But on the other hand, if the market is saying there are some risks, and on average it will underprice risk, perhaps its now when the tail risks are present and should be bet on. That's the tough thing about OTM put buying and other forms of tail betting. A lot of the huge profits come in periods where options are overpriced (or they 'look' overpriced) In the 87 crash situation, it would be VERY easy to consider taking profits before Black Monday. Markets were down -14% in a short period of time. They were oversold (and options were probably expensive) but the real money was made AFTER they were expensive. Taleb says in a video that in the US stock market 80% or so of the kurtosis is explained by the largest outlier (or something to that effect). If you miss out an big event because you took profits too soon (as the options looked 'expensive'), you might have to wait 30 years to get lucky again Its hard very this approach, you got to wait so long and need so much luck to see any green
I never have been a big fan of Monish Pabrai but its interesting what he says on this video He talks about how he used to run his funds by having a "10 by 10" portfolio. If he found a stock he liked, he would put 10% in there. Overtime, as he made mistakes, he changed his approach. Now he has a checklist to avoid previous mistakes (from him and others) and now a typical allocation is 5% and he 'will go as low as 2%'. He also talks about how the key is to not lose money, key is to protect the downside. This looks like a version of 'idiot proofing' a investment plan to me. Which I think is smart
In line with this idea I run into a paper from Robert Merton http://www.people.hbs.edu/rmerton/onestimatingtheexpectedreturn.pdf Its quite relevant to the portfolio research I'm doing. Merton shows that estimating returns so one knows what is the 'risk reward ratio' of a portfolio is quite difficult. Apparently, historically there has been a lot variation in the actual excess returns vs what some models (with assumptions in it) have predicted. There are many problems with estimating returns to build a balanced portfolio: If the model assumes excess returns (mostly the equity risk premium) are constant, when there is a shift in the equity risk premiums, the portfolio will have too much or too little in stocks. If the model assumes excess returns are related to the historical variance (volatility) of the stock market, then when volatility shifts (from historical norms) the model will break down. And volatility shifts quite a bit. Merton talks about " an average annual standard deviation of 27.9 percent for the period July 1926 to June 1946 versus 13.8 percent for the period July 1946 to June 1978" If the model assumes the market always wants the same risk reward ratio (so its constant) and that risk reward ratio is related to variance, similar problems emerge as variance changes. People might want to take more risk in certain periods as well so the ratio might not be constant But of course, all those 3 model assumptions are certainly wrong. People's tolerance for volatility change, historical volatilities change and the required 'risk-reward' ratio changes. So far my tests have dealt with this problem by trying to find the best historical portifolio and then going from there. The issue is that I'm using historical risk premiums that might not be the same for the future. It also relies on historical volatility (the denominator part of the Sortino ratio) which is also subject to change. The only part of my system that is solid is the reliance on the best Sortino ratio (risk-reward ratio), I'm assuming that because people HAVE to invest, the best Sortino will be preferred over other portfolios and that is constant (one can always lever up the best portfolio if they want more returns). An alternative approach is this one http://thierry-roncalli.com/download/erc.pdf In order to remove the problems of having to estimate returns, this ERC approach simply takes historical volatility and builds a portfolio with the % allocations based on those volatilities. No asset class is allowed to have a higher risk allocation than another. So if stocks have 3x the volatility of bonds, if one had to build a portfolio of these 2, the portfolio would have 75% in bonds and 25% in stocks. 3-1 ratio of bonds to stocks (the inverse of the volatility ratio). I think that makes a lot of sense, and it could be superior since it isolates only one problem (estimating volatility), my previous approach relied BOTH on volatility and estimates of return. So there was 2 ways it could go wrong
If I weight by Standard deviation and Standard deviation of Negative Returns then my optimal portfolio is: Stocks are 2.7 times more volatile (SD) than 10y bonds and 1.07 times the vol of gold. If one were to build a only stock and bond mix, it would be 73% bonds, 27% stocks. To get the gold allocation I remove bonds and stocks at the same ratio all the way down to a 15% allocation to gold (where I'm finding that the diversification/rebalance benefits start to degrade) so: -10.95% in bonds -4.05% in stocks Final result is 62% bonds 23% stocks 15% gold 2.69 units of 10y bonds for 1 unit of stocks to 0.65 unit of gold By SD of Negative Returns (which I like more than SD) Stocks are 3x more volatile (to the downside) then bonds and 1.43 than gold 75% bonds and 25% stocks. Taking gold to 15% and removing the other things in the right ratio results in -11.25% -3.75% 63.75% 10y bonds 21.25% stocks 15% gold 3 units of 10y bonds for 1 unit of stocks and 0.705 unit of gold What if I put the 30y bond in this? Using the 2-1 duration ratio rule of thumb It would 1.5 unit of 30y bond for 1 unit of stock 31.875% 30y bonds 21.25% stocks 15% gold with 31.875% in cash If I distribute that cash at the current ratio among the stocks and bonds (leaving gold out because the benefits have maxed out at ~15%) +23.906% bonds +7.969% stocks resulting in 55.781% 30y bond 29.219% stocks 15% gold But that might be too much in a single point of the yield curve. If I put intermediate bonds at a 15% weighting (to maximize diversification/rebalance benefits) and remove bond US bonds and stocks (keeping gold at the optimal 15% level) -11.25% bonds -3.75% Final result 44.53% 30y bonds 25.46% stocks 15% intermediate bonds 15% gold That's very close to my previous optimal portfolio but its nice to see that it converged to a similar answer. I like this ERC approach more (using SD of negative returns not SD), its simpler and I can tweak things by putting in a worse assumption of SD of negative returns to make it more robust
