Portfolio Allocation Timing Models

Ask your investment, budget, and other money related questions here
Nomad
Posts: 196
Joined: Wed May 16, 2018 5:23 pm
Location: UK

Portfolio Allocation Timing Models

Post by Nomad » Sat Jun 08, 2019 11:30 am

Hello there
In the past I've favoured a fix asset class percentages in a portfolio.
However, I've been playing around with market timing models on a website https://www.portfoliovisualizer.com/tes ... ingModel=4
Basically, you take money out of asset classes when they aren't doing well and put them into something that is doing really good.
You do this 'rebalancing' every month or quarter etc.

Now, it sounds like a lot of effort but the results seem to be rather amazing, you don't succumb to crushing devaluations of a portfolio due
to a stock market crash - because these generally take a year to 18 months or more to hit the bottom.
Additionally, you are quick to jump onto any new band wagon - while the trend is being your friend.
Generally, you are only 'in' 3 or 4 asset classes at any one time.
Over a few years the gains over a fixed percentage portfolio are impressive.

Has anyone else been doing this and what works for you?

arcyallen
Posts: 86
Joined: Sat Jan 20, 2018 11:20 am

Re: Portfolio Allocation Timing Models

Post by arcyallen » Wed Jun 12, 2019 7:13 pm

Nomad wrote:
Sat Jun 08, 2019 11:30 am
Basically, you take money out of asset classes when they aren't doing well and put them into something that is doing really good.
If by "aren't doing well" you mean asset class has gone down in price and if by "doing really good" means asset class has gone up in price...isn't that selling low so you can by high?

As far as what's worked for me - not doing that :)

Nomad
Posts: 196
Joined: Wed May 16, 2018 5:23 pm
Location: UK

Re: Portfolio Allocation Timing Models

Post by Nomad » Thu Jun 13, 2019 5:01 pm

arcyallen wrote:
Wed Jun 12, 2019 7:13 pm
If by "aren't doing well" you mean asset class has gone down in price and if by "doing really good" means asset class has gone up in price...isn't that selling low so you can by high?

As far as what's worked for me - not doing that :)
It's more based on which markets have momentum. E.g. if the NASDAQ has been in a bull market for 12 months, you assume that it will probably stay
in a bull market for one more month. The alternative is to assume every current trend will someone reverse next month?
Using the website and picking say seven asset classes but only being invested in 3 at a time gives great results for historical data.

You can only lose money in a month where more than 1 of your asset classes changes from a bull to a bear market.
The month after that you would not be 'in' any of the asset classes that have reversed direction.
Last edited by Nomad on Thu Jun 13, 2019 5:12 pm, edited 1 time in total.

bigato
Posts: 2126
Joined: Sat Mar 05, 2011 12:43 pm

Re: Portfolio Allocation Timing Models

Post by bigato » Thu Jun 13, 2019 5:09 pm

This is speculating and goes generally contrary to the usual way of rebalancing portfolios, that is sell whatever category is above the designed percentage of the theoretical portfolio, and buying more of whatever is below the designed percentage. I'm no judging whether this is a good thing or not nor if you should do it. Just noting that it is something else than the common strategy for rebalancing.

Nomad
Posts: 196
Joined: Wed May 16, 2018 5:23 pm
Location: UK

Re: Portfolio Allocation Timing Models

Post by Nomad » Thu Jun 13, 2019 5:19 pm

bigato wrote:
Thu Jun 13, 2019 5:09 pm
This is speculating and goes generally contrary to the usual way of rebalancing portfolios, that is sell whatever category is above the designed percentage of the theoretical portfolio, and buying more of whatever is below the designed percentage. I'm no judging whether this is a good thing or not nor if you should do it. Just noting that it is something else than the common strategy for rebalancing.
The common strategy for rebalancing is to have a fixed ratio of say 5 or 6 asset classes and to preserve the ratio come what may.
Alternatively, this is a dynamic strategy where you are only in say 3 of the 5 or six asset classes at any time e.g. the ones doing the best over
the last six month period.
Generally, markets keep going in the same direction for months if not years, then unpredictable stop and can reverse for months.

This even works on individual stocks like large cap stocks, I tried that too.
I would say that it isn't speculating, it would be speculating if I was better on things changing direction I'm betting next month is similar to the last few - generally that's true and that's why it works. Try it out with historical data.

bigato
Posts: 2126
Joined: Sat Mar 05, 2011 12:43 pm

Re: Portfolio Allocation Timing Models

Post by bigato » Thu Jun 13, 2019 5:25 pm

In order to properly evaluate this strategy, you need to stablish exactly what the rules are and how to apply them, and then you need to run simulations pretty much the way Tyler9000 did with his spreadsheets that then gave birth to his porfolio site. Only then you'd have reliable estimations of how effective the strategy would probably be in the future.

