So lets say you've got a system employed and during your testing and what not you stumble upon something more attractive. At what point and in what way do you decide to move to the new system? Do you immediately exist all existing positions and start from scratch or what?
Also, when testing using the same parameters but on different markets is it totally ridiculous to take the markets that result in negative average trades out of your test? Classic case of portfolio optimization but perhaps those markets don't trade as well.
When to implement a new system
-
- Full Member
- Posts: 10
- Joined: Fri Oct 16, 2009 10:42 am
Re: When to implement a new system
I'll bet that YOU can invent three or four different ways to "move to the new system". Why don't you post some of your ideas here, along with your opinions about the pros and cons of each one?HWG wrote: in what way do you decide to move to the new system? Do you immediately exist all existing positions and start from scratch or what?
Perhaps your thoughts will resonate with readers; or perhaps readers will surprise you by saying "You left out ideas X, Y, and Z (as follows), and these are just as good if not better than the five ideas you proposed."
-
- Full Member
- Posts: 10
- Joined: Fri Oct 16, 2009 10:42 am
Well the plain as day way would be to close out existing positions and start from scratch. Running the new model from the previous ones start date would not work b/c you would have an equity not equal to the actual equity in the account plus existing positions not on. I think the best way would be to start the test from the current day with the current equity and somehow build the new model on top to recognize the existing positions. So if it recognized existing positions it could create exits for them as well as new entry orders.
-
- Roundtable Knight
- Posts: 1436
- Joined: Mon Dec 22, 2003 12:24 pm
- Location: Des Moines, IA
- Contact:
Re: When to implement a new system
I think you are setting yourself up for a hurtin' down the road if you remove poor performing markets. Who says those same markets have to be bad in the future?HWG wrote:...is it totally ridiculous to take the markets that result in negative average trades out of your test? Classic case of portfolio optimization but perhaps those markets don't trade as well.
Also, sometimes even though a particular market lost money over the life of a test, doesnt mean it didnt contribute in some other fashion, like reduce drawdown.
-
- Full Member
- Posts: 10
- Joined: Fri Oct 16, 2009 10:42 am
Re: When to implement a new system
Yeah, I totally get that and figured as much. Best bet is probably to select as uncorrelated a group as I can.Levi wrote: I think you are setting yourself up for a hurtin' down the road if you remove poor performing markets. Who says those same markets have to be bad in the future?
Also, sometimes even though a particular market lost money over the life of a test, doesnt mean it didnt contribute in some other fashion, like reduce drawdown.
You might choose to view this as a "Should I do X, or should I do Y?" kind of question. X stands for (Trade the old system) and Y stands for (Trade the new system).
Or equally well, X could stand for (make the transition from old system to new, according to These Rules) and Y could stand for (make the transition from old system to new, according to Those Other Rules).
Then you could apply the principle of "minimizing the maximum remorse" as discussed for example in some message threads here.In brief: Why assume you must do one OR the other? Why not do a little of each? If you do a little of each you are guaranteed to participate in whichever one has the most favorable outcome, which reduces (the remorse of betting on a horse that doesn't subsequently win the race).
Or equally well, X could stand for (make the transition from old system to new, according to These Rules) and Y could stand for (make the transition from old system to new, according to Those Other Rules).
Then you could apply the principle of "minimizing the maximum remorse" as discussed for example in some message threads here.In brief: Why assume you must do one OR the other? Why not do a little of each? If you do a little of each you are guaranteed to participate in whichever one has the most favorable outcome, which reduces (the remorse of betting on a horse that doesn't subsequently win the race).
Re: When to implement a new system
Perhaps you can also find a compromise here by trying to minimize the maximum remorse of different scenarios: 1) the poor past performers actually turning out to be the best future performers; 2) the past optimal portfolio actually proving to be the best future performing portfolio; 3) none of the previous options; and 4) you figure it out.HWG wrote:Also, when testing using the same parameters but on different markets is it totally ridiculous to take the markets that result in negative average trades out of your test? Classic case of portfolio optimization but perhaps those markets don't trade as well.
-
- Full Member
- Posts: 10
- Joined: Fri Oct 16, 2009 10:42 am
back to my original post, I think what I'm more interested in is determining when to add new markets to my current system. I should've clarified that. I just don't want to add new markets starting say tomorrow b/c then the past trades will be reflected in my current equity. So keep the same system but trade more market is the goal.