I have backtested some of this stuff across hundreds of stocks and ETFs going back to 2008 (I could have gone back farther, but limited the scope to after 2007). I looked at forward returns over periods of 5, 10, 15, 20, 25, 30, 35, 40, and 45 days. Submitted each set of returns to a t-test against randomly sampled entries. Looked at long entries only. End result, only buying at the lower keltner bands produced a statistically significant edge relative to random entries and only for a subset of instruments (less than a quarter...in other cases, there were surplus returns over random entry, but not statistically significant). I did not see any kind of edge from buying at 50, 100, or 200 EMA or using MACD cross-overs relative to random entries that was statistically significant. Actually, for the MACD, I found surprisingly that buying when the signal line is below 0 is better than buying when it is above zero. Maybe that makes sense given the history. Buying the dip has worked well since 2008. Although pretty much everything produced a positive return because the market mostly went higher over that period so even random entry "worked".
For those that have luck with the moving averages, I'm curious as to how you use them. I'm guessing that they are mostly used as a filter for another signal.