I guess the point is that the indexing strategy you take is inconsequential if the queries and tables are poorly designed to begin with. If you normalize the table structures and do things like I wrote above, then the indexing strategy becomes much more clear: index first on those fields which will be most frequently accessed via WHERE clauses (and JOIN expressions), and consider different strategies for clustering based on how many queries are GROUPING on a certain set of data, or are doing RANGE queries (for instance, on your DATE_ID field?), etc.

--

Also, why aren't you using DATETIME fields?