Len’s Journal

[JulianCohen]

Scott McNab wrote:

“Selection bias loosening”

A career in media relations beckons…

Are we having a brain fade here guys ? After working so hard and producing such great systems with low selection bias ! A plot of buysignals seems to indicate that much of this year has been a period of low volatility not particularly conducive to mean reversion….so loosening of selection bias would produce many more trades…..looking back through the years such periods can often be seen but don’t last forever….its almost like the market is tempting us to fall for that mistake….the Sirens call…..?

You’re probably right Scott. Falling for the old ‘slow trading so fiddle with the system’ trap!

However I might have a fiddle just for the sake of it

[LEONARDZIR]

Julian,
Yes there are more signals than I can t take on 10% of the days.
I was also running about 97% or better but I decided to experiment by increasing selection bias( increasing trade frequency) to increase CAR and see what happens with live trading.
On my MOC I changed the stretch and on my MR I decreased the number of days of pullback and stretch.

Since June 1 my MOC is +3.4% and MR is flat. MR was pretty good heading into the month but lost all my June and july profits this month.

[LEONARDZIR]

Scott, I agree with you there is a risk by increasing selection bias. On the other hand I believe my selection bias boundaries are within Nick’s recommendations.

[LEONARDZIR]

A correction and an observation.
I also played with price range in my MOC system to increase CAR and change selection bias.
That said I hadn’t run my selection bias numbers for awhile. So I just ran them on my MOC and MR systems.from 2011 to present

MOC capture 97% days, 94% of trades.
MR capture 94% days and 88% trades.

In order to keep the selection bias on MR similar to MOC I will have to drop the CAR on the MR

So after thinking this through over the last week I see no advantage to trading an MOC and MR system with similar entries. So I will concentrate on my MOC, US MOMO and NASDAQ MOMO

After thinkin

[LEONARDZIR]

I have a question about MOC and selection bias.
MOC system returns appear to be highly dependent on trade frequency. Yet trade frequency is limited by selection bias.
In the last year the US volatility has fallen dramatically and so the number of daily setups has fallen as well.
In my MOC system(Russell 1000) I can take up to 40 positions at 10% per position. On days when there is less than 40 buy setups I have taken to increase my buysetups by loosening my entry criteria with a goal of getting as close to 40 daily buysetups as possible. Presumably when volatility increases (buysetups increase) I will tighten the criteria. So trying to squeeze as much trade frequency as possible without degrading selection bias in a low volatility setting.
Anybody else doing a similar strategy? Appreciate any comments.

[JulianCohen]

How has it worked for you Len? Are you getting more trades and if so, are they trades of good quality? My concern is that opening the selection bias might give poorer quality trades…has that happened?

[LEONARDZIR]

Julian,
I just started doing it so time will tell. However when I back test with more liberal entry criteria in this environment my MOC CAR shoots up dramatically and at least when I test since 1/1/17 there is no selection bias.
I feel like I am adapting more to current market conditions. I think I will have early warning signs that things have changed when my buysetups start increasing in frequency on a daily basis.
I started thinking about this when I noticed that Scott increased his frequency of trading without increasing selection bias by trading US and ASC MOC.

[SaidBitar]

Len Zir wrote:

I have a question about MOC and selection bias.
MOC system returns appear to be highly dependent on trade frequency. Yet trade frequency is limited by selection bias.
In the last year the US volatility has fallen dramatically and so the number of daily setups has fallen as well.
In my MOC system(Russell 1000) I can take up to 40 positions at 10% per position. On days when there is less than 40 buy setups I have taken to increase my buysetups by loosening my entry criteria with a goal of getting as close to 40 daily buysetups as possible. Presumably when volatility increases (buysetups increase) I will tighten the criteria. So trying to squeeze as much trade frequency as possible without degrading selection bias in a low volatility setting.
Anybody else doing a similar strategy? Appreciate any comments.

i think the only way to be sure that everything is ok is to have large sample size and some random data
i think if you modify the Close to become a bit more random and the thing that is controlling the volatility then you can be sure that all is working good or not

Module 25

this is what i use

[LEONARDZIR]

Said,
Ran my MOC with 10% random close and no major decrement in performance. Basically I increased my universe of stocks I use by decreasing the volume requirements. This increased the frequency of trades by approx 20%. Despite the increased frequency of trades since 1/1/17 no selection bias issues.

[JulianCohen]

I would think that if the selection bias stays the same then take the highest CAGR you can achieve. Only when it affects selection bias that is the point to stop widening the criteria.

For me any movement in stretch or criteria affects the selection bias.

[JulianCohen]

Len Zir wrote:

Said,
Basically I increased my universe of stocks I use by decreasing the volume requirements. This increased the frequency of trades by approx 20%

Len if I may, what did you reduce the volume requirements to?

I have kept to AveVol = MA(V,50) > 500000

Are you less than that or did you use a turnover filter/aswell?

[LEONARDZIR]

Julian,
That is exactly what I am doing, raising the trade frequency but trying to limit selection bias to less than 1 to 2%. I went from a volume of 1 million shares filter to 500,000 shares.
I will plan to review my trade frequency and selection bias on a regular basis and adjust accordingly.

[JulianCohen]

OK got you. I’m already at >500K so I don’t fancy going less than that.

[LEONARDZIR]

August

US momo -0.2%
Nasdaq Momo +4.5%
US MOC -0.5%

[Nick Radge]

Quote:

Nasdaq Momo +4.5%

Nice…

[Author]

Posts