Portfolio Charts

[Tyler9000]

Quick update: Testing the results, I found a mistake in my 5-asset calculations. Correcting that, the top returning portfolio with no more than a 10% drop in any individual year is small cap value, emerging markets, long term treasuries, 5-year treasuries, and gold (7.29% real CAGR). Basically, a bunch of Golden-Butterfly-like portfolios top the list (two types of stocks, two types of bonds, and gold).

This new data set is pretty darn cool. My favorite thing is that it shows a bunch of possible solutions rather than one overly-specific solution that will change every year and may contain assets you're not comfortable with. I need to take some time to check the results and think about how to best apply them, but I'll let you know when I have something up to let others explore the data as well.

To answer ThisDinosaur's question -- I actually originally created many of these tools to study and explain the Permanent Portfolio, my personal asset allocation of choice. The more I play with the data the more I think the Golden Butterfly is something not too different that will work well for me personally. But there's no rush to change, and others may have different needs and preferences.

[jacob]

@Tyler9000 - How far back do your datasets go? Sorry if this is well-known or googlable. One issue with US equity (large, mid) is that it's a mature and widely owned market. There are, therefore, not as many buyers (relative to current prices) compared to EMs and smaller (new) companies. Equity ownership in the US is pretty widespread compared to all other countries and has been for a several decades.

[Tyler9000]

I have data to 1972. That's the longest I can find for the wide variety of assets, and it's reconstructed from a multitude of sources. See Simba's Backtesting Spreadsheet.

I see your point. Like I said, backtesting alone doesn't have all of the answers. There's often more to the story, and it's important to consider what economic conditions made a particular portfolio work.

[fips]

@workathome:
You also increased your annual turnover from 0 to 190% and in the second simulation only hold 2 assets
But for a quick check the graph looks nice nonetheless. Nice to see you on board.

@Jacob:
Afaik Tyler uses Simba's backtesting spreadsheet with the datasets going back to 1972.

[jacob]

@Tyler9000 - Yeah, in my experience, backtesting is best used to avoid strategies that are non-obviously wrong rather than trying to find the "bestest" one. This is because the risk of model-fitting/p-value optimization is high (the problem of telling genius from a rising tide). Annoyingly, 1972 is kind of shitty/not-so-random year to start the time-series because it was the beginning of a new economic/regime (i.e. post Bretton-Woods) and marked the beginning of asset inflation because cash had become fully fiat after that. On the other hand, it's a good starting year because you don't have noise from the past, say 1945-1972 (US hegemony) to skew the results. 1972--2001 was a kind of rising tide. The potential problem is that 2001-now has been a new regime, so 1972-2001 skew the current results.

[Tyler9000]

Well on the plus side, the calculations are all real. The asset growth in the 70's and early 80's post Bretton-Woods doesn't look so hot once you account for inflation. And I do like that the 70's are in the data -- too many stock investors like to only look at data starting in the mid-80's and have no real sense of history.

[jacob]

*cough* Dave *cough* Ramsey *cough*

[workathome]

@workathome:
You also increased your annual turnover from 0 to 190% and in the second simulation only hold 2 assets
But for a quick check the graph looks nice nonetheless. Nice to see you on board.

The test includes slippage. It's designed to only hold 50% of the PP with the highest relative momentum. It holds cash if any asset falls below the 210-day simple moving average. It's a pretty commonly tried, basic momentum strategy applied to ETFs based on Faber's "Tactical" paper. I like how it works with the super simple permanent portfolio and doesn't depend on 10+ asset classes. From what I understand, momentum is the most supported market anomaly.

Yeah though, it would work best in a tax-deferred account with the turnover.

[fips]

Slippage is of little relevance in this case but yes, momentum is still one of the most popular market anomalies.
See your inbox to not derail this thread.

[JamesR]

So when I come up with a portfolio that approaches 8% Sustainable Withdrawal Rate (25% Small Cap Value, 25% Emerging Markets, 25% Int'l Small, 4% Long Term Treasury, 7% Commodities, 7% Gold, 7% REIT), how big of a grain of salt do I need to take this with?



Would it be worth having an option that adds a +20% increase on every historical bottom for backtesting? Would making it worse historically help wean out strategies that are a bit too optimistic?

I applied a sma rule and relative momentum to the permanent portfolio and got some nice results.

http://imgur.com/a/8glkl

I don't understand what's going on here?

[Tyler9000]

So when I come up with a portfolio that approaches 8% Sustainable Withdrawal Rate (25% Small Cap Value, 25% Emerging Markets, 25% Int'l Small, 4% Long Term Treasury, 7% Commodities, 7% Gold, 7% REIT), how big of a grain of salt do I need to take this with?

Would it be worth having an option that adds a +20% increase on every historical bottom for backtesting? Would making it worse historically help wean out strategies that are a bit too optimistic?

The important thing to remember about SWRs is that they are merely a maximum limit subject to a very specific set of assumptions. More years of data looking backward or forward can never raise the SWR, but it can absolutely lower it.

