Filters & KISS methods

[olintner]

Quote from nononsense:

May I ask you what kind of systems you test; I mean daytrading ES or so; how many trades average a day?

Hi NoNonsense,

the system in mind for the monte carlo simulation is a very short-term break-out system which trades 150 times a year a market. It is traded with real money over a sample of 20 markets (fixed income, equity indices, fx and energy) since 1998.
Given the large number of trades it's comparatively easy to get meaningful numbers for a resampling of daily returns. The expected return is not effected by the resampling (mean stays the same) but because MDD is path depended you get a good feeling for the sensitivity of your system for missing a few good trades.
Last but not least - for the portfolio MC don't resample on an individual market basis but on the equity curve of the portfolio. Because if you resample on single market basis for the portfolio you expect a non-correlation which just isn't there when times get tough

Good trading, Oliver

 

[TheStudent]

I don't know if you would count risk arbitrage as a system, but there is a paper which shows the returns of merger arb funds over time.

In the early time period, merger returns were often significantly above market with a few outliers that had tremendous returns.

In the later time period, when merger arb had really mushroomed into a significant industry, the grouping of returns was much tighter and clustered around risk free + X. Outliers had pretty much dissapeared on the upside, leaving only downside outliers.

Here's my interpretation :

Tighter clusters => competition causes industry returns to converge

Absence of upside outliers => the additional resources brought about by the growing industry to scan the horizon for opportunities has eliminated golden opportunities

Movement of returns to risk free + X => industry growth means more people sharing the same pie. Everyone has to be happy with a smaller slice.

So ... markets do shift and restructure.