Nomad
Posts: 196
Joined: Wed May 16, 2018 5:23 pm
Location: UK

Re: Portfolio Allocation Timing Models

Post by Nomad » Fri Jun 21, 2019 1:08 am

bigato wrote:
Thu Jun 13, 2019 5:25 pm
In order to properly evaluate this strategy, you need to stablish exactly what the rules are and how to apply them, and then you need to run simulations pretty much the way Tyler9000 did with his spreadsheets that then gave birth to his porfolio site. Only then you'd have reliable estimations of how effective the strategy would probably be in the future.
Yes, that is what the link on the original post allows you to do for either asset classes or even original stock tickers.
I've started doing this strategy as of about a month ago, so I could perhaps create a separate thread on the fund choices and their performance.
Or could post them here or on my journal page.

Seppia
Posts: 1077
Joined: Tue Aug 30, 2016 9:34 am
Location: Italy

Re: Portfolio Allocation Timing Models

Post by Seppia » Fri Jun 21, 2019 2:50 am

What's your time horizon?
If it's "next month" then go for it, if it's "the next 30 years" statistics (and common sense, I would add) would suggest you buy the stuff that's cheap.

Nomad
Posts: 196
Joined: Wed May 16, 2018 5:23 pm
Location: UK

Re: Portfolio Allocation Timing Models

Post by Nomad » Fri Jun 21, 2019 7:31 am

@Seppia
My time horizon is 2-4 years - depending on returns and desired expenditure.
I have started implementing this strategy, details to follow.

frihet
Posts: 190
Joined: Thu Jun 06, 2013 10:33 pm
Contact:

Re: Portfolio Allocation Timing Models

Post by frihet » Sun Jun 23, 2019 4:37 pm

I use a 3 fund momentum strategy for a part of my assets, at the moment about 10% and plan to increase it over time as I gain confidence, sell other allocations.

As i wrote in another thread the inspiration/ info is from "traden om strategier for fonder" cred to those much smarter than me, who share freely.

The basic idea is to create a list of uncorrelated assets. Then once a month rank them on 3 month performance switching to the 3 funds that performs best. Uncorrelated is important as you don't want to switch unnecessary, but capture the inevitable trends in different markets. Diffrent risk categories is also crucial as that is what limits draw-downs during recessions.

Recent example from the forum thread mentioned. This was the list shown to be most conducive for the Swedish fund universe. But I do believe that back testing for the US would show something similar.

Sector Funds

Finance
Healthcare
Real Estate nordic countries
New Technology
Real Estate global

Other Stock funds

Small Caps Global
Russia
Sweden Small Caps
Asia Small Caps
Nordic Countries Small Caps

Medium Risk

50/50 stocks/bond mix fund
Natural Resource fund

Low risk

TIPS
Longer Bonds 1
Longer Bonds 2
Longer Bonds 3

Interesting to note is how gold and short rates was excluded during this most recent back test. They used to be included, over fitting? Time will tell. Also note how small caps are favored over large caps. Makes sense, they have historically performed better and the strategy's aim is to avoid them and switch to lower risk during recessions anyway.

Nomad
Posts: 196
Joined: Wed May 16, 2018 5:23 pm
Location: UK

Re: Portfolio Allocation Timing Models

Post by Nomad » Sun Jun 23, 2019 4:49 pm

@frihet
I'm glad someone else is interested in this approach. It is probably a good idea to gradually easy into it. It is difficult to find uncorrelated asset classes
and an additional issue is that the portfoliovisualizer.com site hasn't 'heard' of most of the funds I invest in but is aware of the sectors.
So I'm using that to identify the likely sectors. Then using my broker site to tell me the returns of the the last 3/6/12 months periods.
I will be revealing my first months returns on 6th July and what I'm investing in for the next month.

bigato
Posts: 2126
Joined: Sat Mar 05, 2011 12:43 pm

Re: Portfolio Allocation Timing Models

Post by bigato » Sun Jun 23, 2019 5:43 pm

Backtesting it, which is the swr that you can expect from this strategy?

