(This article was first published on Commodity Stat Arb, and kindly contributed to Rbloggers)
I was reminded recently it has been a while since my last posting. So here is a quick analysis of a strategy that works pretty well.
Much has been written on the changes to the NG market structure over the past few years; the influence of financial investors and the impact of new technologies. One thing that hasn’t changed is the reliable earning power of spread trading.
The complexity in modeling returns from futures trading, particularly spread trading means that the best way to characterize the profitability of these types of strategies is in terms of $/contract. This can then be directly compared to the $/contract of margin requirements for a reasonable perspective on the investment return. For this analysis, I used recent data from the CME website for initial margin by strategy type for the NG contract:
But first I present the $/contract return side of the equation for some standard spread strategies. These analyses are all based on rolling a XY position in NG futures, with the contract roll being on the penultimate day available. In the usual convention X is the nearby contract and Y is the further out contract. The ratio is 1:1 in contracts. Other roll alternatives may make more sense but this is conveniently tradeable given the TAS contracts available. Other hedge ratios may be more statistically defensible; 1:1 has the advantage of being available as a directly tradable spread (also via TAS), and the analytics is trivial.
Looking at the 5 year historical performance of N(N+1) spreads we see the following results:
spread

Daily avg

Daily stdev

Annualized
avg/stdev

12

13.27

232.51

0.90

23

4.89

172.79

0.45

34

10.81

128.42

1.33

45

3.90

126.33

0.49

56

1.30

97.10

0.21

67

5.48

108.67

0.80

78

9.22

123.99

1.18

89

17.55

179.98

1.54

910

2.78

129.50

0.34

1011

0.96

123.55

0.12

1112

7.32

755.13

0.15

13

18.16

193.36

1.48

12 (34)

24.08

232.20

1.64

This table shows the average daily return and standard deviation in $/contract. So for example, holding a long prompt contract and short 2^{nd} month contract since January 1, 2007 made an average of $13.27 per day with a standard deviation of $232.51 per day. However on an annualized basis this translates into $3,317 and $3,676 respectively or a profit to risk ratio of about 0.9. Not great, but a nice diversified addition to an existing strategy.
The highest individual 1 month spread contract return is in the 34 contract. Specifically being short this contract, which yields a profit to risk ratio for the year of 1.33.
By extension, these strategies can be combined relatively easily:
Long 12 plus short 23 (which is equivalent to long 13) has a 1.48 profit to risk ratio.
Long 12 plus short 34 has a 1.64 profit to risk ratio.
What is nice about these strategies is that the capital (margin) commitment is not immense. An individual spread requires less than $1000 per contract of initial margin. Drawdowns have been limited in the past 5 years as the cumulative profit chart shows:
Overall, this is a nice uncorrelated strategy to add to a portfolio.
To leave a comment for the author, please follow the link and comment on their blog: Commodity Stat Arb.
Rbloggers.com offers daily email updates about R news and tutorials on topics such as: Data science, Big Data, R jobs, visualization (ggplot2, Boxplots, maps, animation), programming (RStudio, Sweave, LaTeX, SQL, Eclipse, git, hadoop, Web Scraping) statistics (regression, PCA, time series, trading) and more...