|
PureBytes Links
Trading Reference Links
|
On Jan 31, 2:42pm, kathy t wrote:
> Subject: preliminary results of my system
> i have recently started backtesting a system ive been working on. it is
> strictly an EOD system. last night i backtested every stock in the sp500
> from 1995-1998. my total trades totalled 295 (round turns). i had 159
> winning trades and 136 losing trades.
That's about 300 trades, or 100 trades per year over the period tested.
Given the large universe of stocks (500), that seems like a farily
infrequent level of trading. Here's a few things to check for and
to watch out for:
1) Does the system tend take its trades mostly at the same time across
the universe of stocks? For example, with 100 trades in a year, you
would want to see the system take a position in 100 stocks out of
the S&P 500 all at once, and then sit pat for the rest of the year.
You wan't have the capital to trade the system, and a correlated set
of trades will defeatt the purpose of drawdowns.
2) Is the system long only? 1995-1998 was very much an up-trend.
3) Watch out for the effect of stock splits on your indicators, and
your results. Given the strong up trend during 1995-98 there were
undoubtedly stock splits. If you're looking at returns in straight
dollar terms, your results will be weighted in favor of more recent
times, and earlier trades will be underweighted.
4) Try Random selection as a base line. Look at the distribution
of trades (length of time held), and look at the frequency of
choice, and run a few runs using a randomized distribution.
Make that your baseline system to compare against others.
5) Merge the results from all 500 stocks into one equity curve.
Calculate the Sharpe Ratio for that equity curve, and compare
it to the Sharpe ratio for the being long the S&P over the same
period.
> currently the only money
> management system is 10% profit or a 10% loss. which ever comes first.
> no trailing stop. the day after the signal i enter the trade if it hits
> my predetermined price and when the price bar hits my stop(either at a
> loss or profit) i exit the next bar at the open so it goes without
> saying that i dont always make or loose 10% but over a large period of
> time i dont think it would matter much.
I'd look to bias the ratio of win/loss in favor of wins. So, for example,
I'd look at raising the target to 15% or 20% and lowering the
worst loss to 5%. The %win/loss will be worse, but hopefully the
ratio of ave. win to loss will be 2:1 or so ... which is trdeable.
> My question is this? How does
> this compare to others out there like yourselves who are writting and
> using systems? what percent profitablity should i be looking for before
> take the plunge? i imagine that with a better money managment system my
> profitability could be much better.
Trailing stops, like the "chandalier exit" popularized by Chuck LeBeau
might be better, because they will tend to let profits run.
--
| Gary Funck, Intrepid Technology, gary@xxxxxxxxxxxx, (650) 964-8135
|