In a new study, Smithers & Co. examines just what valuation metrics have historically had predictive power over the stock market’s direction — and the results don’t bode well for the current market.

The study, co-authored by Andrew Smithers — who warned of both the 2000 and 2008 market crashes (although several years early in the case of the latter) –and finance professor Stephen Wright, found that the metrics currently being used by many bulls actually have had little accuracy in predicting market moves historically, Brett Arends of  MarketWatch reports. For example, the fact that stocks yield more than bonds or other investments on a relative basis has no significance in terms of predicting future returns, Smithers and Wright found. Buying stocks that are “cheap” by historic standards when compared with last year’s or next year’s per-share earnings, and avoiding stocks when they are expensive using those gauges, works better, but is still a very flawed approach, they found — such an approach has given some “disastrous” buy and sell signals over the years, Arends says.

Which metrics do work? Smithers and Wright found that the 10-year cyclically adjusted price/earnings ratio and Tobin’s Q do. So does a complex proprietary measure Smithers helped develop that “is based on the stock market’s previous returns — and basically says that long periods of excellent returns are generally followed by long periods of dismal ones,” Arends says. “None of these three is perfect,” Arends notes. “But all three have proven very good indicators of subsequent returns for over a century.” He adds that their predictive power is over very lengthy periods of time, saying that history has shown that periods of even ten years can be “quite random” for stock returns.

Unfortunately, right now all three show that the market is significantly overvalued and poised to produce weak returns over the long haul. “In other words, today stocks look like good values only according to measures which have absolutely no track record of predicting future returns,” Arends says. “And according to the measures which do have such a track record, they look wildly overvalued. Make of it what you will.”