Professor Lee from NUS sent me his feedback in email. I am posting it here for follow-up discussion.
Here is my laymanâ€™s view.
Moving averages filter off fine scale structures (or short term fluctuations) and reveal the long term trend, a process called smoothing. A longer moving average (one with longer period) is less sensitive to local fluctuations then a shorter one. In other words, a shorter moving average tracks the original data locally better than a longer one.
Given two periods, p1 and p2 (p1 < p2), we have two moving average signals SMA1 = SMA(p1) and SMA2 = SMA(p2). SMA2 is smoother that SMA1, while the later is more sensitive to the fluctuation of the original signal. I would think that if a “regional minimumâ€™â€™ (smallest local minimum in a region) is established for SMA1, then it is time to buy. On the other hand, if a “regional maximumâ€™â€™ is established for SMA1, then it is time to sell. But how do we know whether a regional minimum or maximum has been established. One way is to use SMA2: if SMA1 rises and cut SMA2 then a regional minimum had been established (to some probability or risk and here the choice of p1 , p2 will play a role), then buy; if SMA1 falls and cut SMA2 then a regional maximum had been established, then sell. Of course there are other ways to established regional minimum and maximum. I would also think that p1, p2 are chosen according to the traders investment horizon.
Weighted average will be useful if you have information that influence prices, e.g. quantitative easing, credit rating changes, political developments, and so forth.
I hope the views are useful. Let me know if you have other views.