The statistical edge of trading against retail sentiment

I spent today analysing historical position data off OANDA. I wanted to know what sort of edge I can expect when trading against retail sentiment. If retail sentiment was extremely bullish, what sort of edge could I expect by selling short, and vice versa?

I analysed four currency pairs: the AUDUSD, AUDJPY, USDCAD and EURUSD. OANDA's data only went back one year, so that is the time scope I analysed. I then checked for days where retail sentiment was at least 65% bullish or bearish. Basically I was looking for extreme bullish or bearish sentiment. I then checked how much price moved during each of those days (from the close of the previous day to the close of the current day), and created a total. 

These are the results:

For each currency pair, if you traded with extreme retail sentiment (65%+ bullish or bearish), you would've lost pips (hence the negative value). If you traded against retail sentiment, you would've made pips. This analysis seems to confirm the idea of betting against the crowd.