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

Re: Re[2]: Rocket Science



PureBytes Links

Trading Reference Links

Mark,

I certainly agree that Fuzzy Logic does better controlling known, physical
inputs.  Where I think FL has a place in trading is combining signals for a
fuzzy composite buy/sell decision.

Example:  Assume you have a system that trades T-Bonds based on 1)-seasons,
intermarket correlation with the 2)- Eurodollar contract and separately with
3)- Copper futures, 4)- price momentum and with 5)- on balance volume.  Such
a system could have about a dozen diverse signals that need to be weighted
together to produce a buy/sell decision.  The simple (dare I say crude) way
would be to add up all the signals and divide by 12.  Fuzzy Logic handles
this situation in a more rational manner.

My point is there are possibilities for FL in trading systems.  It is not
the Holly Grail but only a tool that treats real world problems not in terms
of black or white but rather in terms of shades of gray.

As for neural networks, I believe there is a place of them in trading, but
this is a much more difficult task.  Running a NN with 150 weights reminds
me of optimizing the inputs to a trading system and ending up with a highly
curve fit system that is worthless to trade.  However I have seen NN results
that truly appear to predict future price action.  I've also seen them tank.

Marlowe

----- Original Message -----
From: Mark Brown <markbrown@xxxxxxxxxxxxx>
To: Marlowe Cassetti <marlowec@xxxxxxx>
Cc: <IndTrader2@xxxxxxx>; <omega-list@xxxxxxxxxx>
Sent: Thursday, February 03, 2000 1:36 PM
Subject: Re[2]: Rocket Science


> Hello  Marlowe,
>
> MC> Neural networks have been successfully used for speech recognition,
cancer
> MC> cell detection, and other pattern recognition applications.  Again
there is
> MC> a lot of literature out there on the topic.  Probably the "killer app"
for
> MC> NN is in intelligent robotics.
>
> there  is  a  difference in using FL with a set of knowns and using FL
> with  unknowns.  the  markets are full of unknowns. in futures trading
> unlike stocks "even zero" is not the lowest a future can go. so i have
> not  found  any  application  for a NN or FL and I have tried and will
> most  likely  continue to try in the future. but fl and nn's remind me
> of  the  state  fair  -  i never remember why i haven't been in so long
> until i go then ui remember why i haven't been in such a long time.
>
> MC> I suspect that in another generation, FL and NN will be routinely
accepted
> MC> by the general populace.
>
> it  has  its  place  in a fixed world, but i'm skeptical of non absolute
> applications.
>
>
> --
> Best regards,
>   Mark Brown                        mailto:markbrown@xxxxxxxxxxxxx
>
>