Just curious if anyone here implements this type of strategy:
http://epchan.blogspot.com/2007/01/universal-portfolios.html
http://epchan.blogspot.com/2007/01/universal-portfolios.html
Universal Portfolio
- Thread starter digitalnomad
- Start date
Great article.
2019 *should* end with me having rules for a dynamic(-ally optimised), trend-exploiting, mean-reversion-backed ("universal") portfolio. We'll see. EOD data, EOD trades, so it shouldn't be that hard to *code*.... the *optimization* could much longer.
But it's certainly where I'm going. The options end is done; the trend-exploitation end is (very!) well underway. Still, the Developer (me), tells me that one calendar year is pretty damn short......
Last edited:
Sounds interesting, but from the article:
"if you factor in a transaction cost of 10 basis points, the Universal Portfolio scheme actually returns only 4%. This is the second caveat of Universal Portfolios: because of the frequent rebalancing required, transaction costs tend to eat up all the out-performance."
You do get large volume discounts from the IB tiered commission structure, but probably still not enough to warrant a viable retail strategy. Perhaps weekly rebalancing might beat the benchmark.
How bout running tests through recessions? In bull market, so long as you making much more than brokerage, more taxes on it as well? I think @digitalnomad has better idea.
Or if you run a billion dollars fund, the transaction cost should be small enough to make this a winning strategy?
This is powerful, check out 'no regret trading' where you move at some learning rate towards the best performing allocation. There are a few very well known techniques for simple equity strategies. 1) you want to adjust your holdings in a long gamma way on short time horizons (take money from risers and give to the fallers). 2) the weight of total usd equity you give each stock should reflect its performance over medium term. 3) you need a cut off where the real losers are replaced with other stocks, this avoids the type of 'survivorship bias' you get in the spx, which unfortunately is a big deal, like in SPY, most stocks weren't there 10 years ago and at least 6% of the annual return can be attributed to this changing of the constituents. So you want to combine all these. Idea 1) take the current 3 month moving average, compare to the 3 month moving average 2 months ago. Divide the two and use this return as your weight (so you get like 1.1,0.91, etc etc as weights) 2)Intraday as the stock moves around the weights change much more slowly than the portfolio holdings so you can rebalance very frequently. 3) when r<0.8 you replace with another symbol and re-weight the whole thing. 4) Additionally against this sell a 2 month call on SPX struck 1-2% OTM in about 80%+ of the total notional, which you can roll at expiry and restrike on corrections lower through the strike before expiry. Def 25%+ this year. Can improve the options hedges by selling topside puts thus reducing the short delta hedge and many other options here. Like?
Additional note. Use forwards/futures to buy initially then for your hyper high frequency@zero latency gamma trading of the weights use the tighter and more liquid equities. helps funding and they should mostly trade well below the spot price. Often significant (see for example DJIA 3 month forwards)
Sorry the call you sell should be 'in the money' depending on how low volatility you want the strategy to have. Definitely worth looking into the options part though like different expiries, nationals and rules for how you role it. V possible to reduce volatility and increase expected return versus not including it.
I have done long only monthly rebalancing in an IRA for a friend, implementing a combination of volatility based sizing and industry rotation yielding him -2% for the year which I find pretty dull. Looking at the big picture though, the benchmark ended 5x worse, and only large caps were held in the portfolio. He thinks I’m a savant.