Evaluation of Schannep's Dow Theory versus 21 years of buy and hold (1998-2019)

As I wrote in my last post, there are many projects expanding the Dow Theory in the pipeline. However, before we embark upon discovering new uses for the Dow Theory, we'd better remain on solid ground. To this end, we are going to compare the performance of Schannep’s Dow Theory to that of buy and hold (BAH) when applied to the SPY, which is the ETF representing the S&P 500. 

When I write "Schannep's Dow Theory" I mean the set of rules which includes "capitulation" (more about here, and the links given therein), the so-called "bull/bear market definition" (more about it here) which entails a confirmed decline or advance (by the S&P 500 and the Industrials) of -16% or +19% and, of course, the typical signal (secondary reaction, rally/pullback, breakup/breawdown). 

Not included in this post is the so-called "timing indicator" which is Schannep's proprietary indicator and constitutes the "second" leg of his real life trading with half of the trading funds committed to it. As far as Schannep discloses in his book (a must read for everyone really intend on becoming a good investor) and on his website "thedowthery.com" the "timing indicator" is based on momentum and monetary policy (being easy monetary policy bullish). As I wrote in the past, since I don't know the entrails of such an indicator, I cannot use it in my blog. I hope one day, I'll get to know it, since I feel it is a very good one. From his publicized record on his website, we can deduct that the "timing" indicator has really trounced buy and hold on a permanent basis (since 1953 so it is not a statistical fluke). It has also outperformed his also outperforming Dow Theory which is quite a feat.  As with Schannep's Dow Theory, the "timing indicator" has real money committed to it since time immemorial (since the sixties). So we are not talking of something theoretical or unproven.

From Schannep's website, I can see that the tracking of the "timing indicator" began on 12/31/1953 with a starting capital of USD 10,000. At 08/31/1998 (start of our test) the accumulated capital amounted to USD 4,836,715 at 12/31/2019 (latest date available) the accumulated capital had risen to USD 31,894,18. Using Investopedia CAGR calculator, I obtained a CAGR 9.40% (whereas, as we will see below, buy and hold merely made 5,21%). So, the "timing indicator"is a vital leg of Schannep's trading.


Hence, this post does not do full justice to the whole of Schannep's trading system, as it only focuses on one of its two legs: The Dow Theory as applied by Schannep. Nonetheless, as we will see during the period evaluated Schannep's Dow Theory has also outperformed buy and hold, while drastically reducing drawdowns. Such an outperformance has been documented to exist and persist since 1953. Therefore, as with the "timing indicator" we are dealing with a very robust way of trading. This is why I have always been an ardent defender of "Schannep's Dow Theory". And the more I blog and invest, the more convinced I become about its robustness.

To compare apples to apples in terms of average return we have to carefully select our starting and ending point. As I wrote here, the proper way to compare the Dow Theory and BAH is to measure the performance from one primary bear market bottom to the last one during the period under observation.

In our specific test, we begin at 08/31/1998 (primary bear market low) from which point a new primary bull market swing started. We finish at 08/14/2019, date of the last primary bear market low. Hence, the rally (primary bull market) that got started after 08/14/2019 to date is not to be taken into account, neither for Schannep’s Dow Theory nor for BAH. 

All in all, our test spans 20 years, 11 months and 13 days. I feel this is a considerable long stretch of time to make a meaningful comparison between buy and hold and the Schannep’s Dow Theory. Please mind that when evaluating long term systems (with infrequent trades, as with the Dow Theory) to derive meaningful conclusions we need to take a sufficiently long time. Gary Antonacci in his great book (which every investor should read) “Dual Momentum Investing” cautions us against taking a too short amount of time when evaluating investing strategies. Hence, on page 119 of his book he writes:

“The diverging VR chart patterns show us that backtests performed on only 15 years of data are unlikely to give reliable forward-looking predictions. One of the most frequent mistakes I see others make is using only a limited amount of data, such as 15 years, to develop a model and then expect it to give reliable results going forward in time”
While Antonacci refers to backtesting, the takeway is clear: When dealing with investing strategies which do not generate many trades annually we have to take a sufficiently long time period. If we are going to compare BAH to Schannep’s Dow Theory, it is nonsensical to take, i.e. 5 years. This is too short a period of time. This is why I take solace in the Dow Theory (be it Schannep’s or the “classical”) as it has a sufficiently long track record which spans ca. 120 years for the “classical” and almost 70 years for Schannep’s. The risk of being the Dow Theory a statistical fluke is close to non-existent. Furthermore, we should bear in mind that the Dow Theory is “absolute momentum” and momentum has been documented to consistently work at least since 1800. This is an aspect that Antonacci has proven beyond any shade of doubt in his book. Hence, I feel we stand on solid ground. 

