Global Macro Trading Journal

[Daal]

Brazil's stock market returned 39% last year in BRL terms but one of the duration sensitive bond indices returned 31%. So it didnt trail it by all that much. Another ones produced from 21 to 25%. Cash returned 14%. Inflation was ~6.25%.
My minimax portfolio probably would have beaten cash by a few percent(the gains are offset by losses in USD net of gains from AGG) but over an entire cycle, I doubt most folks trying to time markets will beat it. And the ones that do must be very good at what they do and deserve their fees. But I dont think there are many of these out there. Especially if they have bearish tendencies. Its easy to be bearish and 'beat' markets by staying out of plunges. Its another story when your apocalypse doesn't come and risk premiums keep shrinking. Thats when funds go out of business. I learned this lesson very well from 2009 to today. All the bearish geniouses from 2008 proved to have just been lucky to have a bearish thesis during the right time. They kept their bearishness over the cycle and missed out huge.

 

[Daal]

Found some further data on IG corporates
(Jan 1983 - Dec 2016)

https://www.portfoliovisualizer.com/efficient-frontier

1987-2016 with Total Bond Market (I assume its an index like AGG) and long-term corporates

The huge Sharpe ratio (specially compared to Treasury bonds) and limited correlation to the stock market provides further proof that my suspicion that IG corporates (with limited duration) were pretty good at minimizing 'regret' was correct. Seriously, this thing looks like the ultimate asset class for risk averse investors. Folks try to make 8-9% in stocks and get like 2-3% due human nature. They were better off gunning for a 5-6% historical and actually getting it
Right now they are on the more expensive side (spreads are quite rich) but still, I think its a very important asset class to monitor

 

[Daal]

Looking at the site I'm seeing that short-term Treasury has a return, SD, Sharpe of
"Short Term Treasury 5.58% 2.31% 2.415"
1983-2016
So short-term corporates are as good as T-Bills in terms of Sharpe. But yet, they will produce a higher real return. This thing is like the holy grail

 

[Daal]

That said, right now its probably a pretty bad time to get involved in IG corporates. I'm sure a lot of folks are aware of that sharpe ratio and they are playing all kinds of carry strategies. They borrow at low rates and jam into corporates trying to make a nice spread. And this has going on for years. That is probably why IG corp spreads are so tight (around 50-75bps). There is not much juice left at this point in the cycle and the Sharpe ratio going forward will be pretty bad because you dont get much extra return but get plenty of volatility if spreads blow out.

So, right now, they are pretty bad. But as an asset class, long-term, they are pretty awesome

 

[Daal]

Looking at the site I'm seeing that short-term Treasury has a return, SD, Sharpe of
"Short Term Treasury 5.58% 2.31% 2.415"
1983-2016
So short-term corporates are as good as T-Bills in terms of Sharpe. But yet, they will produce a higher real return. This thing is like the holy grail

The numbers for short-term corporates are probably a little bit overstated because it is essentially mimicking Short Term Treasury. Both return and vol look pretty close. So if one were to go a little bit further in duration in the corporates, the sharpe ratio would probably drop. But the return would pop too. 1.5-2 looks more accurate overall. Only thing that is competitive is Total Bond Market, which I assume is a mix of bond things, like an AGG type index

 

[Daal]

I wish I knew about all of this in 2009. I could have made a very good profit VERY safely back then. All of that while remaining defensive.

When markets bottomed in March and started to rip, I build a short position in stocks that I thought were going to go bust. They included some financials and certain REITs. I think that was an ok bet, every pop had been a short for over a year so I was playing that. And I did some fundamental work in those companies.

But my mistake was not doing a "Soros dual book" trade. I should have longed about 25-50% of my networth in IG corporates to 'hedge' those shorts, I didnt. But what kills me is that on some level I understood this concept back then. I was long Citigroup and ILFC (AIG subsidiary) bonds, I thought the US gov would stand behind these debts. This provided me some hedge for my shorts and it did cushion my losses a little, that was my goal. But I needed to have a bigger 'bullish' book to offset my bearish book. I could easily have built that with LQD. Those things were priced for a great depression. And even in the 30's IG bonds didnt do that badly, especially in real terms.
I didnt do that and as a result, I got squeezed hard on the shorts without full compensation on the longs.

Soros on the Trump rally, got killed on the shorts but made on industrials and financials. Overall he was probably flat even though he was wrong.
I could have gotten the same outcome. 2009 was a good year for IG corporates. So was 2010,11,12.

That mistake cost me dearly, it wasn't just about the money it was more the overall lesson. I would have learned to close my eyes and take risk when everybody is afraid much earlier in my investing career. This would have produced better returns down the line, in other investments. I also would have the financial/emotional reference of having made the right decision right in the eye of a hurricane. That's the type of thing you carry forever with you as a moment to be proud of. As Howard Marks says, you can count on one hand the amount of people that predicted the crisis, reversed stance and bought risk in 2009. It would have been nice to have done that, even if a more defensive version of it

 

[Daal]

Thats why I sometimes go crazy on the big Brazil fund manager who kept trashing the market in 2016. When the market is down so much, valuations are cheap and sentiment is terrible at the very LEAST you go defensive with an "long IG corporates" type position. And IG corporates are EQUIVALENT to 20-30% equities and 70-80% gov bonds. So essentially, it makes NO SENSE to have 0% equities.

