Don't if anyone else experiences the same as this ?
I test with a basked of 90/110 futures markets over 25 years - Normally I would test vanialla for smooth spaces and the perhaps add in risk management afterwards (i.e. group risk / correlation limiter)
Seems strange to me that by far the best results are achieved by not using any such risk managers - as soon as I apply a multitude of group risk parameters or correlation parameters that reduce markets traded simultaneously, etc - the performance cannot get anywhere close to the orginal performance - in terms of MAR or Draw down length ?
Seems a ballsy move to trade 100 markets and take all trades - but perhaps the backtesting is telling me something ?
This is applicable, incidentally to about 500,000 paramer runs over 3 or 4 different trend following systems that are fairly vanilla
Chris
Strange Conundrum - Or Not ?
Perhaps by taking all those positions, presumably with a small % risk, the power of diversification overcomes correlation. For although sugar in NY is similar to sugar in London, they are not exactly the same. So while taking a half size position in both, one may be stopped out, but the other holds on to become a winner. A smaller account may not have the ability to trade far and wide at a lower %risk, thus the group limiters may become more relevant with "smaller" accounts with higher relative bet sizes on smaller portfolios.. Thats my theory.