I'm using MC with IQfeed. I have had really bad problems with back-adjusted data for CL (QCL#C). I devenloped a daily system for CL and out-of sample results were significant. After I started to paper trade it I found out that the back adjusment does some crazy stuff for systems signals.
System gave me some signals in first month and the result were poor. After rollover and adjustment the system changed those historical signals and somehow those signals were much better. After couple of days i reloaded the data and again those signals were changed and back-testing result were totally different from the previous ones. The system works well in not back adjusted data but of course rollover caps are not included in it.
Does anyone else have similar problems and is there any solutions for this? Also do you prefer using back or nor back-adjusted data when trading commodities?
With a backadjusted contract, the signals should not change (see exception below) and remain the same after rollover. With a continuous futures contract all pastsignals will be typically false.
Exception: Your trading system should only be based on relative prices (moving averages, oscillators, etc.) but not take into account absolute prices (such as round numbers). If it is based on absolute price, past signals will be false for both continuous and backadjusted data.
The following user says Thank You to Fat Tails for this post:
Hopefully I get this right. I'm sure that my system is only using relative prices even thought rules are pretty complex. Also system is working similar as it was couple months ago with non back-adjusted data.
Now when I'm saying non back-adjusted data i'm referring IQfeed's continuous future contract QCL#. The back-adjusted continuous contract is QCL#C and that is the problem. I have found that IQfeed is doing the rollover adjustment as I though by adjusting previous month to the newest and in 2008 prices for this ticker were about 190 at the highest point ( i found out some of your older posts where you had found similar prices for merge back-adjusted data.) The problem should not be that either if I understand have understood something right.
Can you correct if I'm wrong:
1) continous non back-adjusted contract is just contract which doesn't adjust anything. Caps are where they should be and the price is always the price that the current contract was traded.
2)This is the thought one. Continuous(perpetual) contract is something that i don't really get even thought you explained it briefly in the earlier discussion. Hopefully this is not the same as 1).
3) back-adjusted continuous contract adjusts previous contracts values to the new one so the cap is eliminated.
This is really new topic for me and I haven't find similar problems with my earlier systems for instruments like NQ and ES. I think this is common for some commodities due to the physical delivering(bigger rollover caps).
Correct. The backadjusted contract should show prices as high as 190 for CL in 2008. This is due to the Cushing contango, which made it much more expensive for long-only funds to roll WTI futures compared to Brent futures.
- non-backadjusted: single month contracts with rollover gap on rollover day
- backajdusted: single month contracts with rollover gap removed, prior contracts aligned to current front month contract
- perpetual (continuous) contracts: typically prices calculated as a weighted average of two consecutive contracts, with a continuous adjustment of the weighting factor
For more information on perpetuals, read paper per link below: