|
PureBytes Links
Trading Reference Links
|
I've been developing an End-Of-Day version of Oddball (because EOD
is all I am able to trade). I can test it only up to October 2001
because that's my most recent data.
Does anyone have EOD data for Advancing, Declining, and Unchanged
issues on the NYSE and NASDAQ exchanges, from October 1999 to
the present? And rollover-adjusted mini S&P and NASDAQ futures
contracts?
It took a while to figure out what to do with the basic Oddball
concept. Mark didn't design it for EOD data. I found I could
get better performance by dumping the RateOfChange function from
Oddball and instead use two Momentum measurements, the first being
the Momentum of (adv-decl)/(adv+decl+unch) and the second being the
Momentum of the first Momentum. Momentum approximates slope, which
is change per unit TIME, rather than change compared to the previous
bar (what you get with the RateOfChange function). By using two
momentums, I get a rough measure of velocity and acceleration of my
adv-decl oscillator.
It worked noticeably better with two momentums than with just one.
I could also get better performance using (adv-decl)/(adv+decl+unch)
rather than just Advancing Issues, even if I optimized either one.
I also improved performance somewhat by using an adaptive money
management stop as well as an adaptive trailing stop, plus the
strategy exits any position if it is ever unprofitable after 1 day.
I'm particularly proud of my adaptive trailing stop. It's my own
invention, on which I plan to write a TASC article.
The strategy works quite well and smoothly over October 1999 to
October 2001 (2 years). Prior to that the performance is sloppy
and choppy. I don't know why this would be. Using Advance-Decline
issues for signals implies that one expects a correlation, or
cause+effect relationship between the data that generates the
signals and the subsequent price data. It just doesn't exist prior
to mid-1999.
The system has an expectation of $0.99 profit per $1 risked. By
comparison, Vladimir's Swinger2-EOD system has an expectation of
about $0.40/$1. In the book "Trade your way to financial freedom"
Van K Tharp advises that anything with an expectation of $0.50/$1 or
so (depending on the number of opportunities to trade) is a decent
system. The "worth" of a system is then determined by a score,
calculated as expectation*opportunities, where opportunities is
number of trades per year.
In my opinion, this is the only way to compare system performance.
Comparing net profit and other results just don't do the job,
because you don't get a result that accounts for risk.
Optimizing for this score is a pain. Tradestation can optimize
only for canned results, not arbitrary user-generated results like
expectation score. To do it, you have to set View->ChartOptions
Strategy tab to save 20,000 or so optimization results, then
optimize so that the total results doesn't exceed 20,000, import
everything into Excel, calculate the scores, sort the list, and find
the parameters that give you the highest score.
I'd like to see how it performs for 10/2001 to the present. Anybody
have the data for E-Mini S&P and E-Mini NASDAQ rollover futures,
advancing, declining, and unchanging issues?
Performance Summary: All Trades
(commission = $20/RT, slippage = $75)
October 1999 to October 2001
Total Net Profit $65,587.50 Open position P/L $512.50
Gross Profit $99,532.50 Gross Loss ($33,945.00)
Total # of trades 115 Percent profitable 48.70%
Number winning trades 56 Number losing trades 59
Largest winning trade $9,375.00 Largest losing trade ($1,615.00)
Average winning trade $1,777.37 Average losing trade ($575.34)
Ratio avg win/avg loss 3.09 Avg trade (win & loss) $570.33
Max consec. Winners 7 Max consec. losers 8
Avg # bars in winners 6 Avg # bars in losers 1
Max intraday drawdown ($4,177.50) Profit Factor 2.93
Max # contracts held 1
Account size required $8,177.50 Return on account 802.05%
--
,|___ Alex Matulich -- alex@xxxxxxxxxxxxxx
// +__> Director of Research and Development
// \
// __) Unicorn Research Corporation -- http://unicorn.us.com
|