Dangerous Nonsense: Market Timing

Using a simple system it is said to be possible to achieve a better risk adjusted return than would be achieved by buy and hold. Maximum draw-down may be less severe and standard deviation may be lower.

Much is made by the retail investment community (and indeed others who should know better) of the advantages of algorithmic market timing based on limited testing over limited timeframes and over an even more limited number of instruments.

Systematic market timing only works in the aggregate, over a wide variety of instruments and over a long time frame. I am very aggrieved when I see articles which use a small number of instruments, conduct very limited testing and conclude that “it works”.

Asset allocation “schemes” are amongst the worst offenders: commentators perform limited tests on 4 index trackers and conclude that applying some simple system will soundly beat the performance of every fund manager (and his uncle) who has ever lived. Often such schemes use a single fixed monthly date to re-allocate between the sectors. Often they tout the sort of performance metrics which may be achieved in the long term by one manager in a million, and then perhaps as much by chance as anything.

It is nonsense, pure and simple.

Much the same applies to “trading your equity curve”. The equity curve (or output) of a systematic investment strategy can itself be “market timed”. In the same way that you can exit the S&P 500 when it dips below the much vaunted 200 day moving average and re-enter when it re-crosses on the upside, you can stop and start your own systematic trading on the same basis.

But much the same considerations apply. If you have a single trading method, it’s going to be a toss-up: sometimes it will work sometimes it won’t. If you trade a large number of systems you stand a better chance of benefiting overall from such a technique.

I will give one small example. The results set out below represent a monthly momentum system back tested on a portfolio of 1,000 US stocks for the period 1st January 1997 to date. The system used was a 20 stock enhanced version of the Smart Beta Stock Momentum System (https://anthonyfjgarner.net/quantechinvestments/smart-beta-stock-momentum-strategy/). The system used for each test run was identical in all respects except for the rolling date upon which re-allocation took place. “Without cut out” represents the equity curve without using market timing. “With” represents the identical system but trading ceases when the 65 day momentum of the equity curve goes negative and re-commences when it re-enters positive momentum over the said look back period.

As can be seen, there is surprising variance in the results when you bear in mind that the only difference in the 31 test runs is the re-allocation date. Nonetheless there is evidence that some advantage may be gained by applying a stop/go form of market timing if it is applied to enough different systems/equity curves.

Clearly this very limited series of tests proves nothing. But it may encourage further research.

System CAGR Max DD Av Daily DD Median DD Std Dev Day System CAGR Max DD Av Daily DD Median DD Std Dev Day
Without Cut Out

