A consistently interesting site for traders and investors is the CXO Advisory.
Bollinger Bands is the topic of one of their recent blogs. They test the bollinger Bands as a method of making overbought/oversold trading decisions on the S&P500 Index. I agree with their conclusions for the most part but caution against drawing such a broad conclusion from such a restrictive test. For example returns of a continuous buy and hold on the index are used as the benchmark for comparison to the results of using a bollinger band signal for trades of 21 days duration. CXO rightly points out that the gains and losses of the bollinger strategy are all short term and thus taxed more heavily than buy and hold and also that no interest earned component is considered for those times the bollinger strategy is out of the market. These are not small items. The interest earned on idle funds with such a strategy would be significant since the strategy is out of the market almost half the time over 57 years. Also the requirement of 21 days trade length has the very limiting effect of cutting the very best trades short. The arbitrariness of that exit strategy heavily dilutes the value of the conclusion. Furtthermore it is not possible to buy and hold the index over such a long period. Index components change requiring trades and incurring costs and taxes all of which reduce the buy and hold return by some uncalculated amount. So while I agree with CXO's conclusion as stated for the test done I find the bollinger band results somewhat encouraging of further study.
I want to reiterate none of this is to criticize CXO Advisory who generously provide interesting and useful research at no cost to readers like me. I merely want to raise points regarding how results should be evaluated and utilized.