The quick answer is when no one else is buying them.
Suppose we measure the number of NYSE, NASDAQ, and Amex stocks making fresh 65-day highs during each trading day. Since 2004 (N = 675 days), when fewer than 150 stocks across all the exchanges are making 65-day highs (N = 69), the next 20 days in the S&P 500 Index (SPY) average a gain of 1.77% (51 up, 18 down). That is a considerably more favorable return than the average 20-day gain of .49% (415 up, 260 down) for the entire sample.
How about when everyone is buying, however? On Thursday, we had 1157 stocks across the exchanges make fresh 65-day highs. Since 2004, when we've had 1000 or more 65-day highs (N = 61), the next five days in SPY have averaged a loss of -.03% (30 up, 31 down). That's weaker than the average five-day gain of .12% (372 up, 303 down) for the sample overall. By 20 days out, however, SPY was up on average by .56% (43 up, 18 down)--a very respectable win:loss ratio.
In short, when no one is buying turns out to be a great time to look at stocks, but returns are respectable when everyone is buying. We typically see some pullback in the near term after a period of popularity, but over a 20-day period returns are quite good. The notion that "overbought" markets are due for scary corrections does not hold water. Rather, it seems as though upside momentum tends to persist over an intermediate-term time frame.