recorded primary bull market highs of 08/28/2019. On that date IEF equalled (but did not better) its last recorded primary bull market highs of 09/04/2019, and hence there was no confirmation. On 02/24/2020 IEF did better its primary bull market highs and, therefore, we can declare the secondary reaction has ended, and the primary bull market as reconfirmed. From the reconfirmation date of 02/24/2020 TLT and IEF went parabolic reflecting the current chaos which is plaguing all markets. However, during the last few days both ETFs are declining. The current decline does not meet the time requirement for a secondary reaction yet, and hence the secondary trend remains bullish.

One Dow Theorist

No comments:

Post a Comment