|
PureBytes Links
Trading Reference Links
|
Logan467@xxxxxxx wrote:
> On another subject, when you test a system, what criteria do you use? Is it
> the percentage of winning trades, Ratio of average win/ average loss, profit
> factor, return on account or a combination of these and others?
I suppose it matters that you are trading, but you're going to want to consider
strongly the return, since this is what you're really after anyway. After this,
average trade and drawdown are important. Most of the other stuff is informative
- for instance, the average number of bars is helpful to know. Some, like the
percentage of and consecutive winners and losers, is useful if you don't like to
sit through a bunch of losing trades. Overall, I'd say that most of the criteria
is useful to some degree, although the first 3 would seem most important.
> What should
> the numbers be for these criterias for the system to be considered a "good"
> system?
How about - makes a lot of money <g>
> And how long should one back test the system for the system to be
> consider "robust"?
I wonder if it really can anyway. Sure, it's a lot better to go back a long way,
but you're going to run into problems when you do. For instance, I used to trade
on 15 min. bars not that many months ago, then 10, then 5, now 2. If I go back
any more than a few months, I know that what I end up with will be way too loose
for today's markets, causing me to enter and exit too late. What worked great
months or years ago may not today, and this is bound to skew the results in a way
you wouldn't prefer. Here's what I do. I test the stuff on a fairly short
period, but compile the data from applying it to hundreds of issues, which makes
the results a lot more statistically significant while staying within current
conditions. Obviously, this should be kept to issues which you trade, which would
limit a lot of people, particularly futures traders.
> Since I
> don't know how to create a continuous futures contract onTS 4.0, I use the
> indexes
> as a proxy for continuous futures contract by changing the "Big Point Value"
> in the symbol universe to the corresponding futures' value.. Then I would change
> the futures contract from
> the nasdaq to the nyse's. If it's still profitable, I would consider that
> "good" as well. My reasoning is that the NYSE and the Nasdaq have subtle
> differences that would show up if the system is not "good", even though the do
> trade in the same direction most, but not all of the time.
You might better working with the SPX and SPY - which are close to the S+P
futures, whereas these other two aren't. I see your point, though, although what
you would presumably end up with here is something that works well with both the
NYSE Comp and the NDX, but possibly less so with S+P futures.
Regards and Good Trading,
A.J.
|