Correlations with the NYSE TICK

UPDATE: I completely forgot one version of the graph when I posted it. I only included the absolute value histogram and not the real value version. Could be the cough syrup. This is why I didn’t trade today. ;-}

As many of you know, I am a big believer in watching market internals for keeping track of intraday context. I’ve found them, and the NYSE TICK in particular, to be a great help in confirming my read on an upcoming market move. I’ve been watching it for years, and over time I began to become aware of how correlated its moves were to to the futures markets and the ES in particular.

I’ve had a study for a while that tracked the correlation of the NYSE TICK index to the ES, and since today was a “sick day” for me and I didn’t trade, I thought I’d update and publish it.

If you’re not familiar with the TICK index, here is a brief explanation.

In this study, we’ll look not only at the correlation of the ES to the NYSE TICK index, but as an interesting side effect, we’ll get some insight into which times of day each instrument is most on the move.

This information can be very useful in anticipating what times of day a move might be made, and it has the potential to keep you out of “crap trades” at the very least. At most, it can factor into your risk management or position sizing rules. Why? How? Simple. Volatility will influence your expectancy over time.

Think of it this way – if you see a setup but it happens to be at the statistically least volatile time of day, then it reduces your chances – mathematically – of realizing that setup’s most profitable outcome (assuming you know what the statistical profitability of your setup actually is). This fact, in turn, increases your overall risk for that setup for that time of day. How you quantify or model that risk is up to you, naturally.

So here we go…

Methodology: I used 5-minute bars from the US Equities RTH session for both the the TICK and the ES from October 19, 2011 to October 20, 2010. This yielded 260 sessions for the ES and 254 sessions for the TICK. The difference in the number of sessions is because the CME is open and trades on a few days per year that the NYSE is closed.

Next, I averaged the range for each bar in each session over the entire sample period. This yielded 78 data points, 5 minutes apart, spanning the entire session from 6:30AM Pacific Time to 1:00PM Pacific Time.

I then created the mean and standard deviation for each instrument’s data set and plotted each data point on a graph so we could compare them to each other on the same absolute value scale. That scale is volatility (range) expressed in standard deviations for the data set.

Here’s how it turned out, in absolute value terms (best for seeing correlations):

…and again in real terms (best for seeing real volatility):

Some Observations

After perusing the graphs above, a few things should jump out at you.

• First, in the lower graph, negative values are below the mean for each data set, positive values are above the mean. In the upper graph, all values appear to be above the mean.
• You should see right away that the period from 7AM to 7:05AM is by far the most volatile time for the ES, followed by the first bar, 6:30AM to 6:35AM.
• Next, the first hour of the day is the most volatile for the ES, while the last hour of the day is the most volatile for the TICK.
• Their movements are more extreme at the bookends of the day, and extremely correlated in the middle of the day. Also, again notice how volatile the TICK becomes at the end of the day. Much more so than the ES, even though the relative volatility of the two is pretty proportional even if less extreme.
• By a long shot, the least volatile time of day positive values is from 8:00AM to about 8:20AM, and again after 11:00AM. After the 9:00AM hour correlations are very high, but volatility is very low. Perhaps it’s best, statistically speaking, to not initiate any trades during those times.
• Notice that there are regular, daily recurrences of volatility.
• When the TICK is at its most volatile, that means those are the times buy and sell programs are most active.
• And finally, closely examine the time patterns. Do you notice how often volatility spikes within 5 minutes on either side of the top or bottom of the hour?

I am sure there other gems in there just waiting to be mined, but this should get you started and maybe even get you thinking about how will you integrate this kind of information into your trading.

However you do it make sure you trade ’em well, amigos.

[hr]

For those who want to play around with the data even more, here it is in raw form as a .csv file.