As part of the Market Study Bootcamp series, today we’re placing the NYMEX Light Sweet Crude Oil futures contract (symbol: CL) in the spotlight. We’re going to try to get to know it a little better in order to see what makes it, er, tick. Get it? Tick? Ahem.
- How frequently does it swing?
- What does the distribution of swing magnitudes look like?
- How long do the swings take?
- When in the session are the swings occurring?
Ok, let’s get going…
We’re going to examine the intraday swings of CL using the data export features of the Pro Edition of the Acme ZigZag study. The study setup is pretty simple:
- Load one year’s worth of ETH bars using the NYMEX Metals/Energy ETH session.
- Apply the Acme Zig Zag to the chart with the swing export feature turned on.
- Fire up Excel and examine the results.
If you Pro-level members are interested in playing along at home, here are the specifics:
- CL ##-## (continuous contract from DTN IQFeed, but you can use the dated contracts too)
- 5 minute bars
- 365 days loaded
- Indicator settings:
- Swings – Magnitude parameter set to 0.25. This value will catch practically all the swings of any significance while not over burdening the data with each and every tiny vacillation.
- Swings – Use Highs and Lows parameter set to true. This will capture the full range of all the swings.
- Swings – Measurement Type parameter set to Points for the sake of mathematical simplicity.
Finally, here’s what a chart set up like the above will look like, more or less:
[box style=”rounded”]NOTE: the graphs below are very wide. Please stretch your browser as wide as possible before clicking on them for maximum detail.[/box]
And Now…. The Data
[iconbox title=”Symbol: CL” icon=”x-office-spreadsheet.png”]
- Swings in sample: 5379†
- Date Range: December 6, 2011 – December 6, 2012
Let’s start with the first question first. How often does the CL swing? Below we have the distribution of swings in our sample:
On the x-axis we have the range of the number of total swings per day, ranging from a minimum of 1 swing to a maximum of 46 swings in a day, and on the y axis we have the number of days in which that number of swings occurred. It’s a little confusing, perhaps, but here’s a for-instance that’s pictured above. At the right extreme of the x-axis we have the number 46, which registers a frequency of 1 on the y-axis. This means that there was only 1 day in the sample where we had 46 swings. On the extreme left, we have 7 days which had but 1 swing. The rest of the data in the sample reads the same way.
So let’s give this graph a little color, shall we? Check out that fat middle! This histogram, if you tip it on its side, has one heck of a Buddha belly. What is this telling us? It’s not a normal distribution, but if it were one standard deviation would include all frequencies from roughly 15 – 34. But again, what’s this telling us? It’s telling us that the CL is one well-oiled rotation machine. It zigs and zags very frequently on most all days and should, theoretically, provide lots of opportunity to catch reversals. The most frequently seen number of swings in a day was 24. Twenty four!
But swing frequency is not enough. We must know more. So let’s now turn our attention to our second question – the swing magnitude distribution?
This time on the x-axis we have the number of points of all swings in our sample, and on the y-axis we have the number of swings of that magnitude/size. What’s most noteworthy is the clean, near-linear distribution of swings. One way of interpreting this graph is that, despite the fact that it moves around a lot, the price action in the CL is actually pretty orderly. This is exactly the kinds of distribution we’d expect in a situation where organic market forces are in play. We’d expect to see a graph dominated by frequent small swings with a steadily-decreasing frequency of big swings – the kind that make home runs and widows. ;-}
You can agonize over stats all you want, but I find it most useful to employ the Pareto Principle when analyzing this kind of data. That means I’ll focus my attention on exploiting the most probable 80% of swing magnitudes. Thus, we should plan on our most probable rewards being between about 25 and about 60 ticks.
On a bit of a side note, one way you can check how well you’re executing your own plan and the quality of your trading rules/guidelines/axioms/principles (please tell me you have some!) is by comparing your actual trading results to this kind of a distribution. Say you’re trying to execute a Pareto-based plan like the one above. If your actual trade outcomes for the CL are not very similar in shape and magnitude to this distribution, you might consider how you could improve you plan, your execution, or both.
Because if you planned your executions perfectly, executed your plan perfectly, caught every swing flawlessly like a time-machine backed trading robot with 20/20 hindsight and a perfectly clear crystal ball – your trading outcomes should look just like this distribution. In other words, this distribution is a kind of performance benchmark. This is the best you could have possibly done. Any deviation from this distribution is the proverbial room for improvement. So, if when you’re at bat, you’re only tapping the slow pitches or only the nailing big ‘uns, maybe you need a little more pepper on your swing, or maybe a little less. ;-}
Now for our 3rd question. How long do the swings take?
Now on the x-axis we have the number of minutes of all swings in our sample and on the y-axis we have the frequency of swings of each duration on the x-axis. Here we have a undeniably clear pattern – the CL’s swings are happening early as well as often. The overwhelming bulk of swings in the CL take between 5 and 30 minutes to complete. If you’re trading CL on the Pareto Principle Plan described above, you’re not going to be in your positions very long. The CL is indeed a guerrilla trader’s delight. Hit ’em fast, harvest them often and get the hell out.
Last and most certainly not least comes our 4th and final Big Question to pop – when exactly is all this action taking place?
On this last graph’s x-axis we have the times of day that a swing started or ended, and on the y-axis we have the number of swings that commenced or concluded at that time. For our purposes in this study, we don’t really care whether it was the beginning or end of a swing. We’re only concerned with understanding how this instrument presents turning points/opportunities.
Here we see, not surprisingly, that the swings most often occurred in the first 2 hours of the RTH session, and then again at the end. No doubt about it, there’s were your opportunities lie. So if you’re on the Left Coast (like I am) you better be a morning person. The RTH open comes pretty early for us. In any case, you probably knew about the timing of opportunities already, maybe intuitively, but there it is very literally in black and white. No guesstimation required.
So now that you and CL are through with the gettin’-to-know-you-chit-chat, is it time to take your relationship to the next level? If so, some questions you might want to ask could be aimed at discovering, say, the top 3 or 4 most frequent (probable) swings by time of day, size and duration, then work at learning to recognize and exploit those opportunities as they unfold, as best as you possibly can.
I leave it to you, gentle readers, to try lifting that particular kimono on your own. Until the next instrument spotlight… trade ’em well.
† Note: 45 swings with a duration of over 1000 minutes were dropped from the sample after the export. These were nearly all due to swings which bridged weekends and holidays. These represented only 0.0008 of the sample size and therefore will have essentially no impact on the study results.