Contracts roll over so you don't need to test monthly contracts at a time. 20-30 trades in sample is a fun curve fit but has zero meaning when it comes to trading.
What do you mean by that? The contract is only the most active for a one month period. You can only really evaluate this on the active contract as the volume and data is way to spotty on an out month contract. Unless someone has patched together data of active months you have to do one month at a time. Even if it was patched together at the seams it would be useless as either the contango or backwardation would make those joints useless. I can apply the system to data of any contract. But again if I was to for instance apply it an old contract, the only time the data would really be relevant is the one month or so when the volume is active. AT any time these systems will need to moved from contract to contract on a monthly basis so evaluating them on a monthly basis would seem correct as well. For example if you look at the results above you will see that I did apply it to an out month contract and have data in that sample of eight months. It only did 36 trades in that eight month period as the data is so thin as that is not an active contract. What I need is the data for the active period on a slew of old contracts. The evaluation will still only be relevant when that contract is in its active periodor approx 1-2 months preceeding expiration. Again, if one were using these systems that is how they would be applied, for a one-two month period preceeding expiration.
hi drbob , have you tried this method on other contracts ? Is it specific to this spread only ? I have developed as well, seen the method work then erode over some time. I would love to collaborate with you if you have interest.
I don't mean to sound offensive but you seem to be quite new to all this. Do a search for "continuous contract". That's the only proper way to do this type of backtesting.
Im not new to trading at all. 15+ years. As I said above, continuous data would be erroneous as the seam between contracts would render it useless with the contango and backwardation that exist in crude every month. The system would triiger false signals on the gaps that would exist at the seams? Understand? The way to test this has to be individual contract data , not merged. I need contract data for a number of months. As far back as possible would be great. But not seamed.
Hi It is not a spread. It is on individual crude oil contracts. The stats there specify which. Some are on Brent and some are on WTI
Ninja certainly can do a continuous backadjusted contract - in theory. In practice, every-time I connect to any variation of IQfeed (that is, Kinetick end-of-day, or Kinetick/IQfeed live, even connecting to the server to download MarketReplay data), my rollover database get overwritten with their crap (for CL in particular, it is really full of shit, they know it and couldn't care less). I am now saving the rollover database each time I update it, and I reload once I disconnect from any data server.
You might not be new to trading, but it certainly sounds like you are new to algo trading. The answer is "merge backadjusted".