Matrix 90% of the time
I find that the matrix metering even in the D100 is really good. I will say that I didn't feel that way until I spent some time studying the picture in the manual of what areas matrix reads, and doing the same study in a couple magazine tests of the metering system. Once I understood where the zones were, how big they were, and how they were shaped, I figured out how to use exposure lock and reframing to compensate for those cases where matrix metering could get things horribly wrong without some help, but will get things wonderfully right with help.
I use spot the other 10%, but use it more like a zone metering system, taking four or five readings on different brightness areas and then manually feeding them to the camera. I don't shoot that many things where one small area is all that I care about, I'm just not that good of a birder.
But wanna know my secret weapon? A handheld incident light meter. I've got one old Sekonic, and one new fancy digital one, both of which I initially acquired to support my medium format work. Over the last couple years, I've discovered that a couple incident readings followed by manual setting of the d100 produce incredibly good results. Most pleased by its results in situations like snow, beaches, wide open skies, things where the in camera metering almost almost always gets it wrong.
There's no way to build an incident-reading inside the camera meter, but if there was a camera that pulled it off, I'd be there in a heartbeat. Someone told me once that all the fancy stuff matrix metering does is an attempt to produce the exposure settings that a reasonably well wielded incident meter would produce.