| Back to Enhanced Slow-Burning Model | |||||||||
| HAUGEN SYSTEMS SLOW-BURNING MODEL BACKTEST | |||||||||
| In this test we constructed portfolios based on a population of the largest 1,000 stocks. Fifty stocks with the | |||||||||
| highest expected return were added at the beginning, four stocks were rebalanced each month, and the | |||||||||
| monthly returns were linked together to compute an annualized return. At beginning of each year, we also | |||||||||
| made sure that the portfolio's sector weights were within 200% of the S&P 500's sector weighting. The | |||||||||
| turnover in these portfolios approximates 100% per year. Then 11 more tests were run, each starting on a | |||||||||
| different month in 1996, but otherwise employing the same rules. Finally, the 12 test's annualized returns | |||||||||
| were averaged and shown below. | |||||||||
| So the first column shows average portfolio returns for the 12 tests done on the standard model. Then, we | |||||||||
| ran a backtest utilizing proprietary improvements to the model developed over the last three years, producing | |||||||||
| new alphas. We then constructed portfolios using the same methodology as with the standard model, | |||||||||
| buy 50 stocks with the highest expected return, rebalance 4 a month, and link the returns. The results | |||||||||
| are presented in the 2nd column. The annual returns for the S&P is included in the right column for comparison | |||||||||
| purposes. | |||||||||
| Standard | Enhanced, | S&P | |||||||
| Model | Slow-Burning | 500 | |||||||
| Model | Index | ||||||||
| Total Return | Excess Return | Total Return | Excess Return | Total Return | |||||
| 1996 | 39.85% | 16.88% | 45.92% | 22.96% | 22.96% | ||||
| 1997 | 28.85% | -4.52% | 38.72% | 5.35% | 33.36% | ||||
| 1998 | 25.34% | -3.23% | 36.36% | 7.78% | 28.58% | ||||
| 1999 | 39.83% | 18.79% | 48.13% | 27.09% | 21.04% | ||||
| 2000 | 1.27% | 10.37% | -4.81% | 4.29% | -9.11% | ||||
| 2001 | -1.54% | 10.34% | 1.80% | 13.68% | -11.89% | ||||
| 2002 | 5.42% | 27.52% | 3.79% | 25.89% | -22.10% | ||||
| 2003 | 30.29% | 1.61% | 31.39% | 2.70% | 28.68% | ||||
| 2004 | 23.87% | 12.99% | 25.73% | 14.85% | 10.88% | ||||
| 2005* | 32.73% | 27.82% | 36.72% | 31.81% | 4.91% | ||||
| Average Linked Annual Ret. | 21.67% | 11.33% | 24.96% | 15.20% | 9.07% | ||||
| Average Annual Return | 22.59% | 11.86% | 26.37% | 15.64% | 10.73% | ||||
| Longitudinal Std.Dev. | 15.41% | 11.45% | 19.23% | 10.64% | 19.51% | ||||
| Information Ratio | N/A | 0.99 | N/A | 1.43 | N/A | ||||
| T-Stat for the Mean Return | 4.40 | 2.97 | 4.12 | 4.29 | 1.65 | ||||
| Probability Total (Excess) Ret < 0 | 7.70% | 16.91% | 9.32% | 8.33% | 28.43% | ||||
| Each test's results were computed as follows | |||||||||
| 1. First the test that began in January had the monthly returns linked together to get an annual total return for 1996. Linking is the | |||||||||
| geometric average calculated by adding 1 to each monthly return, multiplying the numbers together, and subtracting 1. | |||||||||
| 2. Then 1997 thru 2005 had the total returns calculated in a similar way. | |||||||||
| 3. Then the February portfolio's annual returns were calculated the same way, then March, and so on resulting in a table | |||||||||
| with 6 by 12 annual returns | |||||||||
| 4. Then 1996's average annual return was calculated by taking the arithmetic average of all 12 portfolio's 1996 results | |||||||||
| 5. The rest of the year's average annual returns were then calculated. | |||||||||
| The Average Linked Annual Return was then computed by linking these average annual returns together. Like with the monthly | |||||||||
| returns, above, we added 1 to each annual return, and multiplied the numbers together. Then, that product was taken to the | |||||||||
| 12 / 120th power before subtracting by one. The numbers from Greenblatt's books were already geometric averages, so the | |||||||||
| average linked annual return could be calculated the same way as in the rest of the columns. | |||||||||
| The Average Annual Return is the arithmetic average of each test's (eg Jan96, Jan97, Feb96, etc) monthly total return. | |||||||||
| The longitudinal standard deviation is calculated by taking a standard, non-biased standard deviation of the average annual returns | |||||||||
| The information ratio is the average excess return divided by the standard deviation of the excess returns | |||||||||
| The t-stat is the average annual return divided by the standard deviation divided by the square root of the number of years in the study. | |||||||||
| The probability is calculated by first finding the range of returns in the bell curve (eg, 95% of the returns will be within 2 standard | |||||||||
| deviations) and then calculating the ratio of this range and dividing by 2. | |||||||||
| Backtest Parameters: | |||||||||
| Region | U.S. | ||||||||
| Population | Top 1,000 | ||||||||
| # of stocks | 50 | ||||||||
| First Date | 1996 | ||||||||
| Last Date | 12/31/05 | ||||||||
| Sector Constrain | Yes | ||||||||
| Rebalance periodicity | Monthly | ||||||||
| Number to rebalance | 4 | ||||||||