It’s time to revisit the core tenets of our strategy.

This is the definitive source paper for our strategy. I also noticed the best rotation indicator again is the dollar index, and the best EM returns comes in the first 5 years of a Dollar decline.


While only using developed markets indexes from 1974-1993, it was able to see that:

  1.  Buy and hold equal weighted global portfolio.=14.2% annually (1 above)
  2. The parametric contrarian strategy has a shocking great return of 20.7% (2 above)
  3. The blind momentum strategy of buying a portfolio of 3-5 year biggest winners also did poorly and was the worst option at 9.3% (3 above)
  4. The blind contrarian strategy of buying a portfolio of 3-5 year biggest losers did not outperform buy-and -hold the equal weighted global portfolio but  nearly matched it at 13.5%. and handily beat the momentum strategy (4 above)
  5.  The really great news for the sharp ratio for the parametric contrarian strategy is also approximated the same as Buy and hold equal weighted global portfolio.
  6. Using Emerging markets the returns will be higher with somewhat more risk.

This source paper was done more recently and also supports our strategy from 1987-2008 using 70 country indexes. (Our ETF universe has only about 40 to choose from.)



Their Momentum strategy achieved the 16.4% return and it may work for a fast trading mindset of buying equal weighting portfolio that has 6 country indexes of the universe showing best 1 year performance and holding for 1 month only, however this doesn’t account for trading cost and taxes. If those two factors were included in taxable accounts, it could drop effective return significantly by 2-4% annually depending on tax bracket, not to mention trading errors from bid-ask spreads.

Personally this theoretical backtest has a too-good-to-be-true feel about it. Since everyone intuitively chase winners and so momentum strategy is always in crowded trades. So it’s better avoided.

This recent Paper attempts to argue the reasons why a real world portfolio can actually capture the momentum profits in real time.


it shows that trying capture momentum doesn’t work well in real time. trading commissions, taxes, always ruin the results. but personal experience has been that this kind of mind set is too tumultuous and can easily lead to huge mistakes in analysis from market hype and excitement over whatever is hot at the moment.

Notice how small a profit margin the momentum trade can theoretically get? I am not convinced from this paper’s rationals.

This past 7 years to 2016 also proves that momentum’s best days are probably over. Nothing beats holding index funds for years at a time.

charts of this mutual fund vs the RPG:


It could not beat the pure growth index fund. Enough said.

Back to our contrarian strategy:

Their “next 18” contrarian strategy backtest achieved 20.7%  from buying equal weighting portfolio that has 4 country indexes of the universe showing second worst 4-5 year  performance, and holding then for 2 years, and rebalance annually. and this does account for trading cost and taxes, good for collecting dividends, sleeping easily, and simple tax reporting.

Key things to recap here are:

  1. Minimum of 4 and maximum of 6 uncorrelated funds are needed for diversification
  2. Avoid the biggest loser index for mean reversion & choose 2nd, third, forth, and 5th worst.
  3. Avoid the best 1 year index for momentum & choose the smaller ones,
  4. The parametric strategy looks deeper into each country and try to estimate company/industry growth/recovery potential, population change, governance, currency control, rule of law, etc…
  5. Avoid countries that are not investable and favor those with decent growth potential, no matter how badly or well the index is performing currently.
  6. Period dependency is clear for investment results, if momentum/growth stocks is currently favored then contrarian strategy will lag and vice-versa.

Coincidentally the two source papers both calculated at best return of 20.7% for contrarian/mean reversion strategy despite using different time periods and sampling size.