MCS trade skipping settings

[TrentRothall]

When doing MCS what should we be aiming for with the trade skipping settings?

As I understand it one of the reasons for doing MCS is to try and test the extra signals that otherwise don’t get picked up in a single run because of selection bias.

If for example I am testing my mean reversion system on a single run from 2000-2016 it produces around 4250 trades but if I use my MC Code with 10% skip it has around 3900 trades, so you are skipping about 8% of trades

As Nick has said somewhere before MR systems tend to have quieter periods followed by periods where there might be large numbers of possible positions, but using 10% skip you are skipping 1 in every 10 trades even in the quiet periods where there are no selection bias issues, but if we drop the skip back to 5-7% we might not take enough of the trades when selection bias is an issue.

Are we understating performance if we skip trades when there is no selection bias issues? this might not be a bad thing.

I think that makes sense…. :S

[Nick Radge]

People love trying to boil the ocean (what about using this index instead of that? Should I invest my cash in here? What about this? etc etc) but the specifics don’t matter that much.

Remember that backtesting is NOT an exact science. We cannot replicate every possible market environment or portfolio outcome. What we’re attempting to do is get some ‘idea’ or ‘feel’ of what the strategy can and can’t do.

It is more important to reasonably certain and trade, rather then attempt to be absolutely certain and never make a trade.

[TrentRothall]

Yeah ok thanks Nick, i thought that i might have been over thinking things a bit. I am happy with the system so i will just trade it. Thanks

[Author]

Posts