|
Buffalo Bill
 Legend
  Posts: 539
Joined: 10/3/2006
Location: Stafford, VA
User Profile |
OK so you go thru the ATM 3 Auto Trend method and get it where you like it. Now you want to trade it. Soooo how much BT bars does ATM3 need to do a good analysis for solid walk forward results?
Right now the profile goes back to 2010. That's much longer than you need when you deploy the method. I assume you would also check "use all prior data" so ATM is always using the latest WF results. You need enough data for ATM3 to be able to figure out the best combo of strats/symbols but how much is that? Using monthly increment periods for WF, is it 6 months? 1 year?
I found 210 analysis bars improved my results over the default of 160, so I think that's the absolute min, but you need bars PRIOR to the start of the analysis for EMA calcs, etc (if you want things to be accurate and complete). So if you use a 200 EMA somewhere you're at 410 bars loaded as a min for example. So maybe last 2 yrs of bars, or 500 bars, in your data period BT setting in the to do
Thoughts?
IDK.
|
|
Jeff Drake
 Veteran
     Posts: 289
Joined: 11/17/2003
Location: Austin, TX
User Profile |
HI Bill,
Very good post. For the walk forward you do indeed need a warmup period. Since the walk forward is limited to the forward test, the back test can serve as the warmup period. 1 year (~250 bars) is probably sufficient, but 500 bars is what we used in early testing just to play it safe.
Hope that helps,
Jeff Drake
|
|
peanutbrittle
 New User
Posts: 3
Joined: 9/4/2005
Location: Phoennix
User Profile |
What is confusing me is the Back Test Period in the "To Do List" and the Back Test Period in the ATM3 Port Sim. Does BT in the "To Do List" feed settings to the port sim or is Port Sim totally independent of the settings in the BT of the "To Do List"?
And wouldn't a longer BT in Port Sim cover more market states and get better settings?
[Edited by peanutbrittle on 2/25/2020 11:46 AM]
|
|
Barry Cohen
 Sage
      Posts: 6338
Joined: 1/19/2004
User Profile |
The ToDo List test settings are used for voting analysis, & Port Sim can only analyze the results from the ToDo voting analysis. So if you analyze 2 years of data in Real Mode, Port Sim will not be able to give results outside of those 2 years. In other words, if you want a longer test period in Port Sim, you need to increase the test period & analyze in Real Mode first.
|
|
peanutbrittle
 New User
Posts: 3
Joined: 9/4/2005
Location: Phoennix
User Profile |
So my understanding is that selecting "Both" for the BT/FT setting in Port Sim says to do Portfolio performance testing over both the Real Time BT/FT period settings.
The values used would be that found in the BT period in the Real Time mode.
[Edited by peanutbrittle on 2/26/2020 1:17 PM]
|