P/E alone would not work very well. See for example http://www.hussmanfunds.com/html/peak2pk.htm and http://www.hussmanfunds.com/rsi/profitmargins.htm (in short, P/E is affected too much by cyclical changes in profit margins, or you might say: booms inflate the E beyond sustainable levels, thus making the P/E look more favorable than it is).
Here's a random blog post that points to Schiller's normalized earnings measure: http://seekingalpha.com/article/247257-s-p-500-is-expensive-using-normalized-earnings
I think even Price to Sales is supposed to work better than P/E for predicting 10-year returns on a broad index, because it effectively normalizes the margins. (Normalized valuation explains the variance in 10-year returns better than the variance in 1-year returns, I think I've read; you can't rely on things "reverting to mean" in only 1 year.)
Another issue with P/E is that E is more subject to weird accounting effects than for example revenues. For example whether stock compensation is expensed or one-time write-offs are included or whatever can mean you end up with an economically strange earnings number.
btw, a simple way to do what you describe here would be to put a chunk of money into funds that vary equity exposure.
For example John Hussman's fund has an elaborate model that he uses to decide when to hedge. Say you invest 40% bonds, 40% stocks, and 20% in Hussman Strategic Growth. When Hussman fully hedges his fund, you would effectively have 40% in stocks; and when he fully unhedges it, you would have 60% in stocks. This isn't quite the whole story; he also tries to pick up some gains through stock picking, so when fully hedged the fund isn't quite equivalent to cash, more like a market-neutral fund.
(For Hussman Funds in particular, he's considered stocks to be overvalued for most of the last 15 years, and the fund is almost always fully hedged, so you'd want to be comfortable with that.)
There are other funds out there doing similar stuff. There are certainly funds that vary equity exposure though most not as dramatically as the Hussman fund. Some possibilities might be PIMCO All-Asset All-Authority, PIMCO Multi-Asset, perhaps. Or just some value-oriented funds with willingness to deviate from benchmarks.
Definitely read the prospectus on all these and research other options, I just thought it would be helpful to mention a couple of specific examples.
If you wanted to stick to managing ETFs yourself, Morningstar's premium service has an interesting feature where they take the by-hand bottom-up analysis of all the stocks in an ETF, and use that to calculate an over- or under-valuation ratio for the ETF.
I don't know if the Morningstar bottom-up stuff necessarily works; I'm sure they make the "pro" case on their site. On the "con" side, in the financial crisis bubble bursting, they cut their valuation on many companies and they had a high valuation on a lot of the financials that blew up. While I haven't run any stats and don't have the data, in several specific cases it looked like their bottom-up analysis ended up assuming too-high profit margins would continue. Broad-brush normalized valuation measures avoided that mistake by ignoring the details of all the individual companies and assuming the whole index had to revert to mean.
If you're rich, I think you can hire GMO to do a varied-equity-exposure strategy for you (http://www.gmo.com/America/).
You could also look at the "fundamental indexing" ETFs that weight by dividends or P/E or other measures of value, rather than by market cap.
The bottom line is, there are lots of ways to do tactical asset allocation. It seems complex enough that I'm not sure it's something you'd want to manage yourself.
There are also a lot of managers doing this that I personally am not comfortable with because they don't seem to have a discipline or method that they explain well enough, or they don't seem to do enough backtesting and math, or they rely on macroeconomic forecasts that probably aren't reliable, or whatever.
All of these tactical allocation strategies are flavors of active management. I'm most comfortable with active management when it has a fairly objective, testable, and logical discipline to it, such as Graham&Buffett style value investing, Hussman's statistical methods, or whatever it is.
Many people will argue that all active management is bad and there's no way to distinguish among any of it. I am not in that camp, but I do think a lot of active managers are bad, and that it's pretty hard to distinguish among them, and I think active management is more likely to help with risk control than it is to help with beating the market.
Still you should know (and probably already do know, but I'll note for other readers) that there's a strong argument smart people make that you're best off avoiding this whole line of tactical-allocation thinking and just sticking to the pure cap-based index funds.