although I am very happy with how my designer models work out, I would like to understand how to properly revise models if needed.
As far as I understand, one saves the current locked-in live portfolio as a simulation, makes the intended adjustments, and chooses that new simulation to replace the current designer model.
When I save my current designer model from my live portfolios as a simulation and run it again (with what I think should be the same settings), I receive very different results (for example, the designer model currently holding 10 stocks while the simulation with the same settings now holds more than that).
Are those changes due to the changes that were made to the database, so that the simulation works with new data all the way back to the start date, while the designer models adjust on-the-fly?
Probably there are proper instructions and explanations for the differences somewhere but I haven’t found them yet.
So maybe someone could point me in the right direction?
After the revision,
Based on the revised model rank and buy and sell rules the next re-balance day ie., next Monday
the profile will get balanced; sell based on the new rules; and buy based on the new rules;
Please, don’t expect the revised designer model will hold all the current holdings as simulation. but over period of time it will match; as most of the current holdings sold and re-balanced.
As i have 2 models, My SP500 Rev2 - to preserve historical performance with minimum changes;
My SP500 Rev3 clean slate with more than a years of lessons and reflection from Rev2.
My question is regarding how to properly re-vise models.
I thought the proper way was to save your current live designer model as a simulation, change this simulation and overwrite the current live designer model with the modified simulation of that same model. However, even unmodified, the freshly-saved model I save from my designer live model has a different performance than the live model itself.
I got your question, p123 revising some fundamental and earning data time to time to improve accuracy, just re-run existing designer models which is designed few years ago will show degraded performance.
quality of data is definitely improved over periods of time. Y
Humble the market, market will humble back.
Thanks
Kumar
For detailed information specific to your question, I suggest shooting an email over to Paul DiMartino. He is P123’s Customer Service Rep and can be reached at paul@portfolio123.com. I have found his assistance to be invaluable on multiple occasions and highly recommend his service.
/Chris
Note: This has been an unpaid public service announcement.
Thanks - I was just hoping to keep my nitty-gritty questions off P123 staffs’ backs so that they can focus on improving P123, but I think I’ll get back to Paul.
The methodology that you describe is accurate: “save your current live designer model as a simulation, change this simulation and overwrite the current live designer model with the modified simulation of that same model”.
At that point two things happen: We check to make sure that the model is still OK per the Designer Model rules, and we then replace the simulated results with the modified simulation. We do NOT replace the prior out-of-sample history, and it is likely that the next rebalance will have many trades.
Can the simulated results be worse than they were originally? Yes, it’s possible. Simulations can differ from past simulated results on Portfolio123 because of data corrections or engine changes.
