System33 (SUPPORT-RESISTANCE BUGS SHEET signal only)

Systems home page    System33_Support06      Latest trades    Contact     Bookshop    Trivia

SYSTEM33 TEST 3 original version (CODE SHOWS 1% TOLERANCE!!)

This problem is not relevant in 240 minute and above, but if short term then its relevant.

 

BUG 1

 

System is long: close hits max loss stop and exits, but the exit level is ALSO on support, so it buys

it on the SAME BAR immediately after exit.

Then it figures the stop needs to be hit again! Because TS cant determine which trades maxloss is being referred to.

Suggesting the exit stop needs to be NESTED within the support06 code and set to 0.01  or 1%

 

Output of this bug:

 

Makes an instant 3 points loss on each occurrence.

Makes a good trade die when it could do well.

I estimate 500 points has been lost over the testing period with this bug. So lets see if the bug fix works.

 

 

 

Fix ideas.

1.

Obviously best is a DLL, but that's going to cost.

 

2.

Tell the stop if its a support level, then get further away, so 2 stops will be used maxloss and

intelligent maxloss?

 

Tell the signal not to buy if any position was open on the currentbar or the previous bar? But this isnt so good as the above idea. The problem lies in the stop exit and the buy entry coming at the same instant, so a bit of code maybe can cure this.

The new signal will be named support07_v1

 

Idea for bug fix, make a brain variable, that knows +1 if support buy position, and tell the stop not to do anything if it = +1.

Realising this is perhaps dangerous, a 2nd wider stop will have to made with another 10-40 points slop added to it.

Any further buy signals must not take place until that bar is dead and gone in history.

 

The bug is corrected now so all thats left is to optimise the support tolerances.

The signals have now been split into 2 parts which are named thus

 

07Support_v1part1 (just over support signal)

07Support_v1part2 (low cross signal)

 

An RSI(C,5) Filter has been added to stop buying non sold out levels.

Optimisation show the best trigger to be around 41 (eg RSI(C,5) MUST BE LESS THAN 41 before buying this signal.

See optimisation sheet

 

BUG 2

 

Some signals are on the high peaks which obviously isnt correct, so a simple RSI less than N filter should correct this problem.

 

 

BUG 3

 

The tolerance of how many % above or below support needs to be somehome fractal, so it doesnt have to be altered each time.