Put/Call Ratios and Market Predictions

Some feel that Put/Call Ratios are a good predictor of market movements since ...

>Put/Call ratios?
Yes. If you buy a Put then you can, at any time before it expires, decide to sell a stock at some predetermined price
      namely the Strike (or Exercise) price of the Put option.
If you buy a Call then you can, at any time before it expires, decide to buy a stock at some predetermined price
      namely the Strike price of the Call option.

Generally, buying options allows you to profit from rising markets (in the case of Call options) and falling markets (in the case of Put options).
That means when there are lots of Put options (compared to Calls), then investors expect the market to fall.

>And when there's lots of activity in Calls, those investors expect the market to rise, right?
Right. So the Ratio of Put to Calls gives some indication of investor sentiment. Do they expect the market ot rise of fall?

>But if the Put/Call ratio is large, it means the sentiment is that the market will fall, right?
Yes, you'd think so ... for those who trade in options.

>So when the Put/Call ratio is large, you sell ... right?
You'd think so, however, option investors are wrong 90% of the time
... so maybe one should do exactly the opposite to what the Put/Call Ratio suggests.    

>You buy when the Put/Call Ratio is small and ...
And buy when it's large. Yes, that's the theory.

>Large or small compared to what?
Good question. We have to look at the historical data.
Figure 1 shows the Put/Call Ratio and the S&P 500 over about 8 1/2 years.
It would have been a good strategy (over this time period!) to:
      Sell when Put/Call Ratio fell below 0.50 and
      Buy when the Ratio rose above 0.90.

I should point out that, in Figure 1, we actually calculated a 50-day Exponential Moving Average of the Put/Call Ratio (see EMA). ... to smooth out violent oscillations.
>Is 50-day the best choice?
I have no idea, but here's a chart using a 10-day EMA:




Figure 1
>I'll take 50!
Yes.

>How about finding find "the best", using various historical data, other time periods or maybe other markets or ... ?
Yes, but my problem is getting historical data on those ratios
>Excuses, excuses ...