Daniel UK1
Member
Posts: 470
Registered: 4-6-2019
Member Is Offline
|
|
Portfolio Optimisation/Rebalancing of systems
The second largest issue after the strategy itself, at least for myself, is to decide which systems to run and not in your portfolio.
The subject is important and thought it could deserve its own thread, if admin thinks differently, just move it.
I think we all have the same situation, after some time you are sitting on a quite large library of strategies for each market, and then you are
running them on a portfolio.
BUT, what decides which systems to run, how often to rebalance them and what are the defined rules for doing this ?
Myself i am right now picking top x systems per market looking at top NP/DD every 2 years, which would mean around 100 trades before rebalancing.
And would also exchange any systems at any time of the X systems above if it surpasses the largest DD it has had in the past.
What is yours preferred choices? comments on my approach is greatly appreciated.
Its an interesting subject for us all i believe. 
Cheers
Thanks received (1):
+1 loclhero at 2021-06-30 15:20:57
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Im letting pa decided, but force pa to only allow 1 contract per strategy.
This forces more diversification.
Im only likely to re balance when I have a better system to add, or one to kill.
Big topic and I dont claim to be an expert.
Thanks received (1):
+1 Daniel UK1 at 2019-10-04 08:07:50
|
|
|
Daniel UK1
Member
Posts: 470
Registered: 4-6-2019
Member Is Offline
|
|
Thanks for feedback, appreciate that...so as i understand you are letting PA decide, but based on exactly what criteria, ?
You rebalance, when you have a better system to ad, or when one to kill..
Do you have a rule set for this?
- How do exactly define a "better" system ?
- how do exactly define when a system is going to the grave and you "kill" it ?
Sorry for being annoying
Yes best thing i believe would be to run one strategy per contract as you say, however in the long run from an operational perspective, i believe for
myself that it would create issues, since i would prefer to run one strategy per sub account on same market in order to avoid risks such as stratgeies
going long when another are short, and you end up with the stops in place but no position.. (Have actually happened to me), after that i want to rule
out that operational risk. So in the long run and you want to run above 10 contracts on a single market, that would for me require 10 subaccounts or
more...
Problem would be solved though if one would run long and short separately, but not doing that.
Much smarter people than myself in this business are suggesting that in this case, they would prefer to run ALL strategies, and just max positions
that you cant have at one time.. however not liking this due risks of lots of losses and reduced profitable signals, due to clustering.. and
difficulty comparing backtests with reality due to not taking all the signals from a single strategy.
Thoughts appreciated
|
|
|
RandyT
Member
Posts: 123
Registered: 5-12-2019
Location: Colorado, USA
Member Is Offline
|
|
I agree that determining when a strategy has failed is one of the most challenging parts of this game.
I am attaching a paper that I have collected recently that presents an interesting theory as to how to determine when a system has broken. Searching
the web for this paper and authors will also turn up some other discussion.
I think adding this algorithm to PA would be very useful.
Attachment: Login to view the details
Thanks received (2):
+1 loclhero at 2021-05-21 10:57:27 +1 LucaRicatti at 2019-12-19 10:51:59
|
|
|
Daniel UK1
Member
Posts: 470
Registered: 4-6-2019
Member Is Offline
|
|
Right, totally agree.
This would also include when and how to define when to exchange a strategy on same symbol to a better one.
Than you for the paper, will look into it.
Regards
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
I only had a 1 minute look at the paper. This is a complex topic, and ive never spoken to some one who things they have nailed it.
My feeling is anything that looks at the trade sequence alone is flawed. You need to factor in market range, trending, how other different but some
what comparable systems are doing to.
|
|
|
Daniel UK1
Member
Posts: 470
Registered: 4-6-2019
Member Is Offline
|
|
In a perfect world i would like to see a single strategy, as a component in a trading system for one symbol, like an ensemble system.. meaning that
each single strategy votes to go long or short, and according to a set rule, acts upon it... and then each single strategy matters less of the
outcome result.. but this is a quite complex subject and not perhaps a straight solution to be subject of this thread.. But something i work on in my
own research.
Right now, i rebalance my strategies 1-2 times per year, sorted on NP/DD..
How do you do?
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Quote: Originally posted by Daniel UK1  | In a perfect world i would like to see a single strategy, as a component in a trading system for one symbol, like an ensemble system.. meaning that
each single strategy votes to go long or short, and according to a set rule, acts upon it... and then each single strategy matters less of the
outcome result.. but this is a quite complex subject and not perhaps a straight solution to be subject of this thread.. But something i work on in my
own research.
Right now, i rebalance my strategies 1-2 times per year, sorted on NP/DD..
How do you do? |
I manually do similar from time to time, letting pa-pro choose systems.
GSB systems with closeD secondary filter will never go long / short at the same time
|
|
|
SwedenTrader
Junior Member
Posts: 63
Registered: 3-8-2020
Location: Stockholm, Sweden
Member Is Offline
|
|
I just want to add that this checkbox has always worked for me, that several automated strategies in Tradestation can go both long and short at the
same time in the same instrument with no issues. The net result is always correct in the end, it just looks different during the trades since the
position is net 0 and that is a good thing for the margin instead of having separate sub-accounts. That is just more headache i think with several
accounts.
Thanks received (1):
+1 admin at 2021-01-03 16:07:57
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by SwedenTrader | | I just want to add that this checkbox has always worked for me, that several automated strategies in Tradestation can go both long and short at the
same time in the same instrument with no issues. The net result is always correct in the end, it just looks different during the trades since the
position is net 0 and that is a good thing for the margin instead of having separate sub-accounts. That is just more headache i think with several
accounts. |
I think this should be the case on all platforms, but test on a sim account if in doubt
|
|
|
Carl
Member
Posts: 343
Registered: 10-5-2017
Member Is Offline
Mood: No Mood
|
|
Wanted to share this 34 minute video about portfolio construction by Andrea Unger. IMHO the first 27 minutes are interesting.
https://ps.ungeracademy.com/titan
The video is an introduction to Andrea's course and introducing his Titan software.
|
|
|
admin
Super Administrator
Posts: 5069
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
i ONLY looks at a few minutes. Very jaded today. TS had multiple total failures during USA day time. My alertmon woke me up 3 times. Very rare for
this to happen, but reboots, delete cache didn't fix it.) Now doing repair of ts and restore of workspaces from previous zipped up copies.
He basically said drawdown due to systems that are too corealated due to time filters and being a bit similar.
Taken me over a decade to get around this, even though I was aware of the issue.
Thats also part of the logic behind me selling zonetrader and overnightES. These are not GSB systems and trade very differently.
|
|
|
![[*]](images/davis/default_icon.gif){kind=icon}
