[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Optimization Question



PureBytes Links

Trading Reference Links


Check out Oct 2001 Futures Mag article "Walking a Fine Line".

"Enhanced returns were found among optimization strategies 
that selected the worst performing input for the prior period.
For example, regardless of the criteria used for optimization - Sharpe
ratio, average trade or net profit -  the worst input for the shorter
intervals significantly outperformed the best performing input in the
prior period and the average of all inputs."

I wish I could backtest myself.

sincerely,

On Thu, 20 Dec 2001 13:25:27 -0500
"rascal2" <rascal2@xxxxxxxxx> wrote:

rascal2> I trade the S&P futures contract.  I understand the pitfalls of
rascal2> overoptimization and always optimize any systems I develop on several
rascal2> expired contracts before trying them on out-of-range data.
rascal2> 
rascal2> However, I have a friend who used a totally different approach and with good
rascal2> success.  He optimizes his (stock trading) system every weekend on data from
rascal2> the previous four weeks.  He then uses those settings for the forthcoming
rascal2> week, optimizing again the next weekend. He acknowledges that he is "tuning"
rascal2> his system to a specific data set, but does so knowingly with the
rascal2> explanation that the market over the next week will most likely resemble the
rascal2> market over the last month.  By reoptimizing every weekend with the previous
rascal2> four weeks data, he hopes to keep his system in synch with the market.  He
rascal2> likens this to developing adaptive indicators.
rascal2> 
rascal2> I know that theoretically this should not work, but in fact it seems to work
rascal2> pretty well.  His system consistently turns a better profit with smaller
rascal2> drawdowns that it would when optimized over a longer period and then run
rascal2> against the same data sets.
rascal2> 
rascal2> I'd be interested in any thoughts anyone might have on this approach.
rascal2> 

-- 
Kato, Yoshiaki <yskato@xxxxxxxxxxxxxx>