Re: Backcasting Forcasting & working complete models

In article <[EMAIL PROTECTED]>, Reanimater_2000
@yahoo.com says...
> 
> "Bernardz" <[EMAIL PROTECTED]> wrote in message
> news:[EMAIL PROTECTED]
> > In article <[EMAIL PROTECTED]>, Reanimater_2000
> > @yahoo.com says...
> > >
> > > "Bernardz" <[EMAIL PROTECTED]> wrote in message
> > > news:[EMAIL PROTECTED]
> > > > In article <[EMAIL PROTECTED]>, [EMAIL PROTECTED]
> > > > says...
> > > > > Buncha crap.  The world is dripping with massive financial gain to
> be
> > > > > had by anybody who can accurately predict anything subject to
> > > > > empirical validation, horseraces to stock markets to lotteries
> > > > >
> > > >
> > > > These are example where people are competing and learning, it does
> work
> > > > for say sales of oranges in a store?
> > > >
> > >
> > > Suppose you are the manager of a local grocery store. You want to sell
> more
> > > oranges. You have a record of those times when you sold more and those
> times
> > > when you sold less. You find some of the time you sold more and further
> > > divide that group into situations more similar to yours now and other
> > > unsimilar circumstances. Of those times where you sold more where the
> > > circumstances are similar you have backcasted.
> > >
> > > Ideally you would take that time and and the compared times together but
> > > eliminate the latest outcome and then make you model variable enough to
> have
> > > a diverse and differing set of solutions within a range and find those
> that
> > > solved what was removed. These are identical enough in some situations
> to
> > > see an answer not easily seen.
> > >
> > > Two steps forward and one back to proceed in a circuitous but
> directional
> > > manner.
> >
> > Actually PosFocus technique does all this for you automatically. You
> > feed in all the variables and it goes looking though various models and
> > comes up with the one that fits best. I can verify that focus or
> > Backcasting in this situation does give good results.
> >
> > However I have tried this technique on economic figures like interest
> > rates, inflation rates and share prices and it fails miserably.
> >
> 
> Is this an opportunity for some programmer to make millions if he can make
> backcasting models more efficient at predicting interest rates, inflation
> etc..., or should we scrap the technic?

Well depends on your business model. If you are persuasive enough you 
could convince a pension fund to give you a few hundred million to try 
it.


> 
> This PosFocus did it backcast up to a particular time in the past from a
> further time in the past? I mean in simple terms if I have information about
> three days ago and information about two days ago can it pick up trends that
> lead to this day when already done without having information in the model
> about this day. Could PosFocus then devise trend types and features for
> about any situation and match trends that are similar for trying those
> methodologies that worked in situations that look like this but the outcome,
> though known to the programmer, is withheld from the model?


It can do all of that. Most people simply feed into it this weeks sales 
figures and it calculates an expected sale next week together with a S/D 
of the likely error.




> 
> >
> >
> > -- 
> > To the person satisfied with things as they are now, a scary thought is
> > that tomorrow will be different from today.
> >
> > 15th saying of Bernard
> 
> 
> 

-- 
To the person satisfied with things as they are now, a scary thought is 
that tomorrow will be different from today.

15th saying of Bernard