Thursday, April 30, 2009

Seasonal trades in natural gas and gasoline futures

In my book, I mentioned 2 seasonal trades in natural gas and gasoline futures that have been consistently profitable for 14 years. (Mentioned here and here also.) And not only in backtest: I paper-traded them in 2006, and actually traded them in 2007-8, and all 3 years were profitable. How did they fare in this recession year? Quite poorly.

Depending on your exact entry and exit points, the gasoline trade lost about $2,500 per contract of RB. The natural gas trade lost about $7,700 per contract of NG.

You may have heard that natural gas price is at a 6-year low. In fact, we are not seeing any increase in industrial demand for natural gas. Apparently, somebody has forgotten to tell the nation's industrialists that an economic recovery is supposed to be under way.

Will I enter into these seasonal trades again next year? You bet I will.


Anonymous said...

Hello Ernie,
Do you have any position regarding the cointegration of Crud (CL) and Heating Oil (HO) futurs?
Without much investigation, this pair seems to be working nice historicly.

Paul Teetor said...


According to my historical data, CL and HO are indeed cointegrated; that is, their spread is mean reverting. I used an Augmented Dickey-Fuller (ADF) test, and the p-value was only 0.01, indicating that, statistically speaking, there is only a 1% probability that the spread is not mean reverting.

The ordinary least squares (OLS) fit suggests a ratio of 1.2 CL contracts for every 1 HO contract. You can't trade a fractional futures contract, of course. CL is strongly correlated with USO stock, however, so I would suggest padding your 1-contract CL position with enough USO stock to keep the right hedge ratio. Otherwise, bad things could happen.

At the moment (5/1/09), the spread is high, so the indicated trade is short CL and long HO. But that could change quickly!

I also see some mild seasonality in the spread, but not during the Summer months. Wait until December for that.

Hope that helps.


ashutosh kulkarni said...

Hey Ernie,
Natural Gas is definatley in demand, right now according to bloomberg Korean consumption has halved but those spot cargoes are lapped up by the hungry indians. Anyways, look out for Indian RNRL (reliance refineries) to churn out LNG enough for the country' deficit but might releive international pressure ....aah but 6 -9 months to go for that till then happy betting .


Anonymous said...

Thank you Paul for the comprehensive answer.
By the way, when you do your OLS do you use the price or the price Log?
Using the price Log I get a ratio of 1.04 HO to 1 CL.
(daily data starting Feb-05).
Other than that I agree that spread suggests to short CL. Once the spread is back to zero the return will be (as of 5/5) 11%.


DW said...

I'm not too familiar with oil and gas futures so apologies if this is a silly question but does the hedge ratio mentioned factor in different tick values for the two contracts. From what I looked up on the NY Merc, HO has a tick value of $4.2 and CL is $10.

Yona said...

Some more reading on seasonal trading: Condor Options

Unknown said...

With only 12 observations in the back test, how can these nat gas & oil seasonal trades be considered robust, or am I missing something?

Ernie Chan said...

The main strength in these oil and gas trades are that they are partly based on fundamental reasons. So the backtest are just confirmation of these reasons. In fact, all seasonal trades suffer from the sparse data problem if one is merely looking for technical rules.

J.F. said...

Ok... just for yuks...
Mr Teetor found that CL and HO are co-integrated. So I figured, what about ETFs based on those futures?
USO, USL and UNH are ETF's based on futures contracts of crude (USO and USL) and heating oil (UNH). Even though the spread of USL vs UNH looks mean reverting visually, it doesn't bear out statistically.
Using the R package tseries, I used the ADF test (adf.test) and the "Kwiatkowski-Phillips-Schmidt-Shin test" (kpss.test, huh?? don't ask me... I'm not a statistician... I just figured two tests are better than one). Both tests indicate all three series have a unit root (i.e. are a random walk). However, the tests also detect unit roots in the residuals of least mean square regressions between pairs of the USO, USL, and UNH series, i.e. none of the series are co-integrated with each other. In addition, utilizing the Johansen test ( in the urca package) also indicated no co-integration.

Like I said... a long shot that didn't pan out, but an interesting exercise.


Ernie, I loved your book. Keep up the good work.

Paul T, I read your letters and comments they generated on the R financial mailing list on co-integration. Very informative.
Keep up the good work, as well.

nikkus said...

I have been working my self for many years on seasonal patterns in futures and currencies (though the last is my favorite), I have found some really good and attractive patterns. Nevertheless I always failed in making those strategies systematic (criteria of capital allocation as part of a portfolio being the major hurdle, ranking of strenght of one particular pattern in comparaison to another one, filtering weak patterns....). So far I have been putting the emphasize on fundamental explanation or specific flows explanation if any, so my choice remained heavily discretionary. any book or papers to recommend on the above?

Ernie Chan said...

For portfolio allocations, I always recommend Kelly formula, as explained in my book Quantitative Trading or on