Rather than inserting your own returns assumptions, sometimes it helps to look at trends. For your specific portfolio, note that the SWR does not go down at all past 25 years -- that's because one of the most recent 25 year retirement periods is setting new lows and driving down the rates. Plug it into the Hurricane calculator with the same withdrawal rate, and you'll see what that looks like in more tangible terms. Since lighter colors are from more recent start years, portfolio trends are relatively simple to see. If you're looking for something a little less sensitive to timing, try to find a portfolio with an even distribution of Hurricane colors.

So similar to what Jacob said earlier, it's best to use withdrawal rates to determine what not to do rather than think of them as a guarantee. Be smart about it.

[Tyler9000]

Since you asked for it ...

Here is an easy and specific one to start with:
What portfolio of three evenly distributed assets provide the highest overall CAGR since 1972 (regardless of volatility)?

A "seek" feature like this would be neat.

I put in a minimum CAGR I'm satisfied with, for example, and search for the portfolio(s) that hits that, with the lowest volatility.

Or vice-versa (put how much volatility I'm fine with, and search the portfolio that gives the highest CAGR).

Bonus points if I can uncheck assets, so it only searches the ones I want.

I imagine the computing power (and thus time it'd take to run this) might be prohibitive though.

FYI -- the new Portfolio Finder is now up on the site. Enjoy!

http://portfoliocharts.com/2016/03/07/t ... investors/

[JL13]

May I have the opposite? Which portfolio has the highest volatility and the lowest CAGR?

I just want to know where I'm going to die so I don't go there.

[Tyler9000]

May I have the opposite? Which portfolio has the highest volatility and the lowest CAGR?

I just want to know where I'm going to die so I don't go there.

You see those lonely three portfolios floating way out to the bottom left of the cloud? One is simply small cap growth. One is commodities. The other is gold. Put all of your money in just one, and you're going to have a bad time. But buy equal parts of all three, and the resulting portfolio has a CAGR of 5.24% with a worst year a more respectable -28%. Still not a barn burner, but back into the "normal" range.

[JamesR]

Portfolio finder is pretty neat, you're getting pretty good at implementing these things!

And as a tangible proxy for risk I prefer the single worst year of portfolio returns

After playing around with portfolio finder, and then looking up the earlier mentioned 8% SWR portfolio on pixel charts. I noticed it had a longest drawdown of 4 years and the worst year was -37.3%, so I realized that it could be stressful and not that great after all.

I suspect that I'm likely to be more sensitive about the number of years with negative returns than the worst year, so for example I'd easily tolerate -20% for one year, but I'd be pretty worried if it lasted 3 or more years and went below -35% cumulatively. So a tangible proxy of risk could be more of a combination longest drawdown & cumulative drawdown perhaps?


(** I'm just guessing about my feelings because I haven't really had investments for a long time and dealt with up and downs like others on this forum have. I'm also in my early 30's and not necessarily risk averse.)

[Tyler9000]

So a tangible proxy of risk could be more of a combination longest drawdown & cumulative drawdown perhaps?

I've kept it intentionally simple to get everything up and running, but I'm interested in eventually adding other measures as well. Longest and/or deepest drawdown over consecutive years would be interesting.

Yeah, most people overestimate their own pain tolerance for down years until they truly experience it, and sometimes it's hard to get past just talking about returns with young investors who have yet to live through the tradeoffs of high-risk strategies. I've learned over time not to focus so much on discussing "risk tolerance" in absolute terms, but to present alternative more desirable paths to the same goals.

[Dragline]

This is really outstanding work/analysis. Looking at that scatterplot above, one of the things that jumps out at me -- besides it looking like a face down human with no limbs -- is the general curvature of the shape, which shows that trying to goose returns just a few percentage points really increases the possibility of serious drawdowns.

It would be interesting, at least to me, if you were to turn this on its side (flip the x and y axes) so it looks like a human falling backward and see if you can fit a curve to the data. It obvious forms a power-law/exponential function.

I am envisioning that the best way to "play" this would be to take a lower average return/drawdown and leverage it by a factor somewhere between 1 (meaning no leverage) and 2. Then you have a competitive hedge fund.

[Tyler9000]

It would be interesting, at least to me, if you were to turn this on its side (flip the x and y axes) so it looks like a human falling backward and see if you can fit a curve to the data. It obvious forms a power-law/exponential function.

Cool idea -- that would indeed be an informative graphic. I'll see what I can do.

[jacob]

@Dragline - I don't think that's a powerlaw. I think you're seeing one side of a fat-tail of a Bell Curve (which is kind of a powerlaw but only in a mathematically nerdy way) that largely adheres to modern portfolio theory (but not quite since it's fat).

[Tyler9000]

Another way to look at it is that for every given worst year there's roughly a 4% range of potential returns depending on your portfolio choice. The sharp turn around 6-7% is where you've exhausted the diversification potential and have to start adding risk to increase returns.

I also find this interesting:

The orange dots represent the typical spectrum between a 100% total US stock market portfolio on the left and a 100% total bond market portfolio on the right. One might normally think that half the portfolios would be more efficient than the total market and half would be less efficient, but in reality the basic stock/bond blends pretty much follow the inefficient frontier of the larger pool of options. The imbalance above the line represents the diversification bonus, and adding additional assets is more likely than not to have a positive effect on your portfolio.

I also suspect that the "head" at the top left represents the preponderance of all-stock portfolios. I'm looking into that.

This chart is fascinating in its own right, and I'll probably have some sort of write-up soon.