|
PureBytes Links
Trading Reference Links
|
To All RealTraders:
I have apparently misused the term walk-forward testing, when I should have
said "paper trade" or "simulated trading." All my we site is doing is
taking a purely mechanical trading system that I developed and is now "paper
trading" the signals generated by it for all to see. The system is not
being change or optimized in any way. Please note that I have NOT used
optimization in the development of this system. So I guess I am not
"walk-forward" testing this system, I am just hypothetically trading the
signals it generates to see if it makes money and inviting everyone to
watch. The system lives or dies by it's own merits, but I am not trying to
optimize it. Just paper trading to see if it makes money. Sorry for the
misused term, I did not mean to cause such confusion. Now...let's see what
happens with these paper trades.
All the Best !
Bill Shumake
-----Original Message-----
From: steven poser <swp@xxxxxxxxxx>
To: RealTraders Discussion Group <realtraders@xxxxxxxxxxxxxx>
Date: Tuesday, July 07, 1998 5:46 PM
Subject: Re: Walk-forward. What is it.
>Basically, the idea is as follows:
>
>Systems developers often sell you a system based on optimized parameters
>of historical events. Unfortunately, what they typically do is wind up
>curve fitting. Though they might tell you that they developed their
>system based on five years of daily data, they might have so many entry
>and exit conditions as to make any historical test have little
>statistical significance as degrees of freedom wind up being too low.
>
>A way of getting around this and fooling the public is by saying that
>the system is perfectly fine, but its performance is poor because the
>parameters need to be changed. I mean, the market changes all the time,
>right? (This is the sleazy system seller speaking, not me!) So, they
>reoptimize their parameters again. Now RSI is 18 days instead of 11. The
>moving average they use falls to 6 from 9. They trail stops by 1.02%
>instead of 0.77%. In the end though, it is the same s--t. This system is
>still doomed to fail as now they have reoptimized on new past data, and
>it still will not work "walking forward."
>
>Steve Poser
|