|
PureBytes Links
Trading Reference Links
|
At 10:39 AM -0600 12/21/01, Ernie Bonugli wrote:
>Hello Bob,
>
>Friday, December 21, 2001, 10:44:32 AM, you wrote:
>
>BF> Most fixed point-values of stops, etc., can be adapted by relating
>BF> them to the price or the volatility with measures such as
>BF> AverageTrueRange.
>
>Bob, what is adaptive coding?
>
>1) buy at (a price + average true range) stop;
Yes. Maybe: Buy at price + factor * AverageTrueRange stop;
>2) when occilator is at extreme do things like exit position, tighten
>the stops, or do not trade any more, or reverse the position. (ex. ADX
>at extreme point)
ADX tends to be self normalizing so it's values are stable. Buying
when ADX is low (periods of congestion) might take advantage of the
tendency to get in on breakouts but this is a technique. Setting the
trigger points based upon volatility or some other measure of the
market is adapting the technique to the changing market.
>3) shares to trade next = ShareCnt/(TR*BigPointValue)
Yes. This tends to adjust risk to volatility.
>4) exitlong at breakeven(average true range*.8) stop;
Yes. This tends to create a band proportional to volatility.
>5) is mkt trending relax the exit criteria.
Yes. If you define trending is some way related to the numbers such as
RateOfChange / AverageTrueRange.
>6) in general, use mkt strength oscillators to guide the entry and exit values
This is a technique. Adaptive means making the measurement of
"strength" properly in different markets.
>Am I missing the idea? Would you site some relevant examples?
These are good examples. Any time you have fixed numbers that are in
the dimensions of price or dollars, you may have problems. Good
measures have dimensions of percent, or ratio of prices. So instead
of 4.7 points you might use 5.3% of price so that it varies with the
price. Or 2.5 times AverageTrueRange so that it varies with
volatility.
Systems trading stocks tend to do better using percents of price
since these tend to go for longer-term appreciation. Systems trading
futures tend to do better using Range, or True-Range since you mostly
care about small changes using the high leverage. Range and
True-Range are also not affected by back-adjusting contracts (using
fixed offsets) so do not get distorted by back-adjusting (whereas
price does get distorted).
>BF> A "fully developed" autopilot includes code to handle hundreds of
>BF> special cases that occur only rarely. A fully developed trading
>BF> system that can adequately handle 99% of the cases with adaptation to
>BF> changing market characteristics, intelligent bet-sizing, risk
>BF> management, crisis stops, etc., might be several pages of code.
>
> So is the a successful system? Or is this curve fitting? I ask
> because my style of developing a system is as you have described,
> well maybe. Typically, 10% lines to setup oscillators and stuff, 10%
> lines to enter market, 80% lines to exit. Or, simple "adaptive" entry, general
> "adaptive" exits, special case worst case scenarios, etc.. These kinda of
> air-o-plane certainly does not fit the idea of "the simpler code the
> better the system". Please comment.
Is making an autopilot fly around storms, dodge other planes, react
safely to clear-air-turbulence examples of curve fitting? I would not
think so.
>How do I know that I have not curve fitted? That is the question!
> Is the only answer, time will tell?
And it depends upon how you measure the performance. If you simply
measure the duration of the flight or fuel consumption, you might
easily curve-fit good numbers and have a lot of very green-looking
passengers.
But if you measure the standard deviation of passenger comfort from a
nice steady average value, (as the Sharpe Ratio does for trading
systems) you would find that it is very hard to curve-fit to a good
number. You would also find that all of those additions to cover the
special cases DECREASE the standard deviation, (and INCREASE the
Sharpe Ratio) of passenger comfort, making for a less stressful
trip...
Good questions...
Bob Fulks
|