Z8 leverage and lookback period

I have three comments/questions on the use of leverage on the Z8 system:

1. When using no leverage in the Z8assets.csv file (setting it to 1 instead of 4) the algoritm returns an incorrect annual return value. Why is this? Can this be corrected. As soon as you set the leverage to something like 1.1 for instance, everything works properly. It seems like this is a bug.

2. Users should note that in the standard asset list which is included with the distribution there is no margin cost included. This margin cost is coupled to the interest rate and is at present 1.85% (I believe). For the sake of fairness it seems appropriate to include some margin cost as the actual performance will be reduced significantly by this cost.

3. It would be very helpful if the developers would allow to modify the lookback period for calculating the return. Is there anyway to interfere with this value as a user of the Z8 system.

1. Yes, this is possible. When leverage is set to 1, it switches to a no leverage mode and margin is not accumulated anymore. I don't know the reason of this - maybe this was implemented for a special case of direct market access. I'll look into this, we'll probably remove that special case for the next release.

As to my knowlesge, IB are a leverage broker and have no fixed margin cost.

The Z8 lookback period simply depends on historical data of thee default assets. You can not backtest assets that didn't exist at that time.

Thank you for the answers. For the last question I failed to express myself clearly enough. What I meant with "the lookback period for calculating the return" is not identical to the lookback period you refer to. I was referring the LookBack = DAYS = 252 in the MVOtest.c file, which I assume is identical in the Z8 strategy. To calculate the mean return values and covariance matrix 1 year is used. Tests on certain ETF portfolios with MVOtest.c indicate a very clear dependence on this LookBack and substantial improvement in return without increase in DD or decrease in SR is frequently found if this LookBack is shortened to a few trading months only. This occurs not for singular values of LookBack but for a broad range of values significantly shorter than 1 year, indicating that it may be relevant to shorten in this cases the LookBack variable. Hence if there were a Z8.par where one could set this, the flexibility of using the Z8 strategy would be improved.

Thank you for the answers. For the last question I failed to express myself clearly enough. What I meant with "the lookback period for calculating the return" is not identical to the lookback period you refer to. I was referring the LookBack = DAYS = 252 in the MVOtest.c file, which I assume is identical in the Z8 strategy. To calculate the mean return values and covariance matrix 1 year is used. Tests on certain ETF portfolios with MVOtest.c indicate a very clear dependence on this LookBack and substantial improvement in return without increase in DD or decrease in SR is frequently found if this LookBack is shortened to a few trading months only. This occurs not for singular values of LookBack but for a broad range of values significantly shorter than 1 year, indicating that it may be relevant to shorten in this cases the LookBack variable. Hence if there were a Z8.par where one could set this, the flexibility of using the Z8 strategy would be improved.

No, it is not 252, as to my knowledge it is about 8 months in Z8 since that produced better results with most assets.