Optimisation - A waste of time?

General Discussion on the Permanent Portfolio Strategy

Moderator: Global Moderator

Post Reply
User avatar
Hal
Executive Member
Executive Member
Posts: 1349
Joined: Tue May 03, 2011 1:50 am

Optimisation - A waste of time?

Post by Hal » Fri Oct 14, 2022 4:42 am

Just pondering... On reading articles stating that the PP is a "Risk Parity" solution I was wondering if "Optimisation" actually works

So attached are two different optimisation approaches. Risk Parity and Minimum Variance. Neither look successful in 2022. And just for a hypothetical, if the available asset classes were cash/10 yr bonds/ gold/ shares, would a "optimised" approach be best or just 1/4 of each?
Attachments
Returns - Risk Parity.png
Returns - Risk Parity.png (49.04 KiB) Viewed 1404 times
Risk Parity.png
Risk Parity.png (93.53 KiB) Viewed 1404 times
Returns - Min Variance.png
Returns - Min Variance.png (48.82 KiB) Viewed 1404 times
Minimum Variance - Robust.png
Minimum Variance - Robust.png (94.41 KiB) Viewed 1404 times
Aussie GoldSmithPP - 25% PMGOLD, 75% VDCO
Kbg
Executive Member
Executive Member
Posts: 2815
Joined: Fri May 23, 2014 4:18 pm

Re: Optimisation - A waste of time?

Post by Kbg » Fri Oct 14, 2022 8:17 am

The fundamental problem with optimization is that it is a view of the historical past and highly dependent on starting and end dates. Do the same studies and try a couple of alternatives.

Do a rolling 10 year optimization (take 10 year chunks and run the optimization 1972-1981, 82-91 etc.)

Do a move forward optimization (start at 1972 to 2022 and then go forward to 1977 or 1982 but using all data)

Bonus points: do the rolling 10 (or 5) optimization and then forward test for a period of time to see what happened using the optimized portfolio based on the previous period.

This won't take too long and is well worth the time to internalize the lessons a bit deeper.

So what is optimization good for? For me I use it to try and understand asset relationships better under differing economic and market conditions. This implies one also understands/is familiar with the history of both.

To use a current example; we are most like the 1970s to early 80s at this point in time. So what happened? How did things perform? When did they change? How's what happened then comparing to today. Potential answers: Gold...disappointing. Stock/bond correlation becoming/is correlated as expected. Return performance also not great (as per history).

Datawise...it's super hard to depend on things statistically. Just not enough data. Plus the real world and it's conditions are what matter. For example stocks in the US have been a great investment since the 1980s. Japan, not at all. An optimization will mirror reality and that's it. They never project.

My take: useful yes, but not as a means to predict or expect the same outcome in the future.
User avatar
Hal
Executive Member
Executive Member
Posts: 1349
Joined: Tue May 03, 2011 1:50 am

Re: Optimisation - A waste of time?

Post by Hal » Thu Nov 10, 2022 4:16 am

Kbg wrote:
Fri Oct 14, 2022 8:17 am

My take: useful yes, but not as a means to predict or expect the same outcome in the future.
Just for fun. What a difference a year makes :D
Attachments
To Dec 2021.png
To Dec 2021.png (81.16 KiB) Viewed 1076 times
To Oct 2022.png
To Oct 2022.png (96.69 KiB) Viewed 1076 times
Aussie GoldSmithPP - 25% PMGOLD, 75% VDCO
Kbg
Executive Member
Executive Member
Posts: 2815
Joined: Fri May 23, 2014 4:18 pm

Re: Optimisation - A waste of time?

Post by Kbg » Tue Nov 15, 2022 10:56 am

No kidding...and kinda demonstrates the point!

I normally use some form of averaging with segmented data to arrive upon potential allocations (or values for trading systems).

Dual Momentum isn't talked about much anymore, but an application example is to not use a single 12 mo lookback but use 3-12 month lookbacks (for example). Thus, there are 10 signals each month vs. an all or nothing signal and the effect is you phase in and out of percent allocated to US, Foreign or bonds.
Post Reply