33.60%

-50.90

-11.18

-5.62

22.66%

0

With

29.78%

-34.48

-9.72

-6.99

19.31%

0

Without Cut Out

33.85%

-54.26

-12.15

-5.92

23.48%

1

With

30.29%

-36.06

-9.87

-6.68

20.06%

1

Without Cut Out

30.05%

-61.35

-16.68

-9.55

23.69%

2

With

24.38%

-52.05

-17.44

-12.76

19.94%

2

Without Cut Out

30.78%

-48.94

-11.08

-6.09

23.00%

3

With

28.21%

-37.55

-10.02

-6.61

19.42%

3

Without Cut Out

29.80%

-52.89

-13.42

-8.48

23.57%

4

With

27.52%

-48.90

-14.08

-7.77

18.96%

4

Without Cut Out

31.80%

-42.43

-9.80

-5.59

23.83%

5

With

27.16%

-44.88

-12.67

-7.51

19.31%

5

Without Cut Out

31.43%

-41.95

-10.53

-5.79

23.71%

6

With

26.23%

-43.42

-12.34

-7.56

19.18%

6

Without Cut Out

28.48%

-55.79

-12.97

-7.00

23.91%

7

With

24.23%

-40.13

-11.35

-7.63

19.27%

7

Without Cut Out

32.73%

-40.94

-10.54

-6.22

23.93%

8

With

27.26%

-43.26

-12.09

-6.89

19.48%

8

Without Cut Out

29.07%

-44.07

-11.21

-7.14

23.48%

9

With

25.08%

-43.66

-12.12

-7.50

18.84%

9

Without Cut Out

35.11%

-40.01

-8.50

-4.48

21.65%

10

With

30.77%

-29.51

-7.55

-5.39

18.23%

10

Without Cut Out

35.20%

-39.67

-9.26

-5.15

23.71%

11

With

28.89%

-44.32

-11.65

-6.72

19.69%

11

Without Cut Out

35.93%

-40.38

-9.81

-5.24

23.57%

12

With

30.02%

-46.05

-11.26

-6.73

19.29%

12

Without Cut Out

33.00%

-46.47

-11.45

-6.46

24.24%

13

With

28.68%

-48.41

-12.84

-8.18

19.68%

13

Without Cut Out

35.53%

-41.29

-9.62

-5.85

23.50%

14

With

31.63%

-27.22

-7.03

-6.33

19.65%

14

Without Cut Out

35.94%

-42.21

-9.92

-5.78

23.95%

15

With

33.78%

-26.18

-7.36

-5.44

19.89%

15

Without Cut Out

36.52%

-43.61

-9.16

-5.01

23.38%

16

With

27.42%

-43.25

-10.44

-8.11

19.52%

16

Without Cut Out

35.17%

-51.52

-10.06

-5.34

23.68%

17

With

28.79%

-38.85

-9.61

-6.35

19.50%

17

Without Cut Out

30.56%

-52.50

-11.02

-6.57

22.61%

18

With

25.10%

-36.68

-10.21

-8.15

18.13%

18

Without Cut Out

32.60%

-36.76

-9.74

-6.14

23.41%

19

With

28.82%

-24.19

-6.71

-6.39

18.94%

19

Without Cut Out

33.68%

-44.83

-10.50

-6.11

23.23%

20

With

27.58%

-35.85

-9.43

-7.57

18.93%

20

Without Cut Out

34.17%

-50.72

-12.01

-6.98

24.49%

21

With

28.67%

-48.50

-14.45

-9.70

19.53%

21

Without Cut Out

32.44%

-55.85

-13.36

-6.90

23.62%

22

With

26.61%

-47.24

-13.03

-7.91

19.00%

22

Without Cut Out

30.88%

-48.65

-13.07

-8.12

23.67%

23

With

24.30%

-50.59

-15.05

-10.50

19.28%

23

Without Cut Out

31.16%

-50.95

-12.99

-7.71

23.75%

24

With

23.50%

-56.05

-16.72

-9.50

19.06%

24

Without Cut Out

31.38%

-44.98

-10.71

-5.88

22.93%

25

With

24.00%

-48.01

-14.22

-8.63

18.52%

25

Without Cut Out

29.69%

-59.51

-14.41

-7.97

23.13%

26

With

25.77%

-50.48

-14.40

-8.69

18.96%

26

Without Cut Out

32.52%

-45.09

-9.66

-5.65

21.80%

27

With

26.16%

-36.80

-10.73

-7.41

18.22%

27

Without Cut Out

32.13%

-49.77

-11.25

-6.22

22.86%

28

With

27.76%

-45.73

-11.19

-6.21

18.97%

28

Without Cut Out

33.46%

-44.26

-11.10

-6.18

23.74%

29

With

27.09%

-48.04

-12.94

-8.68

19.28%

29

Without Cut Out

31.10%

-45.64

-12.58

-6.93

23.75%

30

With

27.11%

-41.70

-11.52

-7.47

19.44%

30

Without Cut Out

31.42%

-51.72

-12.17

-6.66

24.23%

31

With

27.99%

-47.37

-12.24

-7.29

20.27%

31

Average

32.54%

-47.50

-11.31

-6.40

23.44%

  Average

27.52%

-42.04

-11.63

-7.66

19.24%

 
Categories: Tags:

7 Comments

  1. Hi Anthony, can you give a little more detail about your example system? Is it holding 1000 US stocks each month equal weight without the cut out? Is this the Russell 1k or some other index? Thanks.

    Like

    1. Hi Aaron, yes certainly. The system is this one: https://anthonyfjgarner.net/quantechinvestments/smart-beta-stock-momentum-strategy/ using 20 stocks. IE rotating into the top 20 each month subject to the rules and filters. With/without cut out are identical (except for the cutout!). The 1,000 stocks is the total POTENTIAL portfolio from which the 20 are drawn and includes both currently listed and delisted stocks. The portfolio was chosen on a purely algorithmic basis.
      A

      Like

  2. Not sure that the “retail” community that we’ve observed over the years has been touting any advantages of market timing. The familiar faces of J. Bogle and passive-investing advocates have become a behemoth force in the industry. We think that this will backfire as the expected returns of these products and methodology won’t be able to provide sufficient compounding and growth to the average worker’s “underfunded” retirement asset accumulation ( see first link below ).
    Also, we would disagree about systematic timing working only with a “wide” variety of instruments. We have found it easier ( and less work ) to utilize the SPY, QQQ, and VBR index etfs in tactical asset allocation models, rather than with individual stocks, index futures contracts, or sector etf methods, and still provide commensurate results and risk mitigation.

    https://docs.google.com/presentation/d/1Ua-R53o7c588nUr705hY4YaBEcG1qOrNvtonQIwpVws/edit?usp=sharing

    http://stockmarketmap.wordpress.com/

    Liked by 1 person

    1. Thank you for your comments and the links which I will read with great interest. Comment and disagreement is always welcome and a good way to see other people’s ideas and points of view. Bogle is, in my opinion NOT a passive investor, since an index investment is a trend following procedure in disguise – it invests in growing companies and dis-invests in shrinking, failing companies. As you know I am sure, the majority of listed stocks disappear at a loss. See Eric Crittenden’s excellent research. Fundamental investors try to pick growing companies and ditch losing companies; so do indices and passive investors; so do traditional trend following strategies.

      In all of these cases you get a second or third or 4th chance: if the stock bounces back your “system” (be it an index, a trend following system or mere discretion) is likely to have you re-invested.

      At a time when quantitative analysis looks set to rule the world, it would be a foolish person to deny the validity of systematic investment.

      However I strongly disagree with using a handful of ETFs. I am not suggesting one has to use stocks or futures either. Merely that the handful used by most people will get them into trouble.

      In my daft little book I pointed outlined the simple 200 day moving average system on the Dow going back to the late 1890s. See page 3 if you happen to have a copy. All worked swimmingly (including 1929) until recent years when “something changed”. That “something” may well revert but the point is, it is better to use several ETFs for US exposure not one, quite a few more for international equity exposure and indeed quite a few more for each asset class you want to cover.

      Sure, less work to use fewer but I believe the long term curve will be a lot smoother and more reliable if you don’t restrict yourself in this way.

      But that is simply an opinion.

      Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s