|
PureBytes Links
Trading Reference Links
|
Dennis, Phil and all other that seem to be at loss with this matter
every now and then:
If you want to compare your trading results over different time frames
and different markets, you have to work with percentages. Once you've
done your homework on Ratio Adjusted Data and percentage based
calculations, you can transform those number to dollar terms that will
tell you how much you can expect to win/lose in the future. Not how much
you could have made in the past, which is the only thing TradeStation
will tell you.
Unfortunately, when it comes to system testing and system testing
parameters, all retail programs (including TS) are completely worthless,
because they will tell you nothing about how your system is likely to
perform in the future. For that you will need a whole new set of
percentage based calculations and regression analysis techniques that
are only available in Excel and other spreadsheet programs. Please read
my article in (p. 46) in the latest issue of Futures magazine (Jan.
1999). I've addressed this matter for a long time now, and I'm just
amazed that it needs to be addressed in the first place, and, second,
that people seem to have such a hard time grasping it.
Sincerely,
Thomas Stridsman
Futures magazine
Dennis wrote:
Phil wrote:
> Lately I've been looking at some long-term things with the SP. Trouble
is
> back in the early 80's the market moved maybe 1 or 2 points on a given
day,
> whereas today it can move 20 - 30 points with ease.
<snip>
> I was thinking of calculating a "fake" margin number that would be
> proportional to today's figure scaled down by the ratio of the
volatilities
> then versus now.
This is very important to compensate for in backtesting. Many people
blow it when they do long-term testing of the SP by just trading a
constant 1 contract. What I do is set margin, slippage and commission to
zero. I also set the big point value to 1 so results are in SP points,
not dollars. Then I tell the system to trade more contracts when
volatility is lower. I use average true range but any volatility
measurement will work. For daily data....
cts = factor / xaverage(truerange,100);
if whatever then buy cts contracts;
if whatever then sell cts contracts;
Set factor to some number that makes sense. A factor of approx. 200
would have you trading 10 contracts now and more in the past. Then just
divide your system results by 10 to get the numbers for 1 contract
adjusted to the current level of volatility. Manually add/subtract
whatever seems appropriate for slippage and commission on a per-trade
basis.
Cumbersome but it works and it gives you a *much* more realistic number
for that all-important max drawdown.
--
Dennis
|