I generally only create baseline of alerts for 824 and 825, then 19-25. I have used Glenn Berry's script[/url] at times to create these for me.

Any other alerts outside of this are for specific purposes but I have not used any for performance reasons. Performance side of things I would suggest capturing that information over time (perf counters, wait stats, etc.) and then build your own procedures to alert off of that data. I do this for disk space and then databases with low log space.