 

[Daal]

Thats why I sometimes go crazy on the big Brazil fund manager who kept trashing the market in 2016. When the market is down so much, valuations are cheap and sentiment is terrible at the very LEAST you go defensive with an "long IG corporates" type position. And IG corporates are EQUIVALENT to 20-30% equities and 70-80% gov bonds. So essentially, it makes NO SENSE to have 0% equities.

This manager, in his last monthly report, now disclosed that he has some position in Brazilian equities and that he is buying gov bond duration (equivalent to buying stocks in the case of Brazil). So after talking shit about the market for a year he threw the towel and joined the trend. To his credit it only took him a year. Rosenberg needed several years in the US crisis. Roubini to still day is yet to recommend US equities (i think). I had seen this movie before, that's why I felt so comfortable ignoring the managers, pundits who said the BR market sucked.

Its just too hard to predict a crisis, change your mind and buy near the lows. You cant count on people doing that even if they have a great track record or are good analysts

 

[Daal]

So when I run tests in portfolios using annual data from 1928-2016, the computer suggested to me that the optimal portfolios were:

I speculated that the wide difference in allocation to stocks between the best Sortino portfolio and the best Sharpe portfolio were due to the fact a lot of stock volatility gets 'smoothed' out using annual data. And this effect was greater in the Sortino compared to the Sharpe.
Now, I got monthly data and have started to run tests. It pretty much confirms the suspicions. In monthly data, stocks get punished, a lot. The computer doesn't like it because it so much more volatile. But what is nice is that the Sortino and Sharpe seem to have converged into a consensus.

US Bills = 3month T-Bills
US Bonds = 10y T-Bond
US Stocks = S&P Composite from Robert Shiller

Sep 1926 to Dec 2016. The CAGR is not really the annual rate but the monthly real return rate

What's not important is the specific numbers allocated to US bills vs US bonds. A lot of similar performance to those above can be achieved without US bills

Pretty much the same Sharpe and Sortino as the top performers but wihout US Bills. What I'm focusing on is the ratio between US bonds vs US stocks and the Gold % allocation.
With Gold is easy, the computer is consistently finding that the optimal amount is around 12%. Cool. With US Bonds is more complicated because the ratio is to US stocks seems to be anywhere from 3 to 1 to 4 to 1. But that's only because I'm using the 10y T-Bond. Presumably, if one has access to 30y bonds (like everyone has now). You could own less in bonds and still have more duration. This would ENABLE the investor to own more stocks since one is 'hedged' against turmoil

So while historically the computer wants a 60-70% allocation to US bonds and 20% in stocks, in current times perhaps one can structure something with more duration in a less capital intensive basis (while still retaining some in shorter maturities to have the possibility to pick up in yields during rising rate enviroments). This additional flexibility would enable the investor to own more stocks

 

[Daal]

The thing is, I got even more data then 1926-2016

I got monthly data going back to 1871 for stocks, bonds and the CPI. But not for bills and gold. With gold I can go back to 1879 (with gold its not that hard because the US was in a gold standard so the price didn't change much). So I decided to build a 'synthetic' t-bill series so I could test going back to 1879

Its not great and not very scientific but just to be able to take a look at a longer data I decided to give it a shot. I just took the avg spread difference between 10y bonds and 3m T-Bills (around 106bps) and extrapolated that to the past. So if in any given year the 10y bond was a 4%, i just subtract 106bps and assumed that was the 3m t-bill rate.
It's really a shitty method given that the correlation between 10y bonds and 3m bills from 1926 to 2016 is only 0.08 and the R2 (regression analysis) is 0.51. This makes the 3m bill excessively volatile and unreliable. It also makes the yield curve relationship dissapear since the curve will pretty much never invert or steepen more. But given that the computer wasn't finding much value in T-Bills from 1926 to 2016, I figure that punishing them more won't hurt all that much but I could learn something more about stocks, bonds and gold

As it turns out the results were a little surprising

What is interesting is that the computer converged into this consensus is its best to own 1 unit of the stock market for every 3-4 units of 10y bonds. With longer data the computer wants more bonds but I think that's a little bit of the consequence of the drop in Gold allocations. Gold got hurt bad with more data because of the Gold standard. During the standard the gold price didn't change all that much (nominal changes were small so the changes only happened when inflation moved a ton), and because inflation was kept under control (from 1879-1926) the additional data was negative to gold so the computer didnt had the need to have as much gold to boost the Sharpe/Sortino (When I run this on my Brazil data, I suspect the computer will load up on gold)

But what is also interesting is that on the longer sample, bonds were killing it! The monthly real returns of bonds were 0.1821% and for stocks, they were 0.1878% (compared to 0.1757% vs 0.233% on 1926-2016 data). In the 'extra' period, bonds compounded 0.19% real a month vs 0.1% in stocks. But even though bonds were killing it, the computer still favored a 20% allocation to sotkcs. It seems that 20% is more less a floor of how little the computer wants to own in stocks