Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
I've been interested in finding longer-term and lower-risk ways to trade CL besides doing singles outright. So, I've been testing bull/bear calendar spreads in paper money, and while I've been right over the relevant timeframes, they have not been profitable, and I've found that the behavior of the spread does not necessarily track the underlying. I was not able to find the resources I was looking for to help me really understand the inner workings, so I thought I'd just backtest the spread using daily settlements.
My analysis is ongoing, but my preliminary results do seem to be interesting. I thought I'd share them here. There's a lot more to do, but it is a work in progress.
Dataset: Front: Nov 2014 CL; Back: Jan 2015 CL. Daily settlement values on all days with data available for both spreads (starting Nov 2009). 1,237 daily settlements.
Hypothesis: If I'm bullish on crude, the front month will increase faster than the back month, so you go long the front and short the back. It's the opposite for bearishness. However, while I've read this in multiple places, reality has not quite matched up with theory, from what I've seen, and my gut says that there's a lot more to it: What if the spread is already wide? What if the market is in backwardation? Does the size of the move of the underlying correlate with the size of the change in the spread?
Summary of Data: I looked at 1,237 trading days. Of those, the market was in contango for 379 days and backwardation for 858 days. The front month had 526 up days, and 471 down days (excluding outlier days, discussed below).
Initial Observations: Looking solely at day-over-day changes in settlement value, if the front month went up, a bullish spread (long front short back) would increase roughly 70% of the time (slightly less when in contango). That means that roughly 30% of the time, you'd place the "right" bull spread, and you'd be right about the front month increasing, but you'd lose money.
Still looking at a bull spread, when the market was in contango, the average up move in the front month was bigger than when in backwardation, but the average change in the value of the spread was almost halved (thus, when in backwardation, a smaller move was potentially more profitable than a larger move when in contago).
On the bear side (front month supposedly goes down faster than the back month), the results were fairly consistent with the bull side.
Outliers: There were six days where the front month did not change, but the spread changed. Too small a sample to be of interest right now. There were 233 days where crude moved, but the spread did not. Interestingly, this was a bit less than twice as likely to happen when the market was in contango than when it was in backwardation.
Results: Cropping the vertical axis to enhance visibility of the spread (red line) Same plot, but without cropping Zooming in to just 2014
Next Steps:
These results are not very helpful, and while I had hoped that the plots would reveal something useful in terms of trends and correlation, simply plotting price movement against the spread is clearly not the best approach. Also, these results don't provide much context, ie, if the spread is narrow and the market goes up, will the spread widen? Does the width of the spread as of the prior day affect the range of the spread on the next day? (Does the width affect range?)
In Jan 2009, the difference between the Nov 2014 and Jan 2015 contracts was small. This is futures curve 101 stuff, so presenting the changes in the spread so far out is really pretty silly. I should instead grab the front contract for each respective month, and examine those changes, instead of just grabbing two contracts from issue to expiry.
I re-ran this analysis using the final 30 settlements of each front-month contract, thus getting close to a continuous contract from 1990-2016. The spread is the front-month (long) minus the front-month+2 (short). Thus, if the current front contract month is September, the back month for the spread would be November.
Here are the summary results; the plot follows, though the plot is messy.
Public Data Source: Quandl (free resources)
My Database: MS Azure (SQL)
Script to load data from Quandl to my database: custom C# script
Graphing: R
I used R for the plotting, but I did most of my data analysis directly in SQL, as I'm more familiar with SQL than with R. Wrong tool for the job, I'll admit, but time was an issue. Azure and C# might not be the most sensible tools for this type of thing, but I'm the most comfortable there, so I used them.
Trading: Primarily Energy but also a little Equities, Fixed Income, Metals and Crypto.
Frequency: Many times daily
Duration: Never
Posts: 5,049 since Dec 2013
Thanks Given: 4,386
Thanks Received: 10,207
Have you tried Mth 2 vs Mth 13 and see how that moves compared with Mth2 daily change.
The Mth 1- Mth 2 spread can move can be very heavily fundamentally drive.
