|
PureBytes Links
Trading Reference Links
|
At 10:25 AM -0500 7/11/00, Mark Brown wrote:
>I'm not - so convinced that it can be consistently replicated because
>as you state below the are other factors that are not completely left
>up to the system. A slight error in any of those and not only will
>you loose money but time watching it fail.
>
>So we all know someone who won the lottery by now. So lottery's work?
>
>They "lotteries" like system testing "in particular this system" are
>easily back testable to see what numbers would have worked. Throw out
>the Jap Yen - or others like SF out and give her a whirl BOB.
The tests were profitable on all 15 commodities I tested. It was best
on JY as you noted, but did OK on all 15 with a reasonable range of
parameters.
But without the positions sizing and portfolio allocation, it would
have not done anywhere near as well.
That said, the performance of the optimized portfolio with the best
bet-sizing algorithm I could devise was not all that great:
Sharpe Ratio
Japanese Yen only 0.94
3 Commodity (Optimal weighting) 1.76
15 Commodity (Optimal weighting) 2.20
15 Commodity (Equal Weighting) 1.36
>From the data listed in Mark Johnson's post, I calculated that the
Sharpe Ratio his portfolio was 0.71 over the three-year period. Of
course, his results were real trades with real commissions, real
slippage, etc., whereas those in my table above were hypothetical - a
big difference.
By comparison, the Sharpe Ratio of buy/hold on an S&P 500 index fund
has been about 1.3 for the five years ending 12/31/99 (but we all
know the past five years has been unusually good ones). Of course,
you would have had to have used a leverage ratio of about 3.7
(borrowing 2.7 times your equity) on your index fund investment to
realize the same 71.6% return.
Many decent "surgical precision" type systems trading a single
commodity can have a Sharpe ratio of over 3 so none of the above
would be considered really great by comparison.
Bob Fulks
|