Rob’s Journal

[LEONARDZIR]

Rob, here is my perspective on developing your first system. Keep it simple! You will be amazed that you can get a robust and profitable system with 3 or 4 rules and 5 or 6 filters ( to reduce selection bias. ) The best observation I’ve seen on successful systems is that you should be able to explain it to somebody that knows nothing about trading and you should be able to jot it down on the back of a napkin.
Len

[JulianCohen]

Len Zir wrote:

Rob, here is my perspective on developing your first system. Keep it simple! You will be amazed that you can get a robust and profitable system with 3 or 4 rules and 5 or 6 filters ( to reduce selection bias. ) The best observation I’ve seen on successful systems is that you should be able to explain it to somebody that knows nothing about trading and you should be able to jot it down on the back of a napkin.
Len

I can second that. All the systems I originally built and traded have now been simplified to very simple algorithms.

[LeeDanello]

Rob,

I like to write my systems on toilet paper. That way you can get a continuous piece of paper and least you can recycle the systems when they don’t work out. Seriously as you would know, use the kiss principle. Easier said than done though.

[RobGiles]

Thanks for everyone’s input.

How many degrees of freedom did you end up with on your MR systems?

I’ve tested the following and after tweaking all the variables, results are either poor returns or high returns with very low trade frequency (not enough to bother with):

Price has to trade through bottom BBand (100) [have payed around with 50 & 20 period]
3 lower lows
Price of stock has to be above 100, 50, 20 or 10 period MA (I’ve tested them all. I added this filter as didn’t want to be long a stock that was tanking)
0.5 x ATR Stretch to get a fill following day [tested up to 1.50 x stretch and everything in between)

Adding an Index filter (say S&P500 > 200 period MA) kills it.

Not sure where to go to from here.

Tested over 5 years

[Nick Radge]

Quote:

Not sure where to go to from here.

Where you go from here is have Craig sign off on the code. Once that is done send the code with default parameters set and we discuss how to get a result you deem appropriate.

[RobGiles]

I thought I was told not to send you systems that didn’t show some kind of promise?

[JulianCohen]

You might be surprised at how much can be squeezed out of what you have suggested in your system.

[SaidBitar]

Rob Giles wrote:

Thanks for everyone’s input.

How many degrees of freedom did you end up with on your MR systems?

I’ve tested the following and after tweaking all the variables, results are either poor returns or high returns with very low trade frequency (not enough to bother with):

Price has to trade through bottom BBand (100) [have payed around with 50 & 20 period]
3 lower lows
Price of stock has to be above 100, 50, 20 or 10 period MA (I’ve tested them all. I added this filter as didn’t want to be long a stock that was tanking)
0.5 x ATR Stretch to get a fill following day [tested up to 1.50 x stretch and everything in between)

Adding an Index filter (say S&P500 > 200 period MA) kills it.

Not sure where to go to from here.

Tested over 5 years

here are some hints:
make the BB duration shorter , way shorter than 100
use lower number of STDev
when you are happy with the final results
start cleaning up the conditions
you take out one condition and check the results this you will understand the effect of each condition

[RobGiles]

Thanks Said

Started playing with shorter duration BB last night and got a dramatic improvement and mucked around with the stretch which yielded some interesting results.

[JulianCohen]

Rob Giles wrote:

Thanks Said

Started playing with shorter duration BB last night and got a dramatic improvement and mucked around with the stretch which yielded some interesting results.

Be careful with the stretch. Test for selection bias before you start to like the results too much. In my limited experience, CAGR gets better and better as selection bias gets worse.

It’s all a balancing act.

[LeeDanello]

I use a 5 day Bollinger band with 1 std deviation and catch long trades that fall through the bottom band. No trailing stops just exit at the 5 day MA or a timed stop.

[RobGiles]

Thanks Julien…I’ve just worked that out and had a good session with Nick discussing why.

[RobGiles]

Thanks very much Maurice…..what’s your worst DD if you don’t mind me asking?

[LeeDanello]

I designed it for the ASX and the drawdown on the backtest is about 13 to 14% unleveraged.

[RobGiles]

OK thanks. Good luck with it.

[Author]

Posts