I spend a lot of time reading research papers on areas of tactical asset allocation (momentum, trend following, counter trend following, etc). Lately I have read a few papers on avoiding momentum crashes. Since our firm uses a lot of momentum strategies, this is a subject near and dear to my heart. Momentum crashes are primarily a phenomenon related to individual stocks, which we don't really use, but the research is interesting nonetheless as it confirms my own findings about combining momentum with trend following.
Momentum strategies with individual stocks usually involve either buying high flying stocks, or a combination of going long high flying stocks and shorting weak stocks. Both strategies have shown remarkable performance over the years as momentum is a powerful force in every market. However, these strategies also have a propensity to crash every once in a while. High flying stocks can get way beyond where they should trade and can, and do, have spectacular falls. Long/short investors have also seen times when the dogs did well and the high fliers reverse. We use momentum with index ETFs and funds and we don't really go short anything. While indexes do get frothy, they don't do so to the same extent that individual stocks do. However, momentum investors in indexes can still suffer drawdowns by buying the strongest index in a bear market---in the land of the blind the one eyed man can still lose money.
The fix to this is to combine momentum with trend following. The terms are often used interchangeably but they are different concepts. Momentum involves buying the strongest asset classes, sectors, etc, regardless of direction. Trend following involves only buying assets that are in an uptrend. By combining the two and buying the strongest areas of the market only when they are in an uptrend you can increase the returns and decrease the risk of a plain momentum strategy (not to mention blow away buy and hold).
As an example we have a momentum model we use to trade the S&P 500. It can only be in the S&P 500 when it is in an uptrend. Backtested results of this system from 9/29/03 to 9/14/2012 are as follows:
Average Annual Return: 10.35%
Maximum Drawdown: -8.67%
Worst Month: -4.02%
Mar Ratio: 1.19
Sharpe Ratio: .82
If I take the same model but remove the trend following component, so that the S&P 500 doesn't need to be in an uptrend to buy it, we get the following results:
Average Annual Return: 9.14%
Maximum Drawdown: -14.24%
Worst Month: -11.82%
Mar Ratio: .64
Sharpe Ratio: .60
While this model clearly outperforms buying and holding the S&P 500 it markedly under performs the model that uses a trend following filter.