Charlie You will generally see volume migrate a couple of days before rollover. Beaware of the affects of roll over normally from the 16th-17th onwards most months. Esspecially if the spread is relatively large.
What is going to be messy is the rule-of-thumb that I'm proposing counts back from the 25th of each month unless the 25th is not a business day which next month, the 25th will not be a business day. cT
Well it looks like I missed it by a day this time. Yesterday, Dec 17th the Feb contract did more volume so I'll have to call it the rollover day however it was close all day and around the reports it looked like the Jan had more volume. Today, Dec 18, the Feb was clearly the contract of choice. I'm not ready to pull the plug on the rule-of-thumb since this time we a holiday in the mix and perhaps day-of-week has something to do with it as was suggested earlier in the thread. Perhaps we don't rollover on Fridays . So for next month: Jan 19, 2010. cT
When it comes to ES, it makes hardly any difference whether you make the switch early or not since the front month and the back month contracts trade essentially on par. But I have noticed in the past that this is not the case for CL. The front month would go ballistic as it approaches expiration. I s'pose this is due to contango and backwardation as well as everything in between. Has anyone actually done a research on the price correlation between the two CL contracts (ie. do they move in synch, etc)?
I did a very crude analysis on something related with daily data from the EIA website since 2000. I didn't look at the details on when they roll or a lot of other stuff because I am not trying to trade this, just control for calendar effects in the data while I look at something else. But here's what I found FWIW: 1) regressing daily spread changes (second - front contract prices) on monthly dummies, nothing is statistically significantly different from 0 and point estimates are pretty much +/- 1 penny. 2) looking at daily spread changes by day of the month, the only days that average outside +/- 5 cents are the 14, 16, 18, 20-23 and 31. They are pretty noisy. 3) regressing today's spread change on yesterday's, the spread changes are strongly mean reverting at the daily level across all days, but when you look one day at a time there are alot of days of the month that tend to move in the same direction as the last day. 4) regressing today's back contract price change on the today's front month price change, overall the r-squared is 88%. Most days it is in the 90's, it only drops much below that around the 20-23, with the lowest being 56% on the 22. All of this is contaminated by not knowing the roll date EIA uses, it is not a continous contract. And if I were serious about it I would probably measure the days in days from expiration rather than day of the month. I'd post an excel file but the website won't let me tonight.