Investigating CSI rolling methods

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
Post Reply
Posts: 4
Joined: Tue May 27, 2008 7:56 am

Investigating CSI rolling methods

Post by es175 » Tue Aug 12, 2008 7:53 pm

Greetings to the forum...

I always enjoy posts with data and research results so I thought I'd share some results from today's investigations. (Pretty simple stuff, but it's a start.)

It struck me I've spent a lot of time considering when to buy and sell and rather less on matters such as the nature of the underlying data. With this in mind, I started to explore the effects that different back adjustment methods have on trading performance. The image below shows the results of a shorter term system being tested on a variety of portfolios. The names beside each set of results should be fairly self-evident - C for close, O for Open, 1st for first trigger, 2nd for two consecutive triggers. The last portfolio was created using the fixed calendar rollover dates published on Pinnacle's website. All other factors were of course kept constant.

Hopefully of interest...

back adjusted methodology research results.PNG
back adjusted methodology research results.PNG (39.7 KiB) Viewed 5050 times

Roundtable Knight
Roundtable Knight
Posts: 2986
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo » Tue Aug 12, 2008 10:38 pm

Thank you! Thank you for sharing the results of your research, and congratulations for choosing a study-problem of such immediate relevance & importance.

Also thank you for including Custom Statistics blox that print the average duration of winning trades and average duration of losing trades. These columns of output results help people understand just what kind of system (short-to-intermediate term) is being studied. A child six years of age can plainly see, this is NOT ultra long term trading.

I find it particularly instructive to compare the rows where we "Rollover at the Open" versus rows for the identical circumstances except we "Rollover at the Close". You might think these would give extremely similar results. One way to analyze your data is to conclude, "No, intuition is wrong, these do not give extremely similar results." Another analysis concludes, "they are extremely similar, and now you have a numerical measurement of the test statistics of extremely similar basic systems." In other words, we have numerical measurements of the "noise", the "variability", of test measurements that are basically identical

Thanks again!

Roundtable Knight
Roundtable Knight
Posts: 939
Joined: Fri Jan 30, 2004 3:54 pm
Location: Chicago

Post by RedRock » Wed Aug 13, 2008 1:18 am

fascinating results! It would be interesting to look at fixed roll date (Pinnacle) -1, -2, -3... +1 +2, +3. Seeking footprints of "known" roll date exploitation. (or not)

Roundtable Knight
Roundtable Knight
Posts: 133
Joined: Thu Feb 28, 2008 2:59 am
Location: Stockholm

Interesting topic...

Post by Mats » Wed Aug 13, 2008 8:22 am


I had done some simulations and found that the later roll the better performance and little more drawdown.

However it is not always possible for me to trade with late roll as my broker (IB) requires me to be out First Notice day on some contracts.

What i understand this system here is a short term system. My system is more LTTF and do not seen the big difference in results as posted here.

How late are your brokers to accept rolls and how does your simulations verify my results



Roundtable Knight
Roundtable Knight
Posts: 695
Joined: Mon May 17, 2004 4:18 pm
Location: Boston, MA

Post by LeapFrog » Wed Aug 13, 2008 9:29 am

Wow! What an excellent first post es175.

A lot of fiddly work to run that experiment - thank you very much for sharing.

Taking the Pinnacle roll schedule and the "best of" of your other schedules, I ran a binary test on a very short term system (av. 3 days) and a short term system (av. 40 days). Very little change in results on the very short term system (expected since the chances of a roll happening within a two-three day trade is less than a longer hold period). On the 40 day system I did see improvement in CAGR but an increase in DD to go with it - not the dramatic near doubling that your results achieved. Probably just the different style of systems.

One BIG advantage, for me, of a fixed roll schedule, is being able to plan out my day - knowing when the rolls are coming along - I can work the rolls as it were. On the other hand, using a Vol + OI doube trigger method is more dynamic with the market which appeals also.

Good food for thought and further testing.

Thanks again!

AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2040
Joined: Fri Apr 25, 2003 3:33 pm
Location: London

Re: Interesting topic...

Post by AFJ Garner » Wed Aug 13, 2008 12:23 pm

Mats wrote:Everyone..

I had done some simulations and found that the later roll the better performance /Mats
One thing I noticed was in relation to markets often in backwardation such as energy as the price comes up to meet the spot. Since my greatest fear is delivery risk however.................

Post Reply