Reducing Entry Signals

[ScottMcNab]

I think Levente said to me once that API can access 100 stocks at a time…but could design so cycled through 100 at a time

[LEONARDZIR]

After 6 months of trading my MOC system I am still trying to get my head around the selection bias issue. I believe Nick stated he gets 97% of all trades and days. I have managed to get to 90% of trades and 95% of days where all signals were taken but no better. I improved my signal ratio using Nick’s adaptive position sizing which was helpful.
My question is what is the minimum acceptable % of trades taken and minimal number of days all signals taken to assure realtime results match backtesting? I see different numbers on the forum. I believe Said mentioned that if you take 80% of trades you would be ok although I didn’t see any evidence. Is there any way to predetermine the correct numbers or do you just trade and see if results fall within MCS testing?
Len

numb

[Nick Radge]

Len,
My apologies on tardiness to complete that task. I actually have a sticky note to complete it.

I will get onto it today and also answer those questions.

Nick

[ScottMcNab]

What sort of change (%)/improvement did you see with the adaptive position sizing Len ? Your stats sound pretty good

[LeeDanello]

Doesn’t the trade skipping in Monte Carlo Analysis meant to address the problem of selection bias? So establishing that you can only takes 75% of trades you could adopt a 25% trade skipping parameter to simulate expected results.

[LEONARDZIR]

Scott,
I was playing around with the usual culprits on my MOC system ;like universe, stretch and degree of pullback all the time trying to keep 2:1 leverage. but it wasn’t until I could put on at least 40 positions that I could get reasonable CAR but get reasonable signal/trade and percentage of days with all signals taken. The only quantification I can give you is that at 20 positions at 10% per position my CAR was a little over 40% and with 40 positions at 5% the CAR was about 23% so I assume I will get somewhere between the two.
Len

[ScottMcNab]

Thanks Len. I wasn’t very clear in my question I’m sorry. It wasn’t the CAR (%) I was interested in but what % of days were all signals taken and what % of total trades…going from original position sizing to Nick’s adaptive position sizing…eg from 80% to 90%. I share your interest in selection bias and am thinking I should look to implement an adaptive method perhaps ?

Out of interest, while we talk about this mainly in reference to the MOC systems, would it not equally apply with any system not using a ranking system to select stocks (ie a MRV / swing system using a buylimit or buystop order) ?

[LEONARDZIR]

Scott,
I think this answers your question. I ran my system with max 20 positions and had all signals traded on 86% of days, and 77% of all trades taken. When I ran the system with 40 max positions 95% of days all signals were taken and 89% of all trades were taken. Obviously the tradeoff is lower CAR.but a more reasonable chance of negating selection bias. So Nick’s adaptive positioning allows me to increase the frequency of trades and still have reasonable success of tackling the selection bias issue..
I agree selection bias is probably a bigger issue MRV swing. I just started trading an MRV system with max 20 positions at 10% per position. There are frequent days when I am holding 10 or more positions and have many more signals with potential trades that the system can take..

[ScottMcNab]

Thanks for the info and thoughts Len….much appreciated

[LEONARDZIR]

Maurice,
I think you asked a very important question. Namely does MCS testing with 25% trade skipping account for the possibility of selection bias. I don’t know the answer but I think Nick can answer that question. My guess is that you can still have significant selection bias despite the results of MCS testing.

[Nick Radge]

I’m almost finished the expanded section on Selection Bias. You can read that specific answer HERE

[JulianCohen]

Nick Radge wrote:

I’m almost finished the expanded section on Selection Bias. You can read that specific answer HERE

Bing! :ohmy:

Lightbulb goes on inside thick head. So easy when you know how eh!

[Nick Radge]

I aim to please.

[ScottMcNab]

What little hair I had is rapidly disappearing..

Anyone else finding a similar pattern in backtesting as show below ? .this for 2015…buysignals stay below positions available for almost entire year but then for 3 days in August it all blows up… should I just live with this ?

[Nick Radge]

If its only a single day then I wouldn’t worry about it. That single say is not going to sway the system too far one way or other.

Remember – it’s not about perfection. It’s about finding a balance.

[Author]

Posts