US Election

[Anonymous]

I am using the R1000 for now, but I am working on one for the S&P500 (or smaller universe) to reduce the selection bias

[ScottMcNab]

SP500…doesn’t seem to be any shortage of potential trades…using 25 positions at the moment but will prob go to 40 positions at 5% in the new year

[TrentRothall]

This might be a topic for a new thread but regarding selection bias..

In a MR system is selection bias as important? the scope for potential gain/losses in drastically reduced when only holding a day, as opposed to trend following where missing 1 or 2 trades could ruin a year. As long as a good number of MCS is done and the spread of results is within your risk tolerance does it matter if the system can only take 50 or 75% of trades?

Thoughts?

[SaidBitar]

the smaller the universe the less selection bias you will have, I am using S&P500 and S&P1500 for two different systems.
For the S&P500 i am catching 67.52% of the possible trades and for the S&P1500 i am catching 55%.

So maybe my backtesting will be looking better than the real results due to the difference between total number of possible trades vs total number of actual trades but i will be happy even with the worst MCs run. I tried to reduce the universe and increase the number of positions this will help in increasing the percentage of actual trades vs possible trades but somehow it did not feel correct for me. The reason is that the position size is becoming small and this is making the DD lower but the bad thing that you are leaving the money in cash. this is why i am OK with having number of positions closer to the average positions and having some deviations from the average results of MCs.

[Nick Radge]

Scott – I moved your comments to its own topic HERE because I think its a very important issue to discuss.

Nick

[Author]

Posts