Nomad
Posts: 196
Joined: Wed May 16, 2018 5:23 pm
Location: UK

Re: Portfolio Allocation Timing Models

Post by Nomad » Sun Jun 23, 2019 6:48 pm

I've done a quick test using these asset classes.
VTSMX VEIEX VGSIX GLD VFISX VPACX (total stock market, emerging markets, REIT, GOLD, short term treasury, pacific stocks)

Using a strategy of being in only 3 at a time, the SWR appears to be about 9.5% from 2006 to 2019.

By way of comparison the SWR of those assets in an equally weighted portfolio is 5.7% for the same period.

The difference 5.7% to 9.5% has huge implications for the amount that needs to be saved.

Another thing to mention is that a very small percentage of actively managed funds someone manage to beat their market.
Investing only in these funds in the identified asset classes could yield additional returns.

frihet
Posts: 190
Joined: Thu Jun 06, 2013 10:33 pm
Contact:

Re: Portfolio Allocation Timing Models

Post by frihet » Sun Jun 23, 2019 6:53 pm

bigato wrote:
Sun Jun 23, 2019 5:43 pm
Backtesting it, which is the swr that you can expect from this strategy?
https://i.imgur.com/GfvfuCg.png Here is the data for the allocation above from portfolio visualizer. The Link only Work as web version not on IPad/phones, Seems unbelivable, and is not numbers I would base an ER on. But I do think there is potential for quite a runaway portfolio if you use a traditional 4% SWR. This potential is why I consider this strategy for a part of my assets as a possibility for real wealth building.

bigato
Posts: 2126
Joined: Sat Mar 05, 2011 12:43 pm

Re: Portfolio Allocation Timing Models

Post by bigato » Sun Jun 23, 2019 7:10 pm

Considering only data from 2000 to now is very unreliable, isn't it? It was a very peculiar period in history.

frihet
Posts: 190
Joined: Thu Jun 06, 2013 10:33 pm
Contact:

Re: Portfolio Allocation Timing Models

Post by frihet » Sun Jun 23, 2019 7:27 pm

Yes that's true and that is part of the reason that I would never believe that this is a sure thing, it's a bit of a gamble. After all back tests are about the past. Investing happens n the future. But if you believe in the momentum effect and that asset classes will move in different trends in the future I think it's a fair chance it will continue to work.

I know both 'Meb Faber and the investing for a living blog has come up with similar results from backtests

https://investingforaliving.us/2013/07/ ... -momentum/

https://allocatesmartly.com/list-of-strategies/

The best thing is probably to diversify between several strategies. Some built on momentum and some on value. I've seen that suggested, makes sense.

bigato
Posts: 2126
Joined: Sat Mar 05, 2011 12:43 pm

Re: Portfolio Allocation Timing Models

Post by bigato » Sun Jun 23, 2019 7:30 pm

Well, one thing that you could do is to try and tweak the traditional portfolio allocation strategies for that same short period of years and see if you can come up with something that has similar returns. You probably can, and that just means that if you squeeze a small sample hard enough, you can get pretty much any results you want from it. I personally would only consider it with a much bigger historic data. Other than that you're gambling, which is fine if that's what you want.

Seppia
Posts: 1077
Joined: Tue Aug 30, 2016 9:34 am
Location: Italy

Re: Portfolio Allocation Timing Models

Post by Seppia » Sun Jun 23, 2019 7:32 pm

Quick question: are you considering friction costs in your analysis?

Nomad
Posts: 196
Joined: Wed May 16, 2018 5:23 pm
Location: UK

Re: Portfolio Allocation Timing Models

Post by Nomad » Sun Jun 23, 2019 7:48 pm

Seppia wrote:
Sun Jun 23, 2019 7:32 pm
Quick question: are you considering friction costs in your analysis?
For one of my accounts the cost of buying or selling funds is zero.
For the other account it is £1.25 for funds.
ETF's and shares are different however, they are about £11.95 commission per transaction.
The only ETF I have is a Gold ETF, so it's just one asset out of the lot.
Suffice to say the effect is negligible

Nomad
Posts: 196
Joined: Wed May 16, 2018 5:23 pm
Location: UK

Re: Portfolio Allocation Timing Models

Post by Nomad » Sun Jun 23, 2019 7:52 pm

bigato wrote:
Sun Jun 23, 2019 7:10 pm
Considering only data from 2000 to now is very unreliable, isn't it? It was a very peculiar period in history.
You can use prior data but you have to tweak the asset classes accordingly. For some of them the data starts in 2000.
It wasn't that peculiar, stock market crashes happen a lot and if you keep a sizeable amount of a portfolio in stock all the way down,
a huge amount of capital will be lost - compared to switching the money to something else that isn't crashing like possibly gold.

Post Reply