I noticed that all the backtesting shown on this site looks at the last 10 or 20 years. The market performance during this time is a little unusual in ways that would seem to bias results in favor of a DM strategy as articulated by MDMD (6-month lookback, relative strength, 1-fund holding, etc). For example, two huge crashes within 10 years of each other, both were preceeded by a bull market, and both crashes were slow enough and deep enough that the market timer could switch to another fund well before the bottom, and the bull markets following the crashes were slow enough at first such that the market timer could jump back in before the losing too much growth, etc. I thought I would see what the most data available was and simulate that.
I think I discovered where there isn't any free lunch. And that the flaws can be substantively deleterious to short- and long-term performance. I used the Vanguard funds with the longest-available data on
https://www.portfoliovisualizer.com/test-market-timing-model to conduct MDMD's DM strategy as best as possible from 1986 (or 1987 or 1988 for some funds) through 2014. The 500 Index fund (VFINX), the Small Cap Index Fund (NAESX), and International Growth Fund (VWIGX) were available from 1985 onwards. I was unable to find another international fund available for as lengthy a time period. I also looked at results with that fund excluded. The bond funds most similar to "quality" that investors normally pile into during crashes and with data available were GNMA (VFIIX), Long-Term Treasury (VUSTX), Total Bond (VBMFX), LT Investment Grade (VWESX), ST Investment Grade (VFSTX). These were available starting between 1986 and 1988 depending on the fund. You can argue about which one you think is the "right" one that you would use in a DM strategy. But I looked at them all. I won't include links to all of the simulation results because they are absurdly long URLs. But I will include one and you can easily substitute the fund call letters to see the results for yourself. Note that all these examples are ignoring any transaction costs (through bid/ask spreads, trading fees, capital gains taxes, etc) and assume reinvestment of dividends.
Example 1 LT Treasury 1987+:
Timing portfolio $90,960
Stock market $159,503
Over ~30 years, you end up with about half as much as a buy-and-hold investor, and lag the bond portfolio for about 25 of the years. There's a +$20k positive difference by omitting international.
Example 2 GNMA 1986+:
Timing portfolio $191,625
Stock market $185,465
There's a -$35k difference by omitting international.
Example 3 Total Bond 1988+:
Timing portfolio $177,762
Stock market $156,984
Even with avoiding the 1987 flash crash, the timing portfolio only slightly outperforms. Omitting international has a -$12k difference.
Example 4 LT investment grade 1986+
Timing portfolio $127,063
Stock market $185,465
Omitting international has a -$23k difference.
Example 5 ST investment grade 1986+
Timing portfolio $171,083
Stock market $185,465
Omitting international has a -$40k difference.
So this tells me that the DM strategy, when tested on an out-of-sample set of data, fails miserably. The results posted earlier in the thread from the last 10 or 20 years showed ending portfolios of ~4X the buy-and-hold. So the fact that extending the sample backwards just 10 or 20 years results in similar or much worse performance shows how terrible the DM strategy performs during that extra period--even in the period that excludes the 1987 crash.
In this sequence of returns, for some of these examples, a retiree would go bust quickly. Since their portfolio dives in the first portion of the simulation period and only very slowly recovers 2 decades later, a retiree would be selling a very high percentage of their portfolio to live on. In example 1, the portfolio without any withdrawals would still be below the starting value after 6 years.
Example simulation URL (
note you can't just click on the link here to have all the options load, you have to copy and past the URL Mod Edit: Fixed link. Click away. /end edit.):
https://www.portfoliovisualizer.com/test-market-timing-model?s=y&timingWeights[2]=0&endYear=2014&volatilityPeriodWeight=0&movingAverageType=1&windowSize=6&timingUnits[2]=2&timingModel=4&volatilityWindowSize=0&startYear=1985&assetsToHold=1&multipleTimingPeriods=false&timingUnits[1]=2&outOfMarketAssetType=1&timingPeriods[0]=5&timingWeights[0]=100&volatilityWindowSizeInDays=0&riskControl=false&symbols=VUSTX%2C+VFINX%2C+NAESX%2C+VWIGX&riskWindowSizeInDays=0&timingUnits[0]=2&timingWeights[1]=0&windowSizeInDays=105&volatilityPeriodUnit=1&riskWindowSize=10&rebalancePeriod=1