Up 1.2 percent. Quite a good week, back to all time highs, about six and a third percent up for the year. Stocks and bonds are very negatively correlated right now, so the bonds allocation goes up. On July 2nd, 2020, the strategy rebalanced to:
32% SPY , 44% TLT , 24% GLD
This may end up being my last post reflecting my actual traded portfolio. I will likely start using leverage again in the accounts where it’s allowed. Also I’m going to start implementing some small changes to my traded strategy. The key adjustment will add an ensemble type approach with multiple inputs as opposed to singular inputs for returns, standard deviation and correlation (I tested most of them years ago, but didn’t understand the value in combining them until recently).
But no fear, I’m leaving the same strategy running at the top of the blog, and will still post it’s positions at the end of each week as always. The only difference is, it won’t necessarily be an update from my trading account.
A portfolio of 4 equal parts of SPY, CASH, TLT and GLD would return 9.05% this year with the max drawdown of -1.36% (according to portfoliovisualizer). I wonder how much edge are you getting from the geometric return optimization and how much just from the allocation? Thanks for the update.
I think your the drawdown number is too low. At a minimum that’s monthly data. On a daily timeframe, I believe the drawdown would be closer to 10-11%.
And yes, in certain types are markets, the strategy will perform similar and possibly slightly worse to the PP because it converts itself into a portfolio similar to the PP. So this year you are correct. In other years and markets (like last year), the opposite will be true.
Can you post the 3 assets plus cash Google Sheet. I’ve extended the 2 assets plus cash but would like to verify its accuracy before using. Also, can you let me know how you are deciding on leverage factor.
There’s still a lot more to discuss before getting to that point. I’m not going to bring up leverage until there is a lot more discussion on risk and error management.
When you say that you’re going to potentially start using an ensemble approach for modelling returns, std deviation, and correlation, do you mean you’re considering taking a more predictive approach on how to model these inputs? Or are you referring to something else like adding multiple equity/bond components to the portfolio?
Also, I’m curious have you looked into potentially automating the rebalancing? Seems like you could get rid of a lot of the stress of taking time away from work to rebalance when correlations are in flux.
I already take a more predictive approach to modeling these inputs than what I’ve discussed so far in the blog. There’s more than one good way to estimate these future inputs, and in stead of picking what I think is the best one, or my first estimate, I’m going to use all them to create multiple portfolios and then average those portfolios.
I have looked into automating it. One day I will. But its not really that hard to do it manually every Friday right now. It only takes a few minutes.
When you say that you’re going to potentially start using an ensemble approach for modelling returns, std deviation, and correlation, do you mean you’re considering taking a more predictive approach on how to model these inputs? Or are you referring to something else like adding multiple equity/bond components to the portfolio?
Also, I’m curious have you looked into potentially automating the rebalancing? Seems like you could get rid of a lot of the stress of taking time away from work to rebalance when correlations are in flux.