|
PureBytes Links
Trading Reference Links
|
> In everything I learned (which may not be enough), data of
> "adjusted" continuous contracts, by their very nature, are flawed.
In my experience, continuous contracts are quite valid and are
absolutely essential for my system testing.
> One, why not just use the same-month contract if your trading is
> that long term (a Gann idea)?
I want to test system strategies over *years* of data. I need to be
able to optimize a strategy over a single data stream for that
period. How else would you do it?
Using back-adjusted continuous contracts, I can normally duplicate
the exact same series of trades in a "real" contract or in the
continuous data. The only differences are the entry/exit prices
(which are obviously different in the adjusted contract) and the
INcorrect trades taken at the start of the "real" contract because it
doesn't have enough valid data to "spin up" the system's calculations.
> Two, again for "long-term" positions, why not just "roll over" to
> the next month and adjust the entry price instead of the whole data
> base?
Because the price jump at a roll knocks nearly all indicators haywire
until the price jump slips outside of the "price window" that the
indicator uses. In theory IIR filters like xaverage have an INFINITE
lookback, so in theory you can NEVER lose the bogus results from the
price jump. In reality (especially with TS's limited precision :-)
it does damp out reasonably rapidly, but it still throws everything
off for a while.
Just because you don't use it in your style of trading, Gerald,
doesn't mean it's invalid.
Gary
|