We present some properties of the data from the recent mini flash crashes occurring in individual stocks of the Dow Jones Industrial Average. The top five are: 1) Gaussianity is absent in data; 2) the tail decay of the return distributions follow power laws; 3) chaos and logperiodicity cannot be dismissed at first; 4) chaos and logperiodicity are not good models for the data on second thoughts; and 5) a threshold GARCH fit can also describe the data well, but fails to detect the power law tail decay of most distributions of returns.