.jpg)
.
In thinking more about the post a couple of days ago examining the seasonality of gasoline prices, I was curious how profitable a simple seasonal trading system would be if one could have traded gasoline prices. Keep in mind, there was not an easy way to trade gasoline prices until last year when UGA, the gasoline ETF, was made available. I will take a look at UGA closer at the end of the post.
Basic Trading System Simulation:
- Go long on gasoline in January
- Close the long and go short in Sept.
- Close the short in January and go long again... and so on...
- Start with $10,000 and invest the entire portfolio all the time either long or short.
- Unrealistically assume no slippage (transaction fees, trading fees, etc.)
This trading strategy would have made a whopping 25% compounded annually. There were a total of 37 trades over the span of 18 years, with 78% of those being profitable and the worse trade was for a loss of 8%. Of course this is an idealized, backfit, overfit, impossible system. I would look at this as the best this could have done in the best of worlds. But it still is interesting. The biggest reason this is impossible is because up until last year when the gasoline ETF - UGA got introduced, there would be no way for an individual investor to play this anyway even if they wanted to.
How well does UGA, the gasoline etf, correlate to gasoline prices?
So now that we have UGA that can be traded, and a historical seasonal pattern with gasoline prices the next question is how good could one expect this approach to work with UGA? This plot shows the weekly prices for U.S. average regular gasoline and UGA:
As you can see above, there is a strong overall correlation between UGA and regular gas prices as one would hope. It is not exact which is for many reasons: UGA technically follows front month U.S. gasoline futures, not actual average current gasoline prices, UGA is a tradeable security with all the liquidity that means which also means more volatility, etc. Notice that it looks like UGA leads actual gasoline prices a bit, by about one data point which is a week. This fits the common wisdom that the equity market tends to lead reality. But overall, the fact that the two follow nicely would say that perhaps you could trade the gasoline seasonal effect with UGA which is a good thing, especially since the seasonal trading strategy is a long term strategy over months so the day to day variability in UGA ought to average out.
For the Stat-Geek's: One more look at the lag relationship of UGA and gasoline
Below is a lag cross-correlation function graph. This shows the strength of the correlation between gasoline price and the UGA etf at varying lags between the two series. So what you see is a peak correlation when UGA is offset by one week from gasoline prices. In this case, that one week is saying that UGA is a leading indicator, it predicts gasoline prices with a one week lead time. The cross-correlation function graph is another real nice time series analysis tool.
Where to next? I would like to throw crude oil and the various crude etf's into the mix of this analysis although I am quite sure there will be alot more variability in crude and the crude etf's than gasoline and therefore any seasonal or time series forecasting will be more difficult. Then I will also return to gasoline prices and try to develop a more complex time series model beyond just this simple seasonal model I have done so far.
I guess I need to add a disclosure/disclaimer now to any posts that have to do with trading and investing...
Disclosure: I have no positions in any of the stocks or funds mentioned but I may in the future.
And please see the disclaimer.
...
7 comments: