[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: More on perpetual data



PureBytes Links

Trading Reference Links

Mark Brown wrote:
> 
> > Exactly my point. :-) If you use perpetual data as data 1 in a TS system,
> the TS system report will give you my numbers. It won't include any roll
> expenses because there are none in the data.
> >    Dennis
> 
> OK, I understand that now. it was just your demonstration might have been
> viewed my some that I'm making 20 more points (or whatever) than you are.
> when in fact I'm not, in actual trading I have the same cost and roll over
> that you do and our data is almost identical even though mine is a perp data
> series and yours is the actual price data.  you have to admit its would seem
> its better than many ways of building a continuous price data series for
> testing?
> 
> mb


I'm NO expert but it's my understanding that perp data is built by
averaging several contract months, "gradually" shifting the weights from
the nearby to the next one.  Wouldn't this cause trouble in your P/L
estimates in markets where the contracts arent' as well correlated; for
ex. the grains, meats?  The perp P/L will be some average profit, not
the profit you would get in the new crop or old crop contract.  Same in
meats, one contract can be going down sharply, a deferred contract much
less so.  I think here you must get your P/L numbers from data which
keeps the day to day movement intact for the contract you could be
trading, i.e, individual or gap-adjusted-continuous.

As far as signals go, the same problem with different shape months could
affect any continuous contract trades.  If May is sharply down but Jul
isn't, you may have a signal in May but not July.  Exactly how you
handle that in your testing vs. real trading might cause your results to
vary.

Conrad Bowers