|
PureBytes Links
Trading Reference Links
|
if what you say is true about aberration and the s&p's (i have no
reason to doubt you), then that explains why mark johnson doesn't trade
the damn thing on the stock indices. it appears to blow up when it
encounters excessive volatility. but i'm confused. isn't aberration's
supposedly strongest asset is that it catches the volatile, explosive
moves in a commodity mix? or maybe that's the gotcha, in that you've
got to have a mix or basket of commodities to minimize the drawdowns
from the individual commodities. and it also sounds like that not only
is the system itself is curve fit, but the basket of commodities is
curve fit too.
vegas anyone?
TJ
--- Gaius Marius wrote:
> To keep busy while I was waiting for the markets to open this
> morning, I
> decided to play with Aberration on the S&P. To make the results as
> realistic
> as possible, I used 2 continuous futures contract, 1 back adjusted
> and one
> without any kind of adjustment at all. Data going all the way back to
> March
> of 1992. Before 11/01/97, Big Point Value was $500 per contract.
> After,
> $250. I used $20 as commision and 1 point as slippage.
>
> On a Daily chart, I optimize the length from 2 to 200 with the best
> fit
> found to be 87. It made $31,470 on the back adjusted contract and
> $113,780
> on the non back adjusted (the disparity can be explain in that I
> haven't
> made the adjust for rollovers yet.) It suffered a $53,850 Max DD for
> the
> adjusted and $192,750 for the non adjusted. Profit factor was 1.28
> for the
> adjusted and 1.34 for the non adjusted. Using Bob Fulks' ELA for the
> sharpe
> ratio, aberration has a 0.17 for the adjusted contract and .21 for
> the non
> adjusted.
>
> For intraday trading, I decided to split up the data. My test data
> was from
> March 2, 1992 til March 2, 1996. With my out of sample testing data,
> I used
> March2, 1996 until May 19,1999. Also, this way I can get aroud the
> 13,000
> bars limitation in TS.
>
> First I tried 60 minute data. No matter parameter I tried to optimize
> it
> for, I found that it was unprofitable. Then I tried 45 min bars.
> Same
> thing. Unprofitable from 2 to 200. 90 min bars. It was profitable on
> the out
> of sample test but unprofitable on the test data (I switch because of
> frustration).
>
> The only thing I can conclude is that this system being sold for
> $1800 is a
> curvefitted piece of shit. Trade it at your own risk.
|