This 15% allocation as a rule of thumb for optimum diversification/rebalancing benefits is likely to be true only when there is a small menu of asset classes avaliable. When there is more, the % is smaller. Indeed, Dalio talks about in Market Wizards how 80% of the volatility can be cut with 15 uncorrelated return streams "Ray Dalio: This chart shows how the volatility of the portfolio changes as you add assets. If you add assets that have a 0.60 correlation to the other assets, the risk will go down by about 15 percent as you add more assets, but that’s about it, even if you add a thousand assets. If you run a long-only equity portfolio, you can diversify to a thousand stocks and it will only reduce the risk by about 15 percent, since the average stock has about a 0.60 correlation to another stock. If, however, you’re combining assets that have an average of zero correlation, then by the time you diversify to only 15 assets, you can cut the volatility by 80 percent. Therefore, by holding uncorrelated assets, I can improve my return/risk ratio by a factor of five through diversification." http://valueinvestingbasics.com/ray-dalio-explains-asset-correlations/ 0.00 correlation asset classes are rare but assuming a low correlation its possible that the 15% allocation is too high (as it would only allow a maximum of 6.66 asset classes). If I had more asset classes avaliable I would be able to put smaller %s in them and even forced gold and intermediate bonds to give up some allocation to 'make room' for a new asset class There are also some questions with regards to how rebalancing affects all of this. Dalio's math is based on standard deviation alone, it doesn't consider the effects of rebalancing In any event, my 15% rule seems to work in my case as I don't have access to large menu of assets in the US data case
Well thank goodness you clarified that. I've started day trading again and if I have open positions it's a struggle to fit in dinner sometimes. I was wondering how you manage to day trade, read extensively, think about it all and post here at length. Mystery solved. I'll have to be selective about my hours once I have the data to decide, at my age I don't want to be trading all the time and doing nothing else.
For my Greek backtest I'm getting 2 results for the Computer allocation (68% 10y Bonds, 20% stocks 12% gold). One with Bloomberg data using 10y bond yields with the 2012 'patched' up using my NPV loss but changing it to reflect a bigger rally according to the data. So instead of -21% I used -19% Tks https://www.elitetrader.com/et/members/ch1973.223211/ for help with data BBG data: -46% real loss. Total drawdown -65% on an annual basis. At the depths of the panic this was worse And one with S&P data that already has all the data (no NPV in it) -53% real loss. Total annual drawdown -72%. In the depths of the March debt exchange, this was worse With BBG data and no rebalancing total loss was -40.58% (-54.66% max drawdown) With S&P data and no rebalancing total loss was -50% (-61.25% max dradown) So rebalancing worsed results With my new optimum portfolio (posted earlier today) and no 30y additions, the losses dimish a bit due a higher gold allocation: -43.5% total loss with BBG data -50.3% total loss with S&P data -36.3% total loss with BBG data and no rebalancing -45.22% total loss with S&P data and no rebalancing These results were pretty rough. But it has to be taken into account that the bond market returned -60.92% in that 10y period The stock market returned -89% in that 10y period So there was no hiding. Gold and cash (EUR in Germany/Swizterland or in the backyard) were the safe havens. And foreign assets of course, a simple 15% allocation to foreign assets would have decreased the drawdowns materially What did I learn: -Check and recheck your data, if it looks too good to be true, its probably is -For portfolios that lack enough diversification, rebalancing increases risk (it also didn't help there was not enough time for markets to recover in the period studied) -Global diversification is huge -Despite losing half their money, Greek investors that followed a balanced approach (even though flawled due not having foreign assets) still beat their benchmarks (10y bonds and Stock Indexes). They also learned a big lesson is not investing all they got at home
I try to only be glued to the screen for the first hour (the best hour), after that I even force myself to leave the room, go do other things. I dont want to spend my life watching charts every day for 6 hours for the rest of my life. It just feels like a waste
This market is powerful, I covered IWM short and will reasses if the market reverses. Market is not afraid of anything Trump is doing
Greek Computer portfolio returns vs asset classes (€100 invested adjusted for inflation) Its easy to see why the rebalancing was harmful. The portfolio kept selling a rising gold allocation to plow into a collapsing bond and stock market. This might work out in the end (over the coming years its likely the bond and stock market will perform quite well) but it increased drawdowns and volatility, which is usually not the case