** Eliminating Selection Bias **

[Anonymous]

I have a scenario where I have filled up my total number of positions and then have exits, so the next day I can place orders to take trades on the number that have been sold. When I run the backtest it ignores the new trades that could have been taken based on those exits, almost as if the exits have been ignored. Anybody else seen that?
E.g. 20 is max positions and all 20 have been filled. 18 can be exited the next day which means I can place orders for 18 new positions. The backtest seems to ignore that there is space for new trades because there is an exit on open.

[Nick Radge]

Miguel,
We have the solution for this. I will try and post the code in the coming days in the course.

Nick

[ZachSwann]

Nick in the course notes the hyperlink to the MR template selection bias part 3 it says MOC should MR

[Nick Radge]

Thanks. All fixed.

[Anonymous]

Brilliant! Thanks guys.
Time to dig into some old systems and test them out.

[RobGiles]

Guys

Under “Entry and Exit Signals”, my MOC code is as follows:

LESetUp = C1 AND C3 AND C4;

BuyLimP = L – ATRVal;
Tick = 0.01;
BuyLimVal = round(BuyLimP/Tick);
BuyLimit = BuyLimVal * Tick;
//

---

OnLastTwoBarsOfDelistedSecurity = BarIndex() >= (LastValue(BarIndex()) -1) AND !IsNull(GetFnData(“DelistingDate”));
//Used for the entry
LE = Ref(LESetUp,-1) AND NOT OnLastTwoBarsOfDelistedSecurity;//Long Entry the day before trigger bar
LEPrice = Min(Ref(BuyLimit,-1),Open);
//

---

_SECTION_END();

Do we even need:

OnLastTwoBarsOfDelistedSecurity = BarIndex() >= (LastValue(BarIndex()) -1) AND !IsNull(GetFnData(“DelistingDate”));

or this bit of code in the LE section:
AND NOT OnLastTwoBarsOfDelistedSecurity;

[SaidBitar]

No you do not need it since you are exiting the same day and you do not care if it is the last trading day for the stock

[RobGiles]

Said Bitar wrote:

No you do not need it since you are exiting the same day and you do not care if it is the last trading day for the stock

Thanks Said
Would’ve having that code in my original non CBT code affected my backtest results, as I was never told to remove it when I submitted my code for sign off?

[Stephen James]

It stops you entering a delisted stock Rob, so optional on a system that exits on same bar as entry.

[ZachSwann]

I am looking for the trend version of the selection bias template ?

[Nick Radge]

We prefer to use MCS with trend systems because of start date dependency.

[ZachSwann]

Is there any reason you wouldn’t use them both together?

[Anonymous]

Has anyone worked to convert the MOC template to a short-only version? I have been working on a version that gives me the correct chart signals and appears to be working fine, however, on occasion will show that I get filled at the high of the day when the limit is above the high of the day.

I suspect this is due to the CBT section which I don’t really fully understand in order to edit accurately.

Just thought I would reach out to see if anyone has worked through this already and would be willing to help.

[Stephen James]

Send the file to me Dustin.

[RobGiles]

I am experimenting with a MR system that holds positions up to 7 bars. To eliminate SB I am using the MR_CBT_template. I want to rank all the buy signals by a short period RSI and buy the lowest 20 RSI reading stocks.

Do I simply replace the existing code:

Rank = 100 – ROC(C,3);

with something like:

Rank = RSI(C, Period); ?

How do I get it to rank from the lowest reading to the highest (i.e. so that the weakest stocks are at the top of the list)?

[Author]

Posts