|
PureBytes Links
Trading Reference Links
|
| -----Ursprungliche Nachricht-----
| Von: hans esser [SMTP:he96@xxxxxxxxxxxxxx]
| Gesendet am: Wednesday, September 22, 1999 2:46 PM
| An: omega-list@xxxxxxxxxx
| Betreff: Re: Neurofuzzy, ex-Trailing stops are invalid
|
|
| Dear Sirs,
|
| I hope you dont mind me interfering the battle between the 2 frogs with a
| simple question....
|
We don't mind the intrusion of a Bratwurst...
| Gwen wrote:
| > > | Maybe yes, maybe no.
| > > | If I succeed to build trading systems that are better than 4* DD in
| > > 200 | weeks, | I'm probably not the only one. |
| > >
| > > 4*DD in any 200 trades, not weeks, unless it takes 200 or more weeks to
| > > do 200 trades.
|
| PO replied:
| > Easier in this case...
| > I have never seen, for a system that I have backtested and kept as a valid
| > one, the case where the max DD was above 2 times the max DD during
| > historical training or building period. I usually backtest systems on
| > thousands of bars ( 50,000 at least) and this produces not only 200 trades
| > but rather 2000 or 3000.
|
| Do you talk about INTRADAY data or EOD (or both) - obviously its very easy
| to get a lot of RECENT bars when you would look at 1 or 5 or so min bars.
|
| However going to EOD, i.e. daily charts would require 200 years of data
| <g>......so cant be that...... so you probably use 5min bars (or so), which
| would
| be some 2.5 years of S&P.
|
| Brings me back to my original question:
|
| How valid is a test on a e.g. DAILY S&P chart since inception, i.e. back to
| 1982 in THESE days - most systems I have seen make more (or loose more)
| in last 2-3 years than the entire period before. Is a volatility adjustement,
| i.e.
| trade MORE ctks in old days when swings and ranges were lots smaller the
| cure ? or just a workaround (tm OMGA) ?
|
| Would it, for those reasons of changing markets, it be more recommendable
| to use INTRADAY data as you can have more bars of more recent periods
| than of EOD stuff (yes, I dont believe in any information of DJII chart from
| 1920)
|
| rgds hans
| no web site
| nothing to sell
| might even peek at that sirtrade.com one day
My view is that as long as bars are formed correctly (liquidity is sufficient)
the time frame should not be relevant. So I talk in bars and trades. Since
trading is indeed a matter of statistics, it is irrelvant whether one has 200
trades in 5 days or 200 days in 200 years, except for friction costs that
increase on the short end, and capital requirements that increase on the long
end.
I tested on 30min TBond data from 1980 to 1995. Systems generated 5000 trades
in the period. I did not consider results prior to 1990 to be significant. So I
chose parameter sets that showed consistency throughout the whole period,
independantly of absolute results.
For the SP, I don't believe anything that has been tested simply on pre1996
data. I have a simple first hour breakout system (buy break of first hour high,
reverse on first hour low breakout) which has good performance throughout, but
sometimes really bad drawdowns. I never found a way to reduce it to consistent
levels.
What I look for in any system except the return = 4 * Max DD in 200 trades
criteria, is consistency throughout. No big ups and downs, no surprises, only
steady streams. I never found something satisfactory in that sense, The bond
one I used has an average gross trade of 2 ticks more or less (anywhere from
$36 to $80). Since I traded at $12 costs AND slippage included, I got away with
that for 3 years, but that is not a life... I tell you.
Cheers,
Frogwenn
|