We designed our public dual momentum model, GEM, to be simple and easy for do-it-yourself investors to use. GEM helps protect smaller investors from significant drawdowns while allowing them to earn returns that exceed the market over the long run.
But momentum works best when it incorporates multiple trend determinants. That creates a synergistic effect.
Most investors do not give enough importance to price trends. Greyserman & Kaminski show that simple trend-following has outperformed buy-and-hold and reduced downside excursions back to the origins of many markets. No other investment factor has shown that.
Our models are the culmination of a lifetime of investment research and experience. Most of our proprietary models use a channel breakout approach validated on 100 years of data, as seen in our award-winning research study. We also insist on superior real-world performance.
Richard Dennis taught something similar to this to his “turtle traders.” Jack Dreyfus had the best-performing mutual fund from the mid-1950s to the mid-1960s, trading channel breakouts in stocks making new highs. Nicolas Darvas wrote about making millions with this approach.
Most of our models incorporate reversals from overbought and oversold conditions to complement trend following. Our proprietary models use daily data and adapt quickly to market conditions. They are based on rigorous, academic-quality research, validated against at least 100 years of data. Long historical validation helps reduce the risk of overfitting.
We spend considerable time on portfolio construction and on due diligence to find the best ETFs for our models. Thoughtful portfolio structuring is an essential part of our investing process.
Our models work in many markets. We choose combinations that create optimal, balanced portfolios responsive to different market conditions. Using multiple models with low-to-moderate correlations is the best way to reduce model estimation error and uncertainty. Optimal model combinations may enhance expected returns and reduce expected downside exposure.
We license our proprietary model signals to substantial private and institutional investors, as well as select investment advisors who understand and appreciate our work.
Most trading models exit to a safe-haven asset when not in risk-on positions. Our models are unique in that they switch to other assets or models with positive trends before seeking a safe harbor. This layering can capture additional profits and help reduce whipsaw losses. Our models also use confirming model signals from other closely aligned assets. Here are our current proprietary models.
SMURF focuses on U.S. large-cap stock ETFs and may take modest positions in other stock-market or non-stock ETFs when their trends are positive. SMURF exits to short-term fixed income when trends are no longer positive.
BADASS applies its trend model to ETFs that track blockchain and digital technology stocks. It also allocates a moderate amount to spot Ethereum and Bitcoin ETFs when their trends are positive. BADASS has been our most profitable model.
GLTR applies trend following to gold ETFs. Gold is often mean-reverting and challenging to trade, but our trend strategy handles it well. GLTR also incorporates mean reversion trading. Even without trend-following and mean-reversion profits, gold has outperformed the S&P 500 over the past 25 years. GLTR often has a low correlation to our other models and is a good portfolio diversifier.
FIRM is anchored by short-term fixed-income ETFs. It can also hold short-term positions in other ETFs to exploit short-term reversals. FIRM has the lowest correlation among all our models. FIRM also provides tax deferral on most of its investment income.
GLTR spends about half its time in gold-stock ETFs and the rest in the SMURF positions. SMURF spends around 60% of its time in the stock market and 40% in the FIRM positions. 50/25/25 is a balanced allocation: 50% SMURF, 25% GLTR, and 25% FIRM.
S&P500 | SMURF | GLTR | FIRM | 50/25/25 | |
CAGR | 10.8 | 20.1 | 36.0 | 4.5 | 20.0 |
STD DEV | 16.6 | 11.6 | 15.3 | 2.7 | 8.8 |
ADJ SHARPE | 0.60 | 1.50 | 2.07 | 1.10 | 1.95 |
UPI | 1.05 | 8.58 | 16.26 | 21.63 | 15.02 |
MAX DD | -52.9 | -9.6 | -10.7 | -1.6 | -5.9 |
Results do not guarantee future success nor represent returns that any investor attained. All trading involves risks that may not be foreseen. Results reflect total returns, including reinvestment of interest and dividends but not transaction costs. Positions are rebalanced monthly. Returns and drawdowns are calculated on a month-end basis; intra-month drawdowns would have been larger. CAGR is the compound annual growth rate. UPI is the Ulcer Performance Index, which divides return by the Ulcer Index, which measures the depth and duration of drawdowns from earlier highs. ADJ SHARPE adjusts the annual Sharpe ratio for skewness and kurtosis, per Pezier & White (2006). Our models have evolved. This is their current representation. Model details, including holdings and layering rules, are described in the individual model fact sheets. See our Disclaimer page for more information.
Because of our models’ risk controls, we can use assets more aggressively than buy-and-hold investors. We can further reduce portfolio volatility by combining models with modest correlations and by incorporating low-volatility assets.
Here are some high reward-to-risk initial portfolio allocations using SMURF, BADASS, FIRM, and GLTR in that order. SMURF can end up with allocations up to 80% because of model layering.
SMURF | BADASS | FIRM | GLTR | 20/25/30/25 | 25/20/35/20 | 25/15/40/20 | |
CAGR | 25.1 | 99.0 | 6.5 | 39.5 | 39.4 | 34.2 | 29.9 |
STD DEV | 13.4 | 36.2 | 3.7 | 17.0 | 13.5 | 11.8 | 10.2 |
ADJSHARPE | 1.57 | 2.25 | 1.21 | 1.98 | 2.51 | 2.50 | 2.49 |
UPI | 9.66 | 30.56 | 137.1 | 16.88 | 28.47 | 27.96 | 27.55 |
MAX DD | -8.3 | -17.1 | -0.3 | -10.7 | -5.9 | -5.0 | -4.4 |
AVG DD | -1.4 | -1.7 | 0.0 | -1.1 | -0.6 | -0.5 | -0.4 |
W%MOS | 72 | 83 | 95 | 75 | 79 | 90 | 80 |
Results do not guarantee future success nor represent returns that any investor attained. All trading involves risks that may not be foreseen. Results reflect total returns, including reinvestment of interest and dividends but not transaction costs. Positions are rebalanced monthly. Returns and drawdowns are calculated on a month-end basis; intra-month drawdowns would have been larger. CAGR is the compound annual growth rate. UPI is the Ulcer Performance Index, which divides return by the Ulcer Index, which measures the depth and duration of drawdowns from earlier highs. ADJ SHARPE adjusts the annual Sharpe ratio for skewness and kurtosis, per Pezier & White (2006). BADASS crypto data has been winsorized to the 95th percentile to reduce the influence of extreme outliers. Our models have evolved. This is their current representation. Model details, including holdings and layering rules, are described in the individual model fact sheets. See our Disclaimer page for more information.