|
PureBytes Links
Trading Reference Links
|
The first mistake made in system design or trading in general is to
underestimate the effect costs, commission and slippage, have on returns.
A completely random system would break even without costs; therefore, they
are the 'house edge' the trader must overcome to generate profits. The
higher your overall costs, the more 'predictive' your method must be. The
easiest way to reduce costs as a whole and most importantly as a percentage
of the average winning trade and average losing trade is to trade LONGER
TERM. Since costs are fixed regardless of trade length, this will result
in larger winners and losers per contract or share reducing the negative
effect costs have on trading results.
Example
System A 50% winners average winner to average loser ratio of 2/1 before
costs. Costs are $50 round turn. Average winner is $200 before costs or
$150 after a 25% reduction. Average loser is $100 before costs or $150
after a 50% increase. The W/L ratio goes from a healthy 2/1 to even and the
system goes from a large winner (having a high profit expectancy) to break
even (having a profit expectancy of 0).
System B has the same rules as system A but uses a weekly chart rather than
an hourly one. System B has 50% winners and a before cost ratio of average
winners divided by average losers of 2/1. Costs stay at $50. The time
frame has been expanded 35 times, seven hours a day, five days a week, so
the average winner is now $7000 and the average loser $3500 before costs.
$6950 to $3550 after for a ratio of 1.96 to 1.
>1) Do Mechanical Systems work for a small account ie.$12000?
I do not think so, because proper money management would call for risks of
$600 at the MAX per trade. I would recommend only $240. Small initial
risks force one into shorter term trading. Notice the impact the $50 has
on the small initial risk. We are paying .4% in costs every trade. Most
traders try to trade with WAY too little money; therefore, they are
overwhelmed with costs and driven to short term trading. The
predictiveness or edge of a short term method must be much greater than a
long term method to overcome costs. You can see why the vast majority can
not make any money trading. They blame it on 'people in the know' or not
being able to 'predict' the markets and spend all their time trying to
forecast better. Simple math will show them where there problems lie.
Usually capitalization.
>
>2) Does anyone here trade pure Mechanical, or do most mix Mechanical
> with other techniques? It seems alot of traders use elliot/fib which
> are excellent techniques but impossible to mechanize, am i wasting
> my time/ missing half of the profit if i try to go pure mechanical?
Pure mechanical is used by over 75% of the top twenty CTAs measured by
compound return over the past 10 years. Only one CTA, although a good one,
followed by Barclay's mentions Elliott Wave in its description. They say
they use EW for entries and trend following for trade management and exits.
(I by the way have studied, followed, and used EW for five years. I am
not someone who has not looked at other methodologies. I originally
thought mechanical systems were not effective. When faced with
indesputible mounting evidence, I changed my position. I am still
experimenting with mechanized EW.) I personally have not found any example
of sustained, 10 or more years, success using EW.
>3) How much history do i need to test properly, 25? 50? 100 years?
I would want 1000 trades across many different market 'environments' and
markets. A mechanical system should not just work on some markets. The
prevailling thought that different markets (or stocks) have different
'personalities' is ridiculous in my opinion. In fact that argues agains
technical trading as a whole. Certainly, different systems would have
worked better on different markets in the past, but that is the very
definition of curve fitting. What you need is what will work in the future.
>4) There was recently some talk of trade to rule ratio, some of my
> systems are rather complex, does this mean they'll have less of a
> chance of working?
Yes. In fact probably no chance of working.
>5) I've developed a system that works very well with one of the DOW30
> stocks with consistent equity growth over 15 years including the
> recent downturn. The system does not work on the other stocks.
> Is it reasonable to have different systems for different stocks
> or will this system eventually fail? Should a good system work
> across all markets?
It should work on all mass traded large stocks. A good system will work
over all markets if given enough opportunities. Look at the system on the
Dow 30 AND the other stocks together and base your future returns on that.
>6) this is kind of a weird question, if a system could be made
> consistently profitable by lowering it's commission costs(stocks),
> could a similar result be attained by using the system to trade
> ITM options?
A system CAN be made profitable by lowering costs. Do not forget slippage.
>7) Should i test my exits before my entries?
Test them together. Do not forget that the exit is more important than the
entry. You also should spend as much time on money management parameters
as you spent on entry and exit together. I also strongly recommend you do
your final test by hand.
sb
|