Author: mungofitch

Subject: Re: ETF long term strategy

Date: 4/28/2016

Recommendations: 12

I`d be tempted to test something like this - Backfill every ETF with the historical data of the index it tracks. Skip any ETF you can`t get a whole lot of history for that way. A 20 year minimum sounds good. I got free country-level daily total return data from S&P back to 1989, for example. - Skip funds which track small countries or industries with absurdly concentrated holdings. If the UK is looking nice that`s OK, but Finland was just Nokia for a long time. Anything super concentrated probably has no predictive power for itself. - Starting five years into your data set, do a step forward test. Each year, buy, say, the 6 funds with the lowest rolling-year downside deviation with MAR=10%. Or maybe better, lowest rolling-two-year DD with MAR=10%/year (21% per 2 year period). Hold those for a year, repeat, but this time examine all the history to date for each fund: one year longer each time. Downside deviation is usually thought of as a risk metric, but of course you can`t achieve zero risk on that metric unless you`re getting 10% in every single rolling year, so things with higher returns have a serious tail wind. The general philosophy here: (1) take care of the risk, and the returns will take care of themselves. At least with that metric. (2) since you`re including sectors and perhaps industries, to a certain extent you`ll end up drifting towards those that have the best average long run returns, while avoiding chasing bubbles as might happen with a pure CAGR metric. You don`t want to get put into internet stocks in 2000-2001. The goal is to get put into (say) medical devices and gaming and makeup, not transport and toys and thrifts. (3) I see no need to seek low correlation. I want them all to have very similar high and steady returns! (4) If it ain`t 20 years, it ain`t a backtest worth reading. Jim