Both for BAH and Schannep’s Dow Theory we buy and sell on the close of the very same day when a signal is given. For BAH there are just two signals across the whole test: One to buy on the first day (08/31/1998) of the test and one to sell at the end of our test (08/14/2019).

I performed the test with the help of TradeStation®. To this end, using EasyLanguage® I coded the entry and exit dates for both buy and hold and Schannep’s Dow Theory. Coding Schannep’s Dow Theory signals was time consuming not only because of the number of signals but also because there were some entries which were done piecemeal. Hence, during the period under evaluation, we had several capitulations which resulted in fragmented entries across several dates (which entailed different weightings for each fragmented entry, as on capitulation day we just open commitments of just 25% of our equity). All in all, coding all the signals produced by Schannep’s Dow Theory has not been an easy task. I took due care to check and recheck for bugs. To the best of my knowledge I am offering my readers an accurate rendering of the performance of Schannep’s Dow Theory. However, any shortcomings are my fault, not Schannep’s whose description on his website of all his signals (and actual trading) is fully accurate, as it has been checked by this blogger truly yours many times over. 

I used a starting equity for both strategies of USD 10,000. In the case of Schannep’s Dow Theory profits and losses got reinvested. 

So, here you have the performance of BAH (according to TradeStation®’s performance report):


As to Schannep’s Dow Theory, here you have TradeStation® performance report:


Therefore, Schannep’s Dow Theory made a Total Net Profit of USD 25,383. Please mind that this is the money made on top of the starting capital of USD 10,000. Hence, the final capital was 35,383 USD. The CAGR for such an ending capital was of 6,03% .The CACR of BAH, as per TradeStation® was 5.21% with a Total Net Profit of USD 19,762 (and final capital of USD 29,762). Hence, in absolute terms Schannep’s Dow Theory outperformed BAH by 0,82% p.a. over the last almost 21 years. While 0.82% outperformance p.a. doesn’t look like much, it adds up, as BAH made over the period evaluated USD 5,621 less than Schannep’s Dow Theory. Furthermore, the outperformance was made while being less time in the market and enduring much milder drawdowns. Please mind that I don’t write “tested” as this is not a test but an assessment of what happened in real life.

Things do really get impressive when we look at drawdowns. The SPY experienced from 9/10/2007 to 9/3/2009 a hair curling drawdown of -56.47%, whereas Schannep’s Dow Theory (on a closed trade basis) just suffered a drawdown of -12.98%. From a peak to trough Schannep’s Dow Theory suffered a drawdown of -21.55%. Thus, any way you cut it, Schannep’s Dow Theory did an outstanding job containing drawdowns. And, please believe me, containing drawdowns is the essence of trading. As legendary trader, featured in the first “Market Wizards” book, Paul Tudor Jones said the “most important rule of trading is to play great defense, not great offense”. Keep the powder dry when the going gets tough, and the market itself will take care of the upside when the bear is over. Most investors neglect drawdown protection until it is too late. Forget performance, think first of how are you going to deal with market turmoil and how are you going to keep as much as possible of your capital safe.

From the performance report we can see that Schannep’s Dow Theory was 73.06% of the time in the market. This is quite in line with the ca. 30% of the time which historically the Dow Theory has been out of the market. The important point is to realize that we have been able to outperform buy and hold while being less time in the market (less risk and freed capital for other investments) and by more than halving drawdowns. 

Another interesting statistic is the average duration of the trades taken. Winning trades lasted on average 241.86 days (a little bit less than the average of the past). Losing trades lasted on average 64.20 days. The average duration is shorter than the historical average (more about measures of historical performance here). This is due to trends that in most instances have had less persistence than those seen in the past. In other words, buy signals have encountered a sell signal earlier than in the past, since the advance following a primary bull market signal was been more modest.

I quote an important post written in 2017 which clearly identifies the underlying cause of the shorter duration of trades and, more importantly, proves that even under more noisy conditions, Schannep’s Dow Theory remains profitable and is likely to continue outperforming BAH in the future:


“From 1954 to 2008 the average further advance following a primary bull market signal amounted to 37.98% (median 27.03%). In other words, there was ample bullish action before the primary bull market signal fizzled out. Since we know that, on average, following a primary bull market top, the primary bear market signal (exit) is flashed at ca. 7-8% below the market top, there was an ample margin (37.98% upswing versus 7-8% downswing leading to the exit) to have profitable trades.


What has happened from 2009 to date? Well, as with subsequent declines following primary bear market signals, subsequent advances following primary bull market signals have been muted. My spreadsheet tells me that from 2009 to date, the further advance following primary bull market signals amounted to a mere 12.16% (median 7.6%). Hence, the subsequent advance has been cut in roughly three (37.98% versus 12.16%). No wonder we see “fibrillation”. Sell signals lack follow through and buy signals are likewise afflicted by the same ailment. This results in:


            a) More whipsaws.


b) Shorter duration of each trade (thus, we are suffering trades which last only 2-4 months versus more than 1 year (which used to be usual).


            c) More trades.


However, it is important to note that even under such sputtering bullish action, further advances following a primary bull market signal of ca. 12% are sufficient to result in modestly winning trades. We know that exits tend to be between 7-8% below the top. If following a primary bull market signal we make 12% and from that point we lose 8% we still have ca. 4% left. Of course, this is far from stellar returns, and far from the smooth ride which the Dow Theory tended to give us in the past (less than one trade per annum, 4% outperformance versus buy and hold). Now, we have to suffer more trades and subpar performance (though not losing one’s shirt)


The million question is: Will the market action we have been seeing since 2009 be the new normal in the future? Or it is just an aberration? As I wrote here, I feel we are seeing an aberration, albeit a very painful one in the short run (“short run” for long term trading methods like the Dow Theory means quite a lot of time, since we are not day trading, and a couple of bad trades may take quite a lot of time to be reversed) for those not intimately acquainted with the Dow Theory. Periods of even 10 years of underperformance versus buy and hold are not a rare occurrence. Furthermore, I wrote here that on any given year, the odds favor the Dow Theory underperformance.”


Please also mind that there was a flat period of 728 days. This means that Schannep's Dow Theory was out of the market (flat) for such an amount of time during the 2000-2002 bear market. So Schannep's Dow Theory managed to emerged unscathed from such a devastating bear market. In 2008, Schannep's Dow Theory was a good deal of time out of the market, albeit not completely flat the whole time. This resulted in taking a small hit of -12.98% (see above).

From the Performance Report we can also glance that the ratio between the average winning trade and average losing trade stands at 3.25. This ratio means that on average, when we win, we make 3.25 more money than that lost when there is a losing trade. If we couple this ratio with an almost 60% winning trades and a profit factor of 4.55 (namely total gains are 4.55 greater than total losses), we can say that we are dealing with an excellent trading system. Very few trading systems achieve such impressive performance figures.

One interesting measure is the sharpe ratio (the higher the better). For Schannep’s Dow Theory it was of 0.16 for the Dow Theory it was of 0.08. The higher sharpe ratio tells us that on a risk-adjusted basis (risk to endure in order to make profits) the Dow Theory is a better way to invest. 


Another interesting chart is that of monthly accumulated profits. Let’s first take a look at that of BAH. As you can visually see drawdowns were long (in time) and deep. The equity lows of 2009 were below those of 2002. Not so a pleasant picture.

Monthly accumulative for BAH. Two hair-curling drawdowns
 
Now let’s take a look at the cumulative monthly performance of Schannep’s Dow Theory.
 
Schannep's Dow Theory: The smooth ride and drawdown contention strikes the eye

The flat lines depict periods were we were out of the market. As you can see it has been a quite smooth ride. Declines have been much milder, and, notably, the depth thereof has been much more mitigated. The chart speaks for itself. 

You may be thinking that the outperformance of the Dow Theory is not so remarkable, that most likely any trend following method would have delivered similar results. Well, you'll be disappointed. I performed a test with an exponential moving average with the Industrials and the Transports and it clearly underperforms Schannep's Dow Theory and even BAH (even though drawdowns get reduced by ca. half, so on a risk-adjusted basis even the modest moving average does quite a decent job). In my next post of this saga, you'll read about the specific figures.

Conclusions:



·         Schannep's "timing system" outperformed BAH by 4.19% p.a.


·         Schannep’s Dow Theory outperformed by 0.82% p.a. while achieving dramatic drawdown reduction. 


·         Schannep’s Dow Theory was less time in drawdown.


·   Looking forward it is very likely that Schannep’s Dow Theory outperformance will widen and be in the vicinity of ca. 4% over BAH, while continue keeping drawdowns at bay. Why? Read my post of 2017 linked above.


·         The use of moving averages, even though confirmation slightly improves results, falls greatly short of the performance achieved by the Dow Theory. More about it in the next post of this saga.
 
Sincerely,
One Dow Theorist

No comments:

Post a Comment