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

Re: The radio trading system: An enjoyable research result



PureBytes Links

Trading Reference Links

Mark, it's always a pleasure to see something appear from you on
this mailing list.

I looked at your Crystal Radio trading system.  Cool!  In simpler
terms, you're measuring the amplitude envelope of a bandpass filter,
and your long/short state is simply the increasing/decreasing of
that amplitude.  It looks like it'll work the same without the gain
parameter; not sure what that's for.

Anyway, this made me think of an idea: Your bandpass filter consists
of two RC circuits in series, one low pass and one high pass.  RLC
circuits might work better here (and actually be more analogous to
a crystal radio).  If one uses an RLC filter, one can improve the
bandpass response by driving it with a current source rather than a
voltage source; i.e. assume that the input prices represent current,
and then use the output voltage from the filter.  Every time you
have a filter, pretend you have a current source that converts
voltage to current.  This should improve the response of all the
filters.

This might be overly complex but it's something to think about.

-Alex

>To: tr-list@xxxxxxxxxxxxxxxxxx
>From: "Mark Johnson" <janitor@xxxxxxxxxxxx>
>Date: Tue, 14 Oct 2003 18:29:49 -0700 (PDT)
>Subject: [TR] An enjoyable research result
>
>Last week I gave a talk on futures trading using mechanical
>systems, to a bunch of Ph.D. candidates studying electrical
>engineering.  To make it interesting and intuitive for them,
>I presented the trading systems as electronic filters which
>drive Light Emitting Diodes ("LEDs"), a green one that comes
>on when the system has a Long position, and a red one that
>comes on when the system has a Short position.  A circuit
>schematic diagram of one such electronic filter may be viewed
>at this link (scroll down to Saturday May 31):
><http://www.turtletradingsoftware.com/forum/viewtopic.php?p=1753&highlight=circuit%2A#1753>
>
>This time around, I was talking to people whose research and
>specialty is radio frequency IC's.  You know, "wireless":
>Bluetooth.  802.11.  That kind of stuff.  So I added a new
>filter to the talk, specifically aimed at their field of
>expertise.  And, to my delight, the new filter-as-trading-system
>gave very nice trading performance.
>
>The new "radio inspired" filter has the same block diagram as
>a Crystal Radio, which you may have built in your youth.  The
>block diagram is:  Input -> Bandpass_Filter -> Rectifier ->
>Lowpass_Filter -> Output.  In a radio the input is the antenna,
>the bandpass filter is the coil-and-capacitor tuned circuit,
>the rectifier is the crystal/diode, the lowpass filter is the
>condenser, and the Output is a pair of headphones.
>
>Well, if you turn that into a trading system where the Input
>is a set of Close prices of a commodity futures contract,
>and the output is a Green LED for long and a Red LED for short,
>you get a nice reward for your efforts.
>
>I'll point you in the general direction by giving you some free
>code written in PERL that implements the radio trading system
>using single contracts.  It's on my website at mjohnson dot com
>within the downloadable source code archive file pgms.zip .
>The radio trading system is in that archive, stored as radio.pl.
>Output results (results obtained testing it with PERL, not TR)
>are also downloadable in the file rslt.zip .
>
>Maybe it'll be useful or amusing to you, I hope so.  Maybe it'll
>inspire you to recode the system in TR and change it around a little,
>either the entry/exit rules or the parameter values or the position
>sizing algorithm.  If you do, my guess is you'll be pleasantly
>surprised.  Please remember that you got my code for free and you
>are entitled to all the software support that you paid for, i.e. none.
>It's unsupported free software.  If you need help coding this
>in TR, or if you need help debugging, or if you need help
>learning to use TR effectively, or if you need help because you
>bought TR but you're not really a programmer, I will be happy to
>refer you to a group of expert professionals.  They can probably
>help with whatever problem you're having, at a very reasonable
>hourly rate of pay.  But I hope you choose to do the work yourself
>and have your own fun.
>
>Best wishes,
>-